Advances in Vehicular Networking Technologies

ADVANCES IN
VEHICULAR NETWORKING
TECHNOLOGIES
Edited by Miguel Almeida
Advances in Vehicular Networking Technologies
Edited by Miguel Almeida
Published by InTech
Janeza Trdine 9, 51000 Rijeka, Croatia
Copyright 2011 InTech
All chapters are Open Access articles distributed under the Creative Commons
Non Commercial Share Alike Attribution 3.0 license, which permits to copy,
distribute, transmit, and adapt the work in any medium, so long as the original
work is properly cited. After this work has been published by InTech, authors
have the right to republish it, in whole or part, in any publication of which they
are the author, and to make other personal use of the work. Any republication,
referencing or personal use of the work must explicitly identify the original source.
Statements and opinions expressed in the chapters are these of the individual contributors
and not necessarily those of the editors or publisher. No responsibility is accepted
for the accuracy of information contained in the published articles. The publisher
assumes no responsibility for any damage or injury to persons or property arising out
of the use of any materials, instructions, methods or ideas contained in the book.

Publishing Process Manager Katarina Lovrecic
Technical Editor Teodora Smiljanic
Cover Designer Martina Sirotic
Image Copyright Monkey Business Images, 2010.
Used under license from Shutterstock.com
First published March, 2011
Printed in India
A free online edition of this book is available at www.intechopen.com
Additional hard copies can be obtained from [email protected]
Advances in Vehicular Networking Technologies, Edited by Miguel Almeida
p. cm.
ISBN 978-953-307-241-8
free online editions of InTech
Books and Journals can be found at
www.intechopen.com
Part 1
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Chapter 9
Preface IX
Wireless Networks 1
Seamless Connectivity Techniques
in Vehicular Ad-hoc Networks 3
Anna Maria Vegni, Tiziano Inzerilli and Roberto Cusani
Asynchronous Cooperative Protocols
for Inter-vehicle Communications 29
Sarmad Sohaib and Daniel K. C. So
Efficient Information Dissemination in VANETs 45
Boto Bako and Michael Weber
Reference Measurement Platforms
for Localisation in Ground Transportation 65
Uwe Becker
Coupling Activity and Performance Management
with Mobility in Vehicular Networks 77
Miguel Almeida and Susana Sargento
Ultra-Wideband Automotive Radar 103
Akihiro Kajiwara
An Ultra-Wideband (UWB) Ad Hoc Sensor Network for Real-
time Indoor Localization of Emergency Responders 123
Anthony Lo, Alexander Yarovoy, Timothy Bauge,
Mark Russell, Dave Harmer and Birgit Kull
Hybrid Access Techniques
for Densely Populated Wireless Local Area Networks 149
J. Alonso-Zrate, C. Crespo, Ch. Verikoukis and L. Alonso
Hybrid Cooperation Techniques 165
Emilio Calvanese Strinati and Luc Maret
Contents
Contents VI
Adaptative Rate Issues in the WLAN Environment 187
Jerome Galtier
An Overview of DSA via Multi-Channel MAC Protocols 201
Rodrigo Soul de Castro, Philippe Godlewski and Philippe Martins
Distance Estimation based on 802.11 RTS/CTS
Mechanism for Indoor Localization 217
Alfonso Bahillo, Patricia Fernndez, Javier Prieto, Santiago
Mazuelas, Rubn M. Lorenzo and Evaristo J. Abril
Data Forwarding in Wireless Relay Networks 237
Tzu-Ming Lin, Wen-Tsuen Chen and Shiao-Li Tsao
Experiments of In-Vehicle Power Line Communications 255
Fabienne Nouvel, Philippe Tanguy, S. Pillement and H.M. Pham
Kinesthetic Cues that Lead the Way 279
Tomohiro Amemiya
Transmission Technologies and Propagation 295
Technological Trends of Antennas in Cars 297
John R. Ojha, Ren Marklein and Ian Widjaja
Link Layer Coding for
DVB-S2 Interactive Satellite Services to Trains 313
Ho-Jin Lee, Pansoo Kim, Balazs Matuz, Gianluigi Liva,
Cristina Parraga Niebla, Nuria Riera Daz and Sandro Scalise
Mobility Aspects of Physical Layer
in Future Generation Wireless Networks 323
Asad Mehmood and Abbas Mohammed
Verifying 3G License Coverage Requirements 339
Claes Beckman
Inter-cell Interference Mitigation
for Mobile Communication System 357
Xiaodong Xu, Hui Zhang and Qiang Wang
Novel Co-Channel Interference Signalling
for User Scheduling in Cellular SDMA-TDD Networks 389
Rami Abu-alhiga and Harald Haas
Demodulation Reference Signal Design
and Channel Estimation for LTE-Advanced Uplink 417
Xiaolin Hou and Hidetoshi Kayama
Chapter 10
Chapter 11
Chapter 12
Chapter 13
Chapter 14
Chapter 15
Part 2
Chapter 16
Chapter 17
Chapter 18
Chapter 19
Chapter 20
Chapter 21
Chapter 22
Preface
Its fair to say that the array of commercially available vehicles is beginning to catch
up with the technological advances made available by science during the past decade.
With a lot of eort being employed by manufactures to provide cars with advanced
networking capabilities, we can mainly distinguish the connectivity topologies in two
very dierent groups: the Vehicular Ad-hoc Networks (VANETs), which make use of
the scientic contributions provided by the Mobile Ad-Hoc Networks (MANETs), and
the Vehicle Infrastructure Integration (VII) based networks. VII or Vehicle-to-Infra-
structure (V2I) deserved special atention during the last couple of years, in part, due
to the increase of interest on Cloud based communications, but also given the current
highlight over the paradigm shif towards the Internet of Things. Nevertheless, the
scope of technological challenges that have an immediate impact on the design and
performance of such specic networks is extremely wide and particularly diverse.
This book provides an insight on both, the challenges and the technological solutions of
several approaches, which allow connecting vehicles between each other and with the
network. It underlines the trends on networking capabilities and their issues, further
focusing on the MAC and Physical layer challenges. Mobile oriented technologies set
up the basic requirements for high mobility scenarios. Having this in mind, particular
atention was paid to the propagation issues and channel characterization models.
We tried to cover a vast multitude of topics, which reect the current state of the art
concerning Vehicular Networking Technologies, some of which include dealing with
connectivity issues, networking topologies (VANETs, VII/V2I), MAC solutions, data
forwarding, network/vehicle performance management, link layer coding techniques,
mobile/radio oriented technologies, channel characterization and channel coding
amongst others.
We are thankful to all of those who contributed to this book and who made it possible.
We hope others can enjoy it as much as we do.
Miguel Almeida
University of Aveiro
Portugal
Part 1
Wireless Networks
1
Seamless Connectivity Techniques
in Vehicular Ad-hoc Networks
Anna Maria Vegni
1
, Tiziano Inzerilli
2
and Roberto Cusani
2

1
University of Roma Tre, Department of Applied Electronics, Rome,
2
University of Rome Sapienza, Department of Information Engineering,
Electronics and Telecommunications, Rome,
Italy
1. Introduction
Emerging Vehicular Ad-hoc NETworks (VANETs) are representing the preferred network
design for Intelligent Transportation Systems (ITS), mainly based on Dedicated Short-Range
Communications (DSRC) for Vehicle-to-Vehicle (V2V) communications (Held, 2007). Future
vehicles will be fully networked, equipped with on-board computers with multiple Network
Interface Cards (NICs) (e.g., Wi-Fi, HSDPA, GPS), and emerging wireless technologies (e.g.,
IEEE 802.11p, WiMax, LTE).
Although V2V is potentially the most viable approach to low-latency short-range vehicular
networks, connectivity in VANETs is often not available due to quick topology network
changes, random vehicle speed, and traffic density (i.e. sparse, dense, and totally
disconnected neighbourhoods) (Chiara et al., 2009). As an alternative, longer-range vehicular
connectivity are supported by a Vehicle-to-Infrastructure (V2I) protocol (Held, 2007), which
exploits a pre-existing network infrastructure, for communications between vehicles and
wireless/cellular access points (referred to Road Side Units, RSUs). As a further benefit, V2I
protocols allow access to the Internet and delivery of traditional applications in addition to
dedicated applications for ITS, thus making vehicle communications more versatile.
Both the paradigms V2V and V2I exhibit connectivity problems. Different speed and
traffic densities result in low vehicular contact rate, and limit communications via V2V
protocol, while V2I communications are reduced, especially in highway scenarios, by the
low number of RSUs displaced on the roads. Moreover, V2I limitations are due to particular
vehicular applications, and performance is also strictly dependent on the specific wireless
technology for the RSUs. It can then become advantageous to adopt hybrid schemes
combining V2V and V2I communication into a single protocol and allowing fast migration
from V2I to V2V connectivity depending on the operation context (high/low density,
high/low velocity). Such mechanism is the so-called Vertical Handover (Pollini, 1996).
In this chapter we shall describe the traditional techniques Vertical Handover
algorithms used for seamless connectivity in heterogeneous wireless network
environments, and in particular adopt them in VANETs, where V2V and V2I represent the
main communication protocols. Section 2 deals with the basic features of Vertical Handover
(VHO) in the general context of a hybrid wireless network environment, and it discusses
how decision metrics can affect handover performance (i.e. number of handover

4
occurrences, and throughput). Instead, Section 3 briefly introduces two proposed techniques
achieving seamless connectivity in VANETs. The first technique is a vertical handover
mechanism applied to V2I-only communication environments; it is presented in Section 4
via an analytical model, and main simulated results are shown in Subsection 4.1. The second
approach is described in Section 5. It addresses a hybrid vehicular communication protocol
(i.e. called as Vehicle-to-X) performing handover between V2V and V2I communications,
and vice versa. Subsection 5 illustrates how messages can be propagating via V2X, while
Subsection 6 shows the main phases of V2X algorithm and the simulation results. Finally,
we conclude this chapter in Section 6.
2. Vertical Handover mechanism
Next generation wireless networks adopt a heterogeneous broadband technology model in
order to guarantee seamless connectivity in mobile communications. Different network
characteristics are basically expected for different multimedia applications, and ubiquitous
access through a single network technology is not always guaranteed because of limitation
of geographical coverage. Moreover, since mobile applications require Quality-of-Service
(QoS) continuity in a ubiquitous fashion, cooperation of access networks in heterogeneous
environments is an important feature to assure.
A Vertical Handover (VHO) is a process preserving users connectivity onthemove, and
following changes of network (Pollini, 1996). In this context, VHO techniques can be applied
when network switching is needed (i) to preserve host connectivity, (ii) optimize QoS as
perceived by the end user, and (iii) limit the number of unnecessary vertical handover
occurrences (i.e. the well-known ping-pong effect) (Inzerilli & Vegni, 2008).
VHO schemes can be classified on the basis of the criteria and parameters adopted for
initiating a handover from a Serving Network (SN) to a new Candidate Network (CN).
Namely, we can enlist the following main metrics whose monitoring can drive handover
decisions:
Received Signal Strength (RSS)-based VHO algorithms: when measured RSS drops below
receiver sensitivity it denotes lack of connectivity which requires necessarily a VHO
(Ayyappan & Dananjayan, 2008); (Inzerilli & Vegni, 2008);
Signal-to-Noise and Interference ratio (SINR)-based VHO algorithms: SINR directly
impacts achievable goodput in a wireless access network. A modulation scheme can
sometimes adapt transmission rate a channel coding scheme to measured SINR (Yang et
al., 2007); (Vegni et al., 2009);
Multi-parameter QoS-based VHO algorithms: VHO algorithms can be based on the
overall quality assessment for the available networks obtained balancing various
parameters (Vegni et al., 2007); (Jesus et al., 2007);
Location-based VHO algorithms: they estimate network QoS on the basis of the MT
location relatively to the serving access point (Kibria et al., 2005); (Wang et al. 2001);
(Kim et al., 2007).
In an RSS-based VHO approach, when the measured RSS of the SN drops below a
predefined threshold, the RSS of the monitored set of CNs is evaluated in order to select the
best network to migrate to. The authors in (Ayyappan & Dananjayan, 2008) adopt this basic
approach for VHO and evaluate the performance of a vertical handover mechanism that is
based on the RSS measurements. This approach represents the traditional handover and
simplest mechanism, which however does not aim to optimize communication performance.
Seamless Connectivity Techniques in Vehicular Ad-hoc Networks

5
Differently, the SINR-based approach (Yang et al., 2007) compares the received power
against the noise and the interference levels in order to obtain a more accurate performance
assessment, which brings about a slight increase of computational cost. SINR factor is
considered for VHO decisions, as it directly affects the maximum data rate compatible with
a given Bit Error Rate (BER). Therefore, when the SINR of the serving network decreases,
the data rate and the QoS level decrease too. As a consequence, a SINR-based VHO (Yang et
al., 2007) approach is more suitable to meet QoS requirements, and can be used to
implement an adaptive data rate procedure. RSS-based and SINR-based schemes are both
reactive approaches, which means that they aim to compensate for performance degradation
when this occurs, that is whenever either the RSS or the SINR drops below a guard
threshold. Moreover, we expect that a combination of different VHO decision metrics (i.e.
location information, RSS/SINR measurements, QoS requirements or monetary cost) can
generate most effective and correct VHO decisions (Vegni et al., 2009).
A Multi-parameter QoS-based VHO scheme has been illustrated in (Vegni, et al. 2007),
which is instead representative of a proactive approach performing regular assessment of
the QoS level offered by the current SN, as well as by other CNs. The proposed method
attempts to select the best CN at any time thus preventing performance degradation, and
sudden lack of connectivity. It can be based on the simultaneous estimation of a set of
parameters such as RSS, throughput and BER and in the subsequent evaluation of an
objective QoS metric, which is a function of such parameters. Its effectiveness is directly
dependent on the ability of the objective QoS metric to mimic subjective Quality-of-
Experience of the end-users, and on the accuracy of the assessment of the parameters on
which the metric is based. QoS-based VHO is well suited for multimedia applications like
real-time video streaming. As a drawback, preventive approaches may lead to high
handover frequency and hence lead to algorithmic instability, i.e. the so-called ping-pong
effect. A hysteresis cycle or a hard limitation in maximum handover frequency in the VHO
algorithm can help reducing this phenomenon (Kim et al., 2007).
In location-based VHO solutions, the knowledge of location information is exploited to
assess the quality of the link between SN and the MT, and to predict its future evolution to
some extent on the basis of the MT estimated path. User position can be determined in
several ways (Kibria et al., 2005), including Time of Arrival, Direction of Arrival, RSS, and
assisted GPS (Global Positioning System) techniques. Examples of location-based VHO are
discussed in (Wang et al. 2001), though the proposed technique shows a computational
complexity of the handover decision that is rather high, and establishing and updating a
lookup table to support a handover margin decision turns out to be time-consuming. In
(Kibria et al., 2005) the authors develop a predictive framework based on the assumption
that the random nature of user mobility implies an uncertainty on his/her future location,
increasing with the extension of the prediction interval.
Location-based VHO solutions are the most commonly-used techniques in the VANET
context, where high-mobility of nodes makes it difficult to promptly react to performance
degradation purely basing on RSS measurements.
3. Seamless connectivity in vehicular ad hoc networks
In this section we are introducing two different approaches for seamless connectivity in
VANET scenarios.
The first approach is a Vertical Handover technique based on vehicle speed, where a vehicle
switches from a Serving Network (SN) to a Candidate Network (CN) if its speed is lower

6
than a fixed threshold (Vegni & Esposito, 2010); (Esposito et al., 2010). Such technique i.e.,
called as Speed-based Vertical Handover (SVHO) mainly addresses necessary
connectivity switching in vehicular environments for real-time applications, which require
high QoS levels and seamless connectivity. SVHO does not consider traditional vehicular
protocol (i.e. V2V and V2I), but it simply guarantees a seamless connectivity when vehicles
are crossing an area with overlapping wireless networks.
The problem of a seamless connectivity becomes more challenging in VANETs, because
vehicles move across overlapping heterogeneous wireless cells environments. In such
scenarios, frequent and not always necessary switches from a SN to a CN may occur, often
degrading network performance. Vertical Handover techniques are able to fix seamless
connectivity according to a well-defined decision criteria (i.e. the type of RSU technology,
the RSS indicator, QoS metrics, and so forth (McNair & Fang, 2004)).
Traditional VHO decision metrics cannot be applied in vehicular environments, and may
fail due to the speed and the time that the vehicle is going to spend in a wireless network.
The problem of a seamless connectivity becomes even more challenging as vehicles move at
high speed across overlapping heterogeneous wireless cells environments.
Therefore, in VANETs handovers should be performed on the basis of specific factors, such
as vehicle mobility pattern, and locality information, rather than standalone QoS
requirements. Past solutions have partially but not fully considered these aspects. In (Chen
et al., 2009) the authors deal with a novel network mobility protocol for VANETs, to reduce
both handover delay and packet loss rate. In (Yan et al., 2008) a vertical handover technique
focuses on an adaptive handover mechanism between WLAN and UMTS, based on the
evaluation of a handover probability, obtained from power measurements. In this case, the
handover decision is taken by comparing the handover probability with a fixed probability
threshold, which depends on the vehicle speed and on handover latency.
The second technique is a hybrid vehicular communication protocol, called as Vehicle-to-X
(V2X), which achieves the advantages of both traditional V2V and V2I protocols (Vegni &
Little, 2010). V2X supports VANET scenarios with a heterogeneous network environment,
and aims for vehicles (i) to communicate between them (via V2V), and (ii) to connect to the
Internet (via V2I). V2X permits hybrid vehicular communications, and each vehicle can
switch from V2V to V2I, and vice versa, on the basis of a protocol switching decision metric.
We will illustrate the behaviour of V2X protocol, and analyze how information is
propagated in VANETs with heterogeneous network infrastructure nearby. In the simulated
scenarios, a data push communication model has been assumed, in which information
messages are propagated via localized (limited range) broadcast. Effectiveness of V2X will
be validated via a performance comparison in terms of message dissemination with
traditional opportunistic networking technique (i.e., V2V).
V2X results in a novel opportunistic forwarding technique that is the main approach to
achieve connectivity between vehicles, and to disseminate information. In traditional
opportunistic networking V2V communications exploit connectivity from other
neighbouring vehicles by a bridging technique, where message propagation occurs through
connectivity links which are built dynamically (Agrawal & Little, 2008). Each vehicle acts
as next hop and subsequent hops form a path from a source vehicle to a destination
vehicle. Because V2X does not rely only on V2V but also exploits the potentiality of V2I, it
should be considered as an opportunistic forwarding technique, where messages are
propagated along dynamically generated paths whose links are vehicles, as well as road
side units.

7
4. Speed-based VHO technique
A Speed-based VHO technique (S-VHO) is now described in each single step of VHO
decision (Vegni & Esposito, 2010). We recall that, as all the VHO techniques are mainly
focused on maximization of network performance, and limitation of vertical handover
occurrences, even for S-VHO the aim is to both minimize VHO frequencythe number of
executed VHOs and maximize the throughput measured at the vehicle.
Differently from traditional RSS-based vertical handover approaches (Inzerilli & Vegni,
2008), the proposed S-VHO method does not consider any signal strength parameters: such
information might be out of date, unreliable and its variance may fluctuate significantly,
especially in VANET scenarios, causing unnecessary and unwanted vertical handovers, as
well as throughput degradation. With respect to traditional VHO algorithms, S-VHO takes
a vertical handover decision on the basis of the estimation of the cell crossing time
parameter, which represents the time spent in crossing a wireless cell by the vehicle. No
RSS indicator is considered in such approach, and S-VHO decides to switch from a SN to a
CN on the basis of throughput experienced at the vehicle receiver, and the time spent by
the vehicle in a CN.
It is noticeable that S-VHO method approaches to the technique proposed in (Yan et al.,
2008), except that S-VHO focuses on a vehicle-controlled VHO, due to smart on-board
computer equipped with GPS connectivity, and handover decisions are based on both
vehicle speed, and handover latency (Vegni & Esposito, 2010); (Esposito et al., 2010).
Moreover, S-VHO differs from previous vertical handover techniques because its usefulness
is extended to real time applications for vehicular scenarios (i.e., video streaming, on-line
gaming, Internet browsing, and so on).
We shall describe an analytical framework of S-VHO technique. First, we depict how
throughput in both serving and candidate networks can be estimated; then we show
analytically how a vertical handover decision is taken by the vehicle.
Figure 1 illustrates the vehicular scenario we are dealing with. We assume that a vehicle is
driving at constant speed v [m/s], following a typical Manhattan mobility model, suitable
for a urban area and where the vehicles trajectory is constrained by a road grid topology,
with straight paths (i.e. a highway composed of straight lanes). Each vehicle is equipped
with a GPS receiver, and then the vehicles location is continuously updated and tracked.
Let us denote as t
in
and t
out
the time instants when a vehicle enters and exits a wireless cell
(i.e. an UMTS network), respectively. The distance the vehicle will cross inside the wireless
cell during the time interval T = t
out
t
in
is:

v
x
T
. =
(1)
From (1) we can introduce the Cell Crossing Time [s] parameter, according to the following
assessment, such as:
Definition (Cell Crossing Time). Given a vehicle V, traversing an area covered by a wireless cell
C at constant speed v
, the cell crossing time of V in C, denoted as T [s], is the overall time that V
can spend under Cs coverage.
According to Figure 1, the cell crossing time lasts since the vehicle enters the wireless cell in
P
in
at t
in
, and then exits the wireless cell in P
out
at t
out
, respectively. During T interval, the
vehicle is crossing the x distance. Assuming an omni-directional radio coverage C for the
wireless network, and denoting (i) R
+
\{0} as the radius of the wireless cell, (ii) [0, ]

8
as the angle between the vehicles line-of-sight with the RSU and the direction of the vehicle,
we can use classic Euclidean geometry to obtain
x R 2 cos . = (2)

Fig. 1. VANET scenario where a vehicle moves at constant speed v, following a Manhattan
mobility model. P
in
and P
out
are the wireless cell entrance and exit points, respectively
Since we have assumed that each vehicle is equipped with a GPS receiver, the coordinates
P
in
(x
in
; y
in
) of the entrance, and P
out
(x
out
; y
out
) of the exit point of the wireless cell, with
respect to a coordinate system centered in the cell centre, are easily calculated so that x is
known. It follows that the cell crossing time can be expressed as

out in
out in
y y x R
T
v v x x
2
cos arctan .

= =

(3)
Notice that the computation in (3) is directly performed by the vehicle by assuming constant
vehicle speed, and knowledge of the wireless cell radio coverage.

Fig. 2. VANET scenario with heterogeneous overlapping wireless networks
After defining geometric parameters, we can introduce the throughput (t) [Bits],
evaluated for t = T [s], as:

SN
t B T ( ) , = (4)

9
where B
SN
[Bit/s] is the bandwidth of the actual wireless cell (i.e. Serving Network, SN),
assumed to be constant during T. Equation (4) defines the throughput that a vehicle
would experience by remaining connected with the actual wireless cell, during the cell
crossing time T.
Note that so far we have only modelled the throughput in a VANET network where a
vehicle is crossing a single wireless network, and no overlapping cells are considered. We
now extend the model to heterogeneous environments by capturing vertical handovers
across different access networks as well. In this case, a vehicle is entering an area where two
or more different wireless cells are co-existing and overlapped.
Figure 2 shows this vehicular scenario with heterogeneous overlapping wireless networks
(i.e. UMTS, and WLAN).
Namely, the vehicle drawn in Figure 2 is actually connected to an UMTS network, and is
entering a WLAN network in P
in
at t
in
. Then, UMTS represents a Serving Network (SN)
while WLAN a Candidate Network (CN), respectively.
Since S-VHO aims to maximize throughput, system initiates a handover from the SN (i.e.
UMTS) to the CN (i.e. WLAN), if and only if the estimated throughput measured in the CN
is higher than the throughput in the current SN, such as:

CN SN
. (5)
Let a vehicle be connected to a SN, entering the wireless range of a Candidate Network
(CN). In this heterogeneous scenario, we model the data delivered between the two time
instants t
in
and t
out
, as a positive range function :
+
defined as
( )( ) ( )
CN SN
B T L B T 1 , = + (6)
where is an indicator function, such that = 1 when a vertical handover is executed, and
zero otherwise. Note that when = 1, is equivalent to the throughput obtained in the CN,
while for = 0, is the throughput in the SN. The parameter
+
is a hysteresis factor
introduced to avoid vertical handover occurrence when the two competing networks have
negligible bandwidth difference.
Function in (6) captures the data loss due to the vertical handover latency L [s], that is the
time interval during which a vehicle, traversing an area covered by at least two wireless
cells, does not receive any data due to control plane (socket switching) signalling messages
exchange.
Since during the cell crossing time of a vehicle it is desirable to maximize throughput, S-
VHO technique initiates a handover only when it represents a valid handover, that is when
( ) ( ) 1 0 = > = , (7)
from which it follows:

SN
CN
B
B
L
T
1
> +
, (8)
The inequality in (8) shows that switching decisions may not be necessary even though the
bandwidth B
CN
is higher than B
SN
. Switching becomes necessary only if the time that the

10
vehicle will spend in the wireless cell with higher bandwidth is long enough to compensate
for the data loss due to the switching overhead, namely only if

L
L T
T
1 0 . > <
(9)
This observation leads to the conclusion that the throughput is influenced not only by the
bandwidth of the considered technologies, but by a larger set of parameters, such as (i) the
cell crossing time, (ii) the vehicle speed, and (iii) the overhead of the control-plane protocols
adopted (handover latency). This consideration represents not only a novel definition for
throughput by three previous parameters (i.e. cell crossing time, vehicle speed, and
handover latency), but also gives an important result for vertical handover management and
connectivity switching decisions in vehicular networks. By monitoring and limiting vehicle
speed to a fixed upper bound, it is always possible to perform a valid handover. This result
represents the following Theorem of Speed Upper Bound for valid VHO decisions (Esposito et
al., 2010).
Theorem (Speed Upper Bound). Given a vehicle V, travelling with an average constraint speed v

in a heterogeneous vehicular environment for a distance x, a valid handover for V occurs if v
is
bounded as follows:

( )
( )
CN SN
CN
x B B
v
B L
.

<
(10)
Proof: the claim follows from (8), where we highlighted the term T, such as

( )
( )( )
( ) ( )
( )
( )
CN SN
CN SN
CN SN CN
CN
CN SN
L
B B
T
B T L B T
B B T B L
B L
T
B B
1 ,
,
,
.

>

>
>
>

(11)
By replacing the term T from (11) in the following definition of average vehicle speed, i.e.
v x T / , =

(12)
we obtain the result expressed in (10).
The Theorem of Speed Upper Bound is useful not only for VHO management, but also in
designing V2I protocols, as well as to promote vehicle safety applications. Network
providers may in fact offer lower data rate in those areas where the speed limit is lower, and
so induce vehicles to maintain lower speeds, in order to experience acceptable QoS levels
low jitter and high throughput throughout valid handovers. Users should be more
motivated to respect speed limits, rewarded by QoS enhancement.
S-VHO technique is based on Theorem 1 in order to manage valid handovers for fast users
driving in an heterogeneous vehicular network environment. This approach is acted by the
vehicle itself each time is crossing a wireless network and needs to be connected with it. S-

11
VHO rules according to the following algorithm drawn in Figure 3. S-VHO accepts three
inputs, such as (i) the vehicle speed v
, (ii) the ingress time t

in
of the vehicle into a wireless
cell, and (iii) the GPS location information P
in
, respectively. The algorithm then returns the
handover decision variable {0, 1}, whose value means if a vertical handover is executed
or not.
Let a vehicle be connected to a SN, driving in an area with heterogeneous overlapping
wireless networks (i.e. UMTS, and WLAN). Due to multi-network interface cards equipping
the vehicle, it is able to recognize one or more available CNs to access. A vertical handover
can occur whenever the inequality in (8) is verified.
After each handover execution, the algorithm enters in idle mode for an inter-switch waiting
time period (i.e. T
w
[s]), that means no vertical decision is taken. For example, if a vehicle
travels at 15 m/s, a 10 seconds inter-switch waiting time results in 150 meters covered by
the vehicle, before the algorithm is reactivated. The idle mode approach is a well-know
solution for limitation of vertical handover frequency (Inzerilli & Vegni, 2008); (Inzerilli et
al., 2008).

( )
in in
SN
CN
, ,
,
1
evaluate
*
1 VHO executed
set a decreasing counter to [ ].
0
idle mo
w
w
v t P
inside area with at least two overlapped cells
B
B
L
T
T
T T
T s
T
Input :
Output :
while do
if
then
if then
while do

> +

>
>
( )
de
0 no VHO executed
end
else
end
end
end

Fig. 3. Speed-based Vertical Handover Algorithm
Many handover algorithms incorporate a hysteresis cycle within handover decisions, in
order to prevent a mobile terminal moving along the boundary of a wireless cell to trigger
handover attempts continuously. This phenomenon is well known in the literature as
ping-pong effect (Kim et al., 2007), and hysteresis is largely adopted in practical
implementations.

12
A high number of vertical handover executions can lead to excessive network resource
consumption and also affects mobile terminals performance (i.e. battery life, and energy
consumption).
Ping-pong effect occurs in vehicular environments, specially when vehicles travel on a
border line between two wireless cells, and make frequent, often unnecessary or unwanted
handovers. Limitation of handover frequency is acted by imposing a minimum interval of
time between two consecutive handovers. Namely, the grater is the waiting time the smaller
will be the number of vertical handovers.
S-VHO algorithm first measures the data rate from CN (i.e. B
CN
), and then if inequality (8)
holds, the cell crossing time T is computed as shown in (3). After the cell crossing time
evaluation, S-VHO decides whether the handover would be valid or not by comparing T
with the threshold for valid handover, (i.e. T*).
We can express the threshold for valid handover as

x
T
v
* .
*
= (13)
The term v* is the speed upper bound, obtained from the inequality (10) and formally
expressed as

( )
( )
CN SN
CN
x B B
v
B L
* ,
(14)
where is a positive quantity (i.e. 0). Equation (13) becomes

( )
CN
CN SN
B L
T
B B
* .

(15)
The Theorem of Speed Upper Bound shows that the vehicles speed is strictly limited by
handover latency L, and the hysteresis factor . The impact of both two parameters and the
effects on the speed upper bound are described by the following analytical results.
In Figure 4 (a) we show the impact of the handover latency L [s] on the speed upper bound,
for a given bandwidth ratio of two available wireless technologies (e.g. UMTS, and WLAN
1
).
The hysteresis factor was set to zero to isolate only the impact of handover latency, and the
range of speed was bounded by 35 m/s (typical highway speed limit). As we can notice, for
higher values of handover latency, the speed upper bound (i.e. the maximum speed at
which vehicles experience valid handovers) decreases, and approaches to zero:

L
v lim * 0.
= (16)
This result is reasonable since vehicles travelling at higher speed may not spend enough
time under higher data-rate wireless networks to justify the degraded performance
introduced by the handover overhead of the signalling messages.
For a bandwidth ratio equal to one (i.e. B
CN
= B
SN
), the speed bound is null as the hysteresis
factor has been set to zero; in general (i.e. for 0), the speed bound is still approaching to
zero for L approaching to infinity:

1
The bandwidth ranges were chosen according to WLAN, and UMTS (Laiho et al., 2005) requirements.

13

( )
L L
CN
x
v
B L
lim * lim

. (17)
The second result comes by observing the epigraph and hypograph the set of points
above, and below the drawn curves, respectively. Any point belonging to the epigraph
represents no performance gain in initiating handovers, even if the CN has higher
bandwidth than the SN. In contrast, for any point in the hypograph valid handovers occur.
We can notice that the curve with zero bandwidth gap (i.e. B
CN
= B
SN
) has empty hypograph.
This follows directly from the definition of valid handover, that means a vertical handover
cannot be valid when the data rates from CN and SN are equal.

(a) (b)
Fig. 4. Speed Upper Bound behaviour. Impact of (a) handover latency L, and (b) hysteresis
In Figure 4 (b) we show the impact of the hysteresis [Mbps] on the speed bound, for
different values of the handover latency (i.e. L = {0.01, 0.02, 0.04, 0.08} [s]). The hysteresis
range is [0, B
CN
B
SN
]. It is useful to note that B
CN
B
SN
is the maximum value of after
which no valid handover would occur. We have simulated the case B
CN
B
SN
= 16 [Mbps],
which represents a typical gap in data rates between UMTS and WLAN.

Fig. 5. Simulated VANET scenario with overlapping WLAN and UMTS cells

14
Notice how the hypographic area changes for different values of L: the lower the handover
latency, the greater the hypograph. For example, when the handover latency increases (i.e.
L = 0.08 [s]), the hypographic area significantly reduces. From this observation, it follows
that handover latency should be taken into account when designing protocols for seamless
connectivity in VANETs, and not only focusing on physical parameters or vehicle speed.
4.1 S-VHO simulation results
In this section we describe some simulation results for S-VHO mainly expressed in terms of
throughput maximization, and limitation of ping-pong effect. They will show how S-VHO
works for valid vertical handovers in VANETs.
Network performance, i.e. throughput, delay, and jitter, as well as the number of vertical
handovers, have been obtained with an event-driven simulator. More details of the
simulator can be found in (Vegni, 2010). The simulation scenario depicts a vehicle entering
from a random location, restricted to travel along a grid of streets and intersections. The
vehicle follows a random path inside a grid, according to the Manhattan mobility model, as
shown in Figure 5. The event-driven simulator generates different scenarios, whose
characteristics are similar to previous heterogeneous scenarios described in (Vegni &
Esposito, 2010). Figure 5 depicts one of the simulated scenarios, in terms of data rate
distribution from a set of three UMTS base stations, and twenty WLAN access points,
modelling a 2 km
2
area.
The location of each wireless cell has been generated uniformly at random, and a vehicle
moves in this area with speed in the range [5, 35] [m/s], capturing urban environment as
well as highway scenarios. During its journey, a vehicle requires to download a series of
video frames. For example, we could figure out that passengers are enjoying their travelling
time by means of real time applications, e.g. video streaming and online gaming.
For the cell setup, we have considered typical values of WLAN and UMTS radio
parameters, as described in (Vegni & Esposito, 2010). The cell radius was set to 120 [m] for
IEEE 802.11/a outdoor environment, and 600 [m] for an UMTS microcell; the transmitted
power in the middle of UMTS and WLAN cell has been chosen to be between 43 and
30 dBm, respectively. Finally, the average vertical handover latencies from UMTS to WLAN,
and from WLAN to UMTS, have been set respectively to L
UW
= 2 [s], and L
WU
= 3 [s],
while the data rate of UMTS and WLAN equal to 384 [kbps], and 11 [Mbps], respectively.
S-VHO network performance have been validated in terms of (i) throughput, (ii) jitter, and
(iii) handover frequency. Moreover, a comparison between S-VHO and a previous vertical
handover technique (Yan et al., 2008), based on traditional power measurements, has been
carried out. In both algorithms, the speed of the vehicle is used as handover assessment
criterion. Hereafter we will name the technique described in (Yan et al., 2008) as Speed
Probability Based VHO (SPB). Simulated results
2
will prove the effectiveness of S-VHO for
valid handovers.
Figure 6 shows the throughput expressed as cumulative received bits in a downlink
connection for both S-VHO and SPB techniques, versus the inter-switch waiting time in the
range [0, 50] [s]. The cumulative received bits represent the amount of received bits by a
vehicle moving inside an heterogeneous wireless environment for the simulated period of
time.

2
Results obtained by averaging 100 heterogeneous network scenarios.

15
The throughput performance of S-VHO outperforms at every speed the SPB algorithm. Our
performance improvement are justified by the absence of received signal strength
dependence in S-VHO handover assessment criteria. Moreover, the effectiveness of S-VHO
is clear when vehicle speed is below a given limit (e.g., 20 [m/s]).
On the other hand, SPB does not appear sensitive to either speed or inter-switch waiting
time, and its throughput is limited. However, the S-VHO throughput drops when the
vehicle speed exceeds the desired limit. This is justified by the Theorem of Speed Upper
Bound for valid VHOs.

Fig. 6. Comparison of throughput between S-VHO (white markers), and SPB (black markers)

Fig. 7. Comparison of (left) delay, and (right) number of vertical handover occurrences,
between S-VHO (white markers), and SPB (black markers) algorithms
The average frame delay for both S-VHO and SPB is shown in Figure 7 (left). The 95%
confidence intervals express how the average frame delay increases for higher speeds. This
is because there is not enough time to download the next frame before the signal from the
SN gets too weak. Moreover, S-VHO experiences lower delays compared to SPB, since on
average it performs less handovers.

16
Figure 7 (right) depicts, with 95% confidence intervals, the average number of handovers for
different values of inter-switch waiting time. As expected, the number of vertical handovers
decreases when the system is idle for longer periods (i.e. when T
w
increases). Since our
simulations count all the handovers (valid or invalid), the gap between the S-VHO and SPB
curves represents the number of invalid handovers, executed not taking into account the
handover latency L.

(a) (b) (c)
Fig. 8. Cumulative jitter experienced by a vehicle, for different speeds (a) 15 [m/s] v =
, (b)
25 [m/s] v =
, and (c) 35 [m/s] v =

Jitter performance from S-VHO and SPB have been compared, and are shown in Figure 8 for
different values of speed, and a fixed inter-switch waiting time value (i.e. T
w
= 10 [s]). Each
point represents the cumulative jitter, defined as the difference between maximum and
minimum frame delay. Jitter increases with speed note the scale difference among
different graphs since two frames may be more often coming from different wireless
networks, and also because the cell crossing time decreases when the speed increases. For S-
VHO higher speed implies higher jitter, unless the number of unnecessary handovers is
reduced.
To summarize, S-VHO technique distinguishes from previous approaches by using both
handover latency and cell crossing time estimation to simultaneously improve throughput
and delay. It is driven by the vehicle speed, so that vehicles are required to maintain a given
speed limit to maintain acceptable levels of throughput, delay and jitter. VHOs are limited
to effective and necessary connectivity switching. A handover towards a CN with higher
data rate does not necessary result in a throughput improvement.
S-VHO represents not only a novel VANET protocols for real-time applications, but also it
could be useful for urban safety. As a matter, service providers assisted by vehicular
networks could enforce speed limits and safety, while delivering real-time services as video-
streaming or online gaming.
5. Hybrid vehicular communication protocols
As known (Held, 2007), one of the main characteristics of VANETs is that vehicles move in
clusters interconnected blocks of vehicles due to different traffic scenarios (i.e. dense,
sparse or totally disconnected traffic neighborhoods). For this reason, vehicular connectivity
is not always available, and messages propagation in VANETs is still an open issue.

17
Many authors have addressed how to improve message forwarding by opportunistic
techniques, and multi-hop approaches. In (Resta et al., 2007) a multi-hop emergency
message dissemination is described by means of a probabilistic approach. The authors
derive lower bounds on the probability that a vehicle correctly receives a message within a
fixed time interval. Similarly, in (Jiang et al., 2008) an efficient alarm message broadcast
routing protocol is presented. This technique estimates the receipt probability of alarm
messages sent to vehicles.
In other works the message propagation model is based on the main VANET
characteristics such as number of hops, vehicle position, mobility, etc. In (Yousefi et al.,
2007) the authors consider a single-hop dissemination protocol based on Quality-of-
Service metrics, while in (Chen et al., 2008) a robust message dissemination technique is
illustrated, based on the vehicles position. Finally, the authors in (Nadeem et al., 2006)
present a data dissemination model based on bidirectional mobility of paths between a
couple for vehicles.
Common aspect in all previous works is that data traffic is disseminated only through
vehicles communicating via V2V; no network infrastructure nor V2I protocol have been
considered.
The use of the vehicular grid together with a network infrastructure has been discussed in
(Gerla et al., 2006); (Marfia et al., 2007). The benefits of using the opportunistic infrastructure
placed on the roads result in an enhancement of message propagation. In fact, a Road-Side
Unit represents a fixed node which is able to forward message information to vehicles
driving inside a wireless cell.
V2X relies on the network scenario depicted in (Gerla et al., 2006), but it represents a novel
protocol providing switching from V2V to V2I, and vice versa. It enables vehicles to
communicate via V2V or V2I on the basis of a protocol switching decision. The message
propagation via V2X is then improved by a correct use of vehicular communication
protocols (i.e., V2V and V2I).
The heterogeneous vehicular scenario we are referring to is depicted in Figure 9. Let us
consider a cluster C comprised of a set S of vehicles (i.e., S n {1, 2, , } = ). Then, m Road-Side
Units (RSUs) (i.e., m < n) are displaced in the network scenario. Each vehicle is able to
communicate via V2V on the basis of a fixed transmission range radio model (Vegni &
Little, 2010). We assume that only a limited subset of vehicles in the cluster C, (i.e.,
S l S {1, 2, , } = , with l < n), is able to connect to an RSU via V2I. For example, not all the
vehicles might have an appropriate network interface card, and/or are not in the range of
connectivity of an RSU. Analogously, we assume that only k RSUs (i.e., k h {1, 2, , } = with
h < m) are available to V2I communications.
In such scenario, we are now introducing the Protocol Switching Decision as follows:
Definition (Protocol Switching Decision). A source vehicle V
s
, sending via V2V a message of
length L to a destination vehicle V
d
, will switch to V2I if there exists an optimal path linking V
s

with an RSU. Analogously, a source vehicle V
s
, sending via V2I a message of length L to a
destination vehicle V
d
, will switch to V2V if there exists an optimal path linking V
s
with
neighbouring vehicles.
The optimal path will be defined hereafter.
For the connectivity link from the i-th to the j-th vehicle we define as link utilization time
q
(i, j)
[s] the time needed to transmit a message of length L [bit] from the i-th to the j-th
vehicle, at an actual data rate f
(i, j)
[Mbit/s], such as

18

i j
i j
L
q
f
( , )
( , )
. = (18)
For a link between the i-th vehicle and the k-th RSU, the data rate f
(i, j)
is obtained by the
nominal data rate
( , ) i k
f
by applying a Data Rate Reduction (DRR) factor (i.e.,

(i, k)
[s]) that
depends on the distance from the vehicle to the RSU, such as

i k i k i k
f f
( , ) ( , ) ( , )
. =

(19)
The DRR factor increases when a vehicle is laying within the bound of a wireless cell.

Fig. 9. VANET scenario with heterogeneous wireless network infrastructure partially
covering the grid
Let us now define a Path in a vehicular network as:
Definition (Path). Given the i-th vehicle and the k-th RSU, a path is a sequence of M hops
connecting the i-th vehicle with the k-th RSU, where a single hop represents a link between two
neighbouring vehicles. The path length represents the number of hops M for a single path.
It follows that the maximum number of directed links from a vehicle to an RSU is = lh,
while the maximum number of different paths that can connect the i-th vehicle to the k-th
RSU is n.
From the definition of path, we define the path utilization time Q
(i, k)
[s] from the i-th vehicle to
the k-th RSU as the sum of single link utilization time parameters (i.e., q
(i, j)
), for each hop that
comprises the path, as

n
i k i j j x x k j x
i
x S
Q q q q L f
1
( , ) ( , ) ( , ) ( , ) ( , )
1
.
= + + + =

(20)
The optimal path will be the one, among all the paths n, with the minimized path utilization
time, such as

n
s s
i k j x
s n s n
i
x S
Q L f
1
( ) ( )
( , ) ( , )
1,2,..., 1,2,...,
1
min min .

= =
=
(21)

19
Equation (21) is compared with the link utilization times in V2V communications in order to
detect the most appropriate vehicular protocol. It represents our criterion for the optimal path
detection technique in VANETs where vehicles are V2X enable.
5.1 Data propagation rates
In this section we illustrate how a message is propagated in a VANET incorporating a
heterogeneous network infrastructure, where vehicles are communicating via V2X. For this
purpose, we shall give several definitions of message dissemination rates for different cases.
The network scenario we are referring to is that one as depicted in Figure 9. Vehicles move
in clusters in two separated lanes (i.e., lane 1, and 2), where north (i.e., N), and south (i.e., S)
represent the directions of lane 1, and 2, respectively. The message propagation direction is
assumed to be N.
Each vehicle in the grid is able to know about its local connectivity through broadcast
"hello" messages forwarded in the network. Local connectivity information received by each
vehicle establishes if a vehicle is within a cluster or is traveling alone on the road. In
contrast, a vehicle will know if there are neighbouring wireless networks to access, on the
basis of broadcast signaling messages sent by the RSUs.
Let the vehicles be traveling at a constant speed c [m/s]. The message propagation rate within a
cluster (i.e., v [m/s]) is defined as

x
v
t
, = (22)
where x [m] is the transmission range distance between two consecutive and connected
vehicles communicating via V2V, and t [s] is the time necessary for a successful
transmission, which depends on the single link of connected vehicles. Typical value of
transmission range distance is 0 < x 125 m, as shown in (Agarwal & Little, 2008).
Equation (22) represents the average message propagation rate within a cluster, because it
consists of each single contribution due to each single link (i, j) in the cluster, such as

i j
i j i j i j
i j i j i j i j
x
v v x f
h h q hL
( , )
( , ) ( , ) ( , )
, , , ( , )
1 1 1
, = = =

(23)
where v
(i, j)
[m/s] is the message propagation rate for the link (i, j), and h is the number of
hops occurred within a cluster. The message propagation rate within a cluster v [m/s]
depends on the average message propagation rate for each single hop, and increases for a
low number of hops h.
Now, let us consider v
RSU
[m/s] as the message propagation rate within the network
infrastructure, as

RSU
RSU
,
d
v
T
= (24)
where d is the distance between two consecutive RSUs, and T
RSU
is the time necessary to
forward a message between two consecutive RSUs. T
RSU
is defined as the ratio between the
message length L [bit], and the effective data rate B [bit/s], for the link between the m-th and
(m + 1)-th RSU,

20

RSU
.
L
T
B
= (25)
Notice that v
RSU
is strictly dependent on the message propagation direction: a message is
forwarded to an RSU if it is placed along the same message propagation direction. The
potential for communications between RSUs aims to avoid connectivity interruptions
caused by low traffic densities, and that the V2V protocol cannot always solve.
In Figure 10 we show the data propagation rates for the considered VANET scenario.
Notice that each message forwarded by an RSU to the next RSU has been previously sent
by a vehicle driving inside the wireless cell of the RSU. Moreover, each time an RSU
receives a message from another RSU, it will send the message (i) to the destination
vehicle if it is driving inside the actual RSU coverage, or (ii) to forward the message to
next RSU.
In the first case, the message propagation rate will depend on a downlink connection from
RSU to a vehicle, while in the second case the message propagation rate will be equal to
v
RSU
. By leveraging these considerations, we define the message propagation rate in uplink
(downlink), when a vehicle sends a message to an RSU (and vice versa), as:

UP ( , ) DOWN ( , )
, ,
r r
i m m i
x x
v f v f
L L
= =

(26)
where x
r
is the distance that separates the i-th vehicle and the m-th RSU, while f
is the
effective transmission data rate for the link (i, m) (uplink), and (m, i) (downlink), respectively.
From (24) and (26), it follows that the message propagation rate v
V2I
[m/s] for
communications between vehicles and RSUs via V2I depends on the effective transmission
data rates in uplink and downlink, and on the effective data rate for intra-RSU
communications, such as:

( )
V2I UP RSU DOWN
( , ) ( , )
1
.
r i m m i
v v v v
d B x f f
L
= + + =

= + +

(27)
After defining message propagation rates for communications via V2I, we introduce the message
propagation rate for communications via V2V (i.e., v
V2V
[m/s]), as

( )
V2V
( ), v v c
= + (28)
which depends on the constant velocity c of vehicles and on the effective transmission data
rates within a cluster C, according to (23). The positive or negative sign of v
V2V
is due by the
message propagation direction.
Finally, when no connectivity occurs (i.e., a vehicle is traveling alone in the grid), the
message propagation rate is equal to c, which depends on message propagation direction.
To sum up, we characterize the behaviour of the whole system in terms of six transition
states as follows:
1. Messages are traveling along on a vehicle in the N direction at speed c [m/s];
2. Messages are propagating multi-hop within a cluster in the N direction at speed
( )
V2V
v
+
[m/s];

21
3. Messages are traveling along a vehicle in the S direction at speed c [m/s];
4. Messages are propagating multi-hop within a cluster in the S direction at speed
( )
V2V
v

[m/s];
5. Messages are transmitted via radio by an RSU in the N direction at speed v
V2I
[m/s];
6. Messages are transmitted via radio by an RSU in the S direction at speed v
V2I
[m/s].

Fig. 10. Data propagation rates in VANET scenario with network infrastructure
States (14) are typical for data propagation with opportunistic networking technique for
vehicles communicating only via V2V, while states 5, and 6, have been added for vehicles
communicating via V2I. All the six states can occur when vehicles communicate via V2X.
In our assumptions, we considered two message propagation directions (i.e., forward and
reverse propagation).
In forward message propagation, each vehicle is assumed to travel in the N direction at speed
c [m/s], and the message is propagated in the N direction as well. The message propagation
rate has a minimum value due to the speed of the vehicle (i.e., c [m/s]) since the message is
traveling along the vehicle. When a connection between two consecutive vehicles traveling
in the N direction is available, the message will be propagated via V2V at a rate
( )
V2V
v
+
.
Moreover, if no vehicle connection is available, the bridging technique can attempt to forward
a message to clusters along the S (opposite) direction, whenever they are overlapping with
the cluster along the N direction.
Analogously, it is easy to evince that in reverse message propagation, each vehicle is assumed
to travel in the S direction at speed c [m/s], and the message is propagated in the S
direction as well. The message propagation rate will have a minimum value due to the
speed of the vehicle (i.e., c [m/s]), and a maximum bound when a message is propagating
via V2V at a rate
( )
V2V
v

. Again, if no vehicle connection is available, a message will be
forwarded via bridging to clusters along the N (opposite) direction, whenever they are
overlapping with the cluster along the S direction.
Such considerations occur for vehicles communicating via V2V and when opportunistic
networking is available. In contrast, when vehicles are communicating via V2I, the forward

22
message propagation will have a maximum bound equal to v
V2I
, while for reverse message
propagation range the maximum bound is v
V2I
.
The definitions for forward and reverse message propagation rates are given below,
respectively.
Definition (Forward Message Propagation rate): the forward message propagation rate, when a
vehicle is communicating via V2V, is in the range [c,
( )
V2V
v
+
]. In contrast, when a vehicle
communicates via V2I, the forward message propagation rate is in the range [c, v
V2I
].
Definition (Reverse Message Propagation rate): the reverse message propagation rate, when a
vehicle communicates via V2V, is in the range [c,
( )
V2V
v

], while for vehicles communicating via
V2I, the range of reverse message propagation rate is [c, v
V2I
].
5.2 V2X algorithm
This section illustrates how V2X takes a protocol switching decision.
The algorithm for handing over from V2V to V2I, and vice versa, is described by its pseudo-
code in Figure 11. It is mainly based on (i) the Infrastructure Connectivity (IC) parameter,
which gives information if a vehicle is able to connect to an RSU, and on (ii) the optimal path
detection technique. The algorithm accepts one input (i.e., the vehicles IC), and returns the
actual message propagation rate (i.e., {v
V2V
, v
V2I
}).

Input : IC
Output :
v
V2V
, if a vehicle communicates via V2V
v
V2I
, if a vehicle communicates via V2I
while IC = 0 do
A vehicle is connected via V2V, v
V2V
end
else
if IC = 1 then
Optimal path detection, v
V2I
or v
V2V
end
end
if A vehicle communicates with an RSU via V2I then
the RSU tracks the destination's position,
if Destination vehicle is inside the actual RSUs coverage then
Direct link from RSU to destination vehicle
else
The actual RSU will forward the message to next RSU
end
end
end

Fig. 11. Algorithm for protocol switching decisions in V2X

23
Let us consider the following VANET scenario. A source vehicle is communicating with
other vehicles (relay) via V2V in a sparsely connected neighbourhood, where the
transmission range distance between two consecutive vehicles is under a connectivity
bound, i.e. x 125 m.
The source vehicle is driving inside any wireless cell, and is receiving "hello" broadcast
messages from other vehicles nearby. Local connectivity information will notify the vehicle
the availability of vehicles to communicate with via V2V; no RSU presence will be notify to
the vehicle. In this case (i.e., V2V availability, and no V2I) the IC parameter for vehicle A will
be set to 0. Otherwise, when a vehicle enters a wireless network, the presence of an available
RSU to access will be directly sent to the vehicle by means of its associated IC parameter set
to 1.
Finally, a destination vehicle is driving far away from A, and other vehicles (relay) are
available to communicate each other.
In such scenario, the algorithm works according to two main tasks, such as (i) checking IC
parameter, and (ii) tracking the destination vehicle(s). Every time a vehicle forwards a
message it checks its IC value. When IC = 1, the vehicle calculates the optimal path according
to (21) in order to send the message directly to the selected RSU via V2I. Otherwise, the
vehicle forwards the message to neighbouring vehicles via V2V.
By supposing the RSU knows the destination vehicles position (i.e. by A-GPS), if the
destination vehicle is traveling within the RSUs wireless coverage, the RSU will send the
message directly to the destination vehicle. Otherwise, the RSU will be simply forwarding
the message to the RSU that is actually managing the vehicles connectivity. Finally, the
message will be received by the destination vehicle.
Some simulation results are now shown in order to verify the effectiveness of V2X
approach as compared with traditional opportunistic networking scheme in VANET. As a
measure of performance, we calculate the average message displacement (i.e. X [m]) in
VANETs via V2X. The message displacement is a linear function, depending on time, and
varying for different traffic scenarios, message propagation speeds, and network
conditions. It follows that in each of the six states listed in Section 5.1, the message
displacement X(t) will be as follows:
1. ( ) , X t c t = for messages traveling along on a vehicle in the N direction at speed
c [m/s];
2.
( )
V2V
( ) , X t v t
+
= for messages propagating multi-hop within a cluster in the N direction at
speed
( )
V2V
v
+
[m/s];
3. ( ) , X t c t = for messages traveling along a vehicle in the S direction at speed
c [m/s];
4.
( )
V2V
( ) , X t v t
= for messages propagating multi-hop within a cluster in the S direction at

speed
( )
V2V
v

[m/s];
5.
V2I
( ) , X t v t = for messages transmitted via radio by an RSU in the N direction at speed
v
V2I
[m/s];
6.
V2I
( ) , X t v t = for messages transmitted via radio by an RSU in the S direction at speed
v
V2I
[m/s].
States 1, 2, and 5 refer on a forward message propagation, while stated 3, 4, and 6 on a reverse
message propagation, respectively.

24
We simulated a typical vehicular network scenario by the following events:
i. at t = 0 s a source vehicle is traveling in the N direction and sends a message along on
the same direction, (state 1);
ii. at t = 2 s the message is propagated multi-hop within a cluster in the N direction, (state
2);
iii. at t = 6 s a relay vehicle enters an RSUs radio coverage, and the message is transmitted
via V2I to the RSU. Finally, it will be received by other vehicles at t = 10 s, (state 5).
We compared this scenario with traditional opportunistic networking technique in
VANETs, where the following events occur:
i. at t = 0 s a source vehicle traveling in the N direction sends a message along on the
same direction, (state 1);
ii. at t = 4 s the message is forwarded to a vehicle in the S direction, (state 3);
iii. at t = 6 s the message propagates via multi-hop within a cluster in the N direction, (state
2). The transmission stops at t = 10 s.
For comparative purposes, main simulation parameters has been set according to (Wu et al.,
2004), including c = 20 m/s, d = 500 m, typical message size L = 300 bit, data rate
transmission B = 10 Mbit/s (e.g., for WiMax connectivity), and x
r
= 400 m. The transmission
rates in DSRC have been assumed equal to 6 Mbit/s (Held, 2007). We assumed a cluster size
equal to h = 5, and different distances between couples of vehicles (i.e., 100, 75, 50, 40, and
30 m). For each hop the transmission range has been hold (i.e. < 125 m).
Figure 12 (left) depicts the maximum and minimum message propagation bounds for V2X in
forward message propagation mode. Notice a strong increase in the message propagation with
respect to other forms of opportunistic networking: after t = 10 s, the message has been
propagating for approximately 30 km in V2X (Figure 12 (left)), while only 1.5 km in
traditional V2V (Figure 12 (right)). The high performance gap is mainly due to the protocol
switching decision of V2X, which exploits high data rates from wireless network
infrastructure. In contrast, opportunistic networking with V2V is limited to use only DSRC
protocol.

Fig. 12. Forward message propagation for (left) V2X protocol, (right) traditional
opportunistic networking
Analogously, we simulated how a message is forwarded in reverse message propagation mode,
where vehicles are traveling in an opposite direction (Figure 13). In this case, the message

25
propagation rates are in the range [c; v
V2I
] and [c;
( )
V2V
v

] [m/s], for V2X and traditional
opportunistic networking scheme, respectively. Once again, while V2X assures high values
for message displacement (i.e., at t = 10 s, a message has been propagated up to around
70 km, as shown in Figure 13 (left)), traditional V2V can achieve low values (i.e., at t = 10 s,
messages have reached 1.3 km far away from the source vehicle (see Figure 13 (right)).
Notice the fluctuations of message displacement in forward and reverse cases with V2X (i.e.
50, and 70 km, respectively). They are mainly due to traffic density, and RSUs positions (i.e.
inter-RSU distance). In general, high performance are obtained with V2X, while low
message propagation distance with traditional V2V.

Fig. 13. Reverse message propagation for (left) V2X protocol, (right) traditional opportunistic
networking
6. Conclusions
In this chapter we have discussed application of VHO in the context of VANETs in order to
optimize application delivery through a mixed V2V/V2I infrastructure. Vertical handover
strategies can be applied to assure VANET connectivity context-aware, and content-aware.
Various metrics can be adopted to trigger handover decisions including RSS measurements,
QoS parameters, and mobile terminal location information. This last represents the most
common parameter used to drive VHO decisions.
Hence, a geometrical model has been presented where GPS-equipped mobile terminals
exploit their location information to pilot handover and maximize communication
throughput taking into account mobile speed. The proposed technique has been described
via both analytical and simulated results, and validation of its effectiveness has been
supported by a comparison with a traditional vertical handover method for VANETs (Yan et
al., 2008).
Moreover, we have described a hybrid vehicular communication protocol V2X and the
mechanism by which a message is propagated under this technique. V2X differs from
traditional V2V protocol by exploiting both V2V and V2I techniques, through the use of a
fixed network infrastructure along with the mobile ad-hoc network. In this heterogeneous
scenario, we have characterized the upper and lower bounds for message propagation rates.
Validation of V2X has been carried out via simulation results, showing how V2X protocol

26
improves network performance, with respect to traditional opportunistic networking
technique applied in VANETs.
7. References
Held, G. (2007). Inter- and intra-vehicle communications, CRC Press.
Chiara, B.D.; Deflorio, F. & Diwan, S. (2009). Assessing the effects of inter-vehicle
communication systems on road safety, Intelligent Transport Systems, IET, Vol. 3,
No. 2, June 2009, pp. 225235.
Pollini, G.P. (1996). Trends in handoff design, IEEE Communication Magazine, Vol. 34, No. 3,
March 1996, pp. 8290.
Inzerilli, T. & Vegni, A.M. (2008). A reactive vertical handover approach for WiFi-UMTS
dual-mode terminals, Proceeding of 12th Annual IEEE International Symposium on
Consumer Electronics, April 2008, Vilamoura (Portugal).
Vegni, A.M.; Tamea, G.; Inzerilli, T. & Cusani, R. (2009). A Combined Vertical Handover
Decision Metric for QoS Enhancement in Next Generation Networks, Proceedings of
IEEE International Conference on Wireless and Mobile Computing, Networking and
Communications 2009, pp. 233238, October 2009, Marrakech (Morocco).
Vegni, A.M. & Esposito, F. (2010). A Speed-based Vertical Handover Algorithm for VANET,
Proceedings of 7th International Workshop on Intelligent Transportation, March 2010,
Hamburg (Germany).
Vegni, A.M. & Little, T.D.C. (2010). A Message Propagation Model for Hybrid Vehicular
Communication Protocols, Proceeding of 2nd International Workshop on
Communication Technologies for Vehicles, July 2010, Newcastle (UK).
McNair, J. & Fang, Z. (2004). Vertical handoffs in fourth-generation multinetwork
environments, IEEE Wireless Communications, Vol. 11, No.3, (June, 2004), pp. 815.
Esposito, F.; Vegni, A.M.; Matta, I. & Neri, A. (2010). On Modeling Speed-Based Vertical
Handovers in Vehicular Networks Dad, slow down, I am watching the movie,
at IEEE Globecom 2010 Workshop on Seamless Wireless Mobility 2010, December 2010,
Miami (USA).
Inzerilli, T.; Vegni, A.M.; Neri, A. & Cusani, R. (2008). A Location-based Vertical Handover
algorithm for limitation of the ping-pong effect, Proceedings on 4th IEEE International
Conference on Wireless and Mobile Computing, Networking and Communications,
October 2008, Avignon (France).
Kim, W.I.; Lee, B.J.; Song, J.S.; Shin, Y.S. & Kim, Y.J. (2007). Ping-Pong Avoidance Algorithm
for Vertical Handover in Wireless Overlay Networks, Proceeding of IEEE 66th
Vehicular Technology Conference, pp. 1509-1512, September 2007.
Chen, Y.S.; Cheng, C.H.; Hsu, C.S. & Chiu, G.M. (2009). Network Mobility Protocol for
Vehicular Ad Hoc Network, Proceeding of IEEE Wireless Communication and
Networking Conference, April 2009, Budapest (Hungary).
Yan, Z.; Zhou, H.; Zhang, H. & Zhang, S. (2008). Speed-Based Probability-Driven Seamless
Handover Scheme between WLAN and UMTS, Proceeding of 4th International
Conference on Mobile Ad-hoc and Sensor Networks, December 2008, Wuhan (China).
Laiho, J.; Wacker, A. & Novosad, T. (2005). Radio Network Planning and Optimisation for
UMTS, 2nd edition, Chapter 6.

27
Resta, G.; Santi, P. & Simon, J. (2007). Analysis of multihop emergency message propagation
in vehicular ad hoc networks, Proceeding of the 8th ACM International Symposium on
Mobile Ad Hoc Networking and Computing, pp. 140149, September 2007, Montreal
(Canada).
Jiang, H.; Guo, H. & Chen, L. (2008). Reliable and Efficient Alarm Message Routing in
VANET, Proceeding of the 28th International Conference on Distributed Computing
Systems Workshops, pp. 186191.
Yousefi, S.; Fathy, M. & Benslimane, A. (2007). Performance of beacon safety message
dissemination in Vehicular Ad hoc NETworks (VANETs), Journal of Zhejiang
University Science A.
Chen, W.; Guha, R.; Kwon, T.; Lee, J. & Hsu, Y. (2008). A survey and challenges in routing
and data dissemination in vehicular ad hoc networks, Proceeding of IEEE
International Conference on Vehicular Electronics and Safety, Columbus (USA),
September 2008.
Nadeem, T., Shankar, P. & Iftode, L. (2006). A comparative study of data dissemination
models for VANETs, Proceeding of the 3rd Annual International Conference on Mobile
and Ubiquitous Systems, pp. 110, San Jose (USA), July 2006.
Gerla, M.; Zhou, B.; Leey, Y.-Z.; Soldo, F.; Leey, U. & Marfia, G. (2006). Vehicular Grid
Communications: The Role of the Internet Infrastructure, Proceeding of Wireless
Internet Conference, Boston (USA), August 2006.
Marfia, G.; Pau, G.; Sena, E.D.; Giordano, E. & Gerla, M. (2007). Evaluating Vehicle Network
Strategies for Downtown Portland: Opportunistic Infrastructure and Importance of
Realistic Mobility Models, Proceeding of MobiOpp 2007, Porto Rico, June 2007.
Agarwal, A. & Little, T.D.C. (2008). Access Point Placement in Vehicular Networking,
Proceeding of 1st International Conference on Wireless Access in Vehicular Environments,
Dearborn (USA), December 2008.
Wu, H.; Fujimoto, R. & Riley, G. (2004). Analytical Models for Information Propagation in
Vehicle-to-Vehicle Networks, Proceeding of ACM VANET, Philadelphia (USA),
October 2004.
Ayyappan, K. & Dananjayan, P. (2008). RSS Measurement for Vertical Handoff in
Heterogeneous Network, Journal of Theoretical and Applied Information Technology,
Vol. 4, Issue 10, October 2008.
Yang, K.; Gondal, I.; Qiu, B. & Dooley, L.S. (2007). Combined SINR based vertical handover
algorithm for next generation heterogeneous wireless networks, Proceeding on IEEE
GLOBECOM 2007, November 2007, Washinton (USA).
Vegni, A.M.; Carli, M.; Neri, A. & Ragosa, G. (2007). QoS-based Vertical Handover in
heterogeneous networks, Proceeding on 10th International Wireless Personal
Multimedia Communications, CD-ROM no. of pages: 4, December 2007, Jaipur
(India).
Jesus, V.; Sargento, S.; Corujo, D.; Senica, N.; Almeida, M. & Aguiar, R.L. (2007). Mobility
with QoS support for multi-interface terminals: combined user and network
approach, Proceeding on 12th IEEE Symposium on Computers and Communications, pp.
325332, July 2007, Aveiro (Portugal).

28
Kibria, M.R.; Jamalipour, A. & Mirchandani, V. (2005). A location aware three-step vertical
handover scheme for 4G/B3G networks, Proceeding on IEEE GLOBECOM 2005, Vol.
5, pp. 27522756, November 2005, St. Louis (USA).
Wang, S.S.; Green, M. & Malkawi, M. (2001). Adaptive handover method using mobile
location information, Proceeding on IEEE Emerging Technology Symposium on
Broadband Comm. for the Internet Era Symposium, pp. 97101, September 2001,
Richardson (USA).
Vegni, A.M. (2010). Multimedia Mobile Communications in Heterogeneous Wireless
Networks -Part 2, PhD thesis, University of Roma Tre, March 2010, available online
at http://www.comlab.uniroma3.it/vegni.htm.
Sarmad Sohaib
1
and Daniel K. C. So
2
1
University of Engineering and Technology, Taxila
2
The University of Manchester
1
Pakistan
2
United Kingdom
1. Introduction
Inter-vehicle communication is envisioned to play a very important role in the future,
improving road safety and capacity. This can be achieved by utilizing cooperative relaying
techniques where the communicating nodes exploit spatial diversity by cooperating with
each other (Laneman et al., 2004). This alleviates the detrimental effects of fading and offers
reliable data transfer. The source node broadcasts the signal to the destination node directly,
and also through the relay nodes. Both the direct and relayed signals are combined at the
destination. However, conventional cooperative communication systems require frame or
symbol level synchronization between the cooperating nodes. The lack of synchronization
results in inter-symbol interference (ISI) and degrades the system performance. This problem
will be more severe in inter-vehicle communication as maintaining synchronization in fast
moving nodes is very difcult. In this chapter, we present the major asynchronous cooperative
communication protocols that can be employed for inter-vehicle communications. These are
the asynchronous delay diversity technique (Wei et al., 2006), asynchronous space-time block
code (STBC) cooperative system (Wang & Fu, 2007), and asynchronous polarized cooperative
(APC) system (Sohaib & So, 2009; 2010).
2. Conventional cooperative communication system model
A three node cooperative network containing the source (S), relay (R) and destination (D)
nodes is shown in the Fig. 1. The information will be transmitted from the source node to the
destination node directly and also through the relay node. Both the direct and relay signals
are combined at the destination using combiners (Brennan, Feb 2003). In general, there are
two kinds of relaying modes; amplify-and-forward (ANF), where the relay simply amplies the
noisy version of the signal transmitted by source, and decode-and-forward (DNF), where relay
decodes, re-encodes and re-transmits the signal.
The conventional ANF channel model is characterized by transmitting and receiving in
orthogonal frequency bands or time slots (Laneman et al., 2004; Sohaib et al., 2009). Here we
consider the ANF scheme with the relay node transmitting at the same frequency band as the
source node, but in subsequent time-slot.
Asynchronous Cooperative Protocols
for Inter-vehicle Communications
2
R
S D
h
sd
h
sr
h
rd
Fig. 1. Cooperative communication netwrok.
The channel

h
ij
between the i-th transmit and j-th receive antenna is given by
h
ij
=
U1
u=0
h
ij
(u)
_
PL
ij
(1)
where, h
ij
(u) is the normalized channel gain, which is an independent and identically
distributed (i.i.d.) complex Weibull random variable with zero mean. This describes the
random fading effect of multipath channels, and is assumed to be frequenct selective fading
with U the total number of frequency selective channel taps. Weibull distribution is used for
the analysis of APC in vehicle-to-vehicle communication as it ts best (Matolak et al., 2006).
The path loss factor PL
ij
models the signal attenuation over distance, and is given by (Haykin
& Moher, 2004)
PL
ij
=
(4)
2
G
t
G
r
2
_
d
ij
_
= PL
0
_
d
ij
_
(2)
where PL
0
is the reference path loss factor, d
ij
is the distance between i-th transmitter and
j-th receiver, is the path loss exponent depending on the propagation environment which is
assumed to be the same over all links, is the wavelength, and G
t
and G
r
are the transmitter
and receiver antenna gains respectively.
In a typical three node system, single transmission is normally divided into two timeslots
(Peters & Heath, 2008; Tang & Hua, 2007). In the rst timeslot, the source node broadcasts
the signal to the destination and the relay node. The received signal at the destination node
directly from the source node is
y
sd
(t) =
E
s
PL
sd
U1
u=0
h
sd
(u)x(t u) + n
d
(t) (3)
where x is the transmitted signal from the source with unit energy, E
s
is the transmitted
signal energy from the source, h
sd
is the normalized channel gain from the source to the
destination with a corresponding path loss of PL
sd
, and n
d
(t) captures the effect of AWGN
at the destination. Similarly, at the same timeslot the relay node receives the same signal from
the source, given by
y
sr
(t) =
E
s
PL
sr
U1
u=0
h
sr
(u)x(t u) + n
r
(t) (4)
where h
sr
is the normalized channel gain from the source to the relay with a corresponding
path loss of PL
sr
, and n
r
(t) is the AWGN at the relay.
30 Advances in Vehicular Networking Technologies
S-D
S-R
R-D
time
t t+T
Fig. 2. Timing diagram of ANF cooperative scheme.
In the second timeslot the signal received at the relay node is amplied by a factor k
r
and
forwarded to the destination given by
y
rd
(t + T) =
k
PL
rd
U1
u=0
h
rd
(u)y
sr
(t u) + n
d
(t + T) (5)
where T = LT
s
is the timeslot or frame duration with L being the total number of symbols
per frame and T
s
the symbol period, h
rd
is the normalized channel gain from the relay to
destination node having a corresponding path loss of PL
rd
, and n
d
(t + T) is the AWGN
at the destination node. The transmitter estimates path loss through the reverse link and is
assumed to be perfectly estimated. On the other hand, instantaneous channel fading gain is
not assumed to be known at the transmitter, as it requires feedback information. Therefore,
setting identical received signal energy from the direct and relayed link, the amplication
factor k
r
is given by
k
r
=
_
E
s
E
_

h
sd
2
_
E
r
E
_

h
rd
2
_ =
E
s
/PL
sd
(E
s
/PL
sr
+ N
0
) /PL
rd
(6)
where E
r
is the received signal energy at the relay node. All AWGN noises are modeled as
zero mean mutually independent circular symmetric complex Gaussian random sequences
with power spectral density (PSD) N
0
. Exact channel state information (CSI) is assumed to be
available at the receiver only, and not at the transmitter.
For conventional ANF system, the signal in (3) and (5) are combined at the destination node
using diversity combiners, e.g. Maximal Ratio Combiner (MRC). The diversity gain achieved
through cooperation can compensate the additional noise in the relay (Laneman et al.,
2004). Hence, cooperative diversity schemes achieve better performance than non-cooperative
schemes.
Fig. 2 illustrates the timing diagramof ANF cooperative system, where, t is the time when the
source node starts transmitting the data to the destination and relay nodes. The relay node
will start transmitting after a duration of T. Therefore it takes two orthogonal channels for
one complete transmission, thus decreases the spectral efciency of the system. Also frame
level synchronization is required in conventional ANF, which is not always achievable in
wireless communication. The diversity gain achieved through cooperation can compensate
for the additional noise in the relay (Laneman et al., 2004). Hence, the cooperative diversity
schemes achieve better performance than non-cooperative schemes.
31 Asynchronous Cooperative Protocols for Inter-vehicle Communications
Source Destination
Relay
Relay +
delay
Fig. 3. System structure of cooperative communications.
S-R and S-D
time
R1-D
R2-D
Fig. 4. Timing diagram of asynchronous delay diversity cooperative scheme.
3. Asynchronous cooperative systems
In this section we present a brief summary of the three major inter-vehicle asynchronous
cooperative communication systems.
3.1 Asynchronous delay diversity technique
In (Wei et al., 2006), a distributed delay diversity approach is proposed in the
Relay-Destination (R-D) link to achieve spatial diversity as shown in Fig. 3. Error detection
schemes such as cyclic redundancy check (CRC) is employed at the relay nodes to determine
whether the received packet is error free or not. If the received packet is error-free, the relay
node will then forward the information packet to the destination, after an additional articial
delay. On the contrary if the packed is in error, it will be dropped at the relay node. Assuming
the CRC code can perfectly detect any packet error the forwarded signal from the relay is
thus a delayed version of the transmitted symbols. Hence, the destination node will see an
equivalent frequency selective fading channel in the form of articially introduced delays.
Fig. 4 illustrates the timing diagram of this scheme.
To equalize the frequency selectivity, a decision feedback equalizer (DFE) is employed at the
destination node. It also combines the inputs fromthe direct link channel, and relay link ones.
Although this scheme can mitigate the synchronization problem, it uses half duplex relay
node which reduces the spectral efciency due to the bandwidth expansion or extended time
duration. Constellation size has to be increased to maintain the spectral efciency which then
reduces the performance gain over non-cooperative single-input single-output (SISO) scheme.
3.2 Asynchronous space-time block code cooperative system
Instead of using the simple delay diversity code in the R-D link, the asynchronous STBC
is proposed in (Wang & Fu, 2007) to achieve distributed cooperative diversity. The system
and timing diagram for this scheme is identical to that of the asynchronous delay diversity
scheme in Fig. 3 and Fig. 4. At the relay, the detected symbols are mapped into the orthogonal
STBC matrix. Each relay then randomly select one row from this matrix for transmission.
The random cyclic delay diversity technique is then applied to make the equivalent channels
frequency selective. At the destination node the frequency domain equalizer (FDE) is
employed to combine and equalize the received signal.
The scheme has a disadvantage that it could suffer performance degradation due to diversity
loss by random row selection. Similar to the previous scheme, this system also assumes the
relay to be half duplex which results in low spectral efciency.
3.3 Asynchronous polarized cooperative system
Most cooperative communication systems, including (Wang & Fu, 2007; Wei et al., 2006),
employ half duplex relays. This is because full duplex relay that uses the same time and
frequency for transmission and reception is difcult to implement. The transmitted signal
will overwhelm the received signal. In view of this, the asynchronous polarized cooperative
(APC) systemis proposed in (Sohaib & So, 2009; 2010), and is illustrated in Fig. 5. It allows full
duplex relay operation, and does not require frame of symbol level synchronization. In this
scheme every vehicle is equipped with dual polarized antennas that can auto-congure itself
to be the source, relay and destination node. The vehicle working as a source only activates
the vertical polarized antenna for transmission, whereas the destination vehicle congures
the dual polarized antennas for reception. The vehicle working as a relay uses dual polarized
antennas for transmission and reception at the same time and at the same frequency thereby
achieving the full duplex ANF communication and effectively reducing the transmission
duration and increasing the throughput rate. The solid lines represent transmission and
reception on the same polarization, also known as co-polarization. On the other hand, the
dotted lines represent transmission in one polarization but reception in the other polarization,
also known as cross-polarization. The effect of cross-polarization is considered as it is
impossible to maintain the same polarization between the transmitter and the receiver due
to the complex propagation environment in terrestrial wireless communications. For more
practical consideration, path loss is also included in the analysis.
For a relay to operate in full duplex mode the transmission and reception channels must
be orthogonal either in time-domain or in frequency domain, otherwise the transmitted
signal will interfere with the received signal. In theory, it is possible for relay to cancel
out interferences as it has the knowledge of transmitted signal. In practice, however, the
transmitted signal is 100-150dB stronger than the received signal and any error in the
interference cancellation can potentially be disastrous (Fitzek &Katz, 2006). Due to this reason,
the installation of co-polarized antennas at the relay node in place of dual-polarized antennas
is not feasible for full duplex relay. However, with dual-polarized antenna the transmitted
signal on one polarization is orthogonal to the received signal at another polarization, thereby,
enabling the relay to communicate in full duplex mode, not the overall system.
The source node will broadcast using vertical polarization. The vertically polarized received
signal at the relay node is the same as (4).
The received signal at the relay node is amplied by a factor k
r
, and transmitted immediately
to the destination node through horizontal polarization. Radio propagation and signal
V-pol antenna
H-pol antenna
V-pol antenna
Source
Relay
H-pol antenna
V-pol antenna
Destination
h
v
rd
h
h
rd
h
v
sd
h
h
sd
h
sr
Fig. 5. Asynchronous polarized cooperative system for inter-vehicular communication.
processing at the relay node will cause some additional time delay , which could be a few
symbols duration and is much shorter than the frame duration T. It must be noted that the
APC systemdoes not require symbol level synchronization, between the source and relay, and
thus can be any positive real number. Fig. 6 illustrates the timing diagram of this scheme.
The vertically and horizontally polarized signal received at the destination, denoted as y
d
v
and y
d
h
respectively, are given by
y
d
v
(t) =
E
s
h
v
sd
x(t u) + k
r

h
v
rd
y
sr
(t u) + n
d
v
(t) (7)
and
y
d
h
(t) =
E
s
h
h
sd
x(t u) + k
r

h
h
rd
y
sr
(t u) + n
d
h
(t). (8)
The received signals of the above equations can therefore be written in matrix form as
y
d
v
(t)
y
d
h
(t)
. .
y
d
=
h
v
sd

h
v
rd
h
h
sd

h
h
rd
. .
H
E
s
x(t u)
k
r
y
sr
(t u)
n
d
v
(t)
n
d
h
(t)
. .
n
(9)
where n is the 2 1 i.i.d. zero mean complex AWGN vector with variance E
_
n n
H
= N
0
I,
and I is an identity matrix. The diagonal elements of H correspond to co-polarization, while
the off-diagonal elements correspond to cross-polarization. The relay amplication factor k
r
is
k
r
=
_
E
s
_
E
_

h
v
sd
2
_
+E
_
h
h
sd
2
__
E
r
_
E
_

h
v
rd
2
_
+E
_
h
h
rd
2
__ (10)
where E
r
is the received signal energy at the relay node given by
E
r
= E
s
E
_

h
sr
2
_
+ N
0
. (11)
Since the source and relay node are spatially separated apart, we can assume the channel
from the source to the destination is not correlated with the channel from the relay to the
T
S-R
S-D
R-D
J time
Fig. 6. Timing diagram of APC scheme.
destination. In other words, the co-polarization elements of the channel h
v
sd
and h
h
rd
and the
cross-polarization elements h
h
sd
and h
v
rd
are assumed to be completely un-correlated. Therefore
E
_
h
v
sd
h
h
rd
_
= E
_
h
h
sd
h
v
rd
_
= 0 (12)
and
E[h
v
sd
h
v
rd
] = E
_
h
h
sd
h
h
rd
_
= 0. (13)
We dene the receive correlation coefcient as
r
=
E
_
h
v
sd
h
h
sd
_
=
E
_
h
v
rd
h
h
rd
_
. (14)
At the destination node, the vertical and horizontal polarized signals are received at different
time due to the signal processing and additional propagation delay caused by the relay.
Because of cross polarization, the delayed signal from the relay becomes an ISI. Therefore
equalization for each polarization is required. As there are two branches from the vertical
and horizontal polarization, diversity combiner is needed. The frequency domain diversity
combiner and equalizer (FDE-MRC) is therefore used and is shown in Fig. 7. Assuming that
DFT
Horizontally polarized
antenna
CP removal
Vertically polarized
antenna
Maximum
Ratio
Combiner
DFT CP removal
Equalizer
(MMSE)
IDFT
Detector
Fig. 7. Receiver structure of the APC MIMO system.
cyclic prex (CP) with duration longer than delay is inserted before transmission from the
source node, and removed at the destination node, the signals received at the destination node
from the source and relay nodes are transformed into frequency domain by taking L points
discrete Fourier transform (DFT). The resulting signal spectras at the k-th subcarrier from
vertical and horizontal polarized branches are respectively given by
Y
d
v
(k) =
E
s
X(k)
_
h
v
sd
+ k
r
h
v
rd
h
sr
e
j2
k
L
_
+ k
r
h
v
rd
N
r
(k)e
j2
k
L
+ N
d
v
(k)
E
s
X(k) H
v
(k) + N
v
(k) (15)
and
Y
d
h
(k) =
E
s
X(k)
_
h
h
sd
+ k
r
h
h
rd
h
sr
e
j2
k
L
_
+ k
r
h
h
rd
N
r
(k)e
j2
k
L
+ N
d
h
(k)
E
s
X(k) H
h
(k) + N
h
(k) (16)
where k = {1, 2, . . . , L}, X(k) is the transmitted signal in frequency domain, N
r
(k) is the relay
noise in frequency domain, and N
v
(k) and N
h
(k) are the effective noises at the vertical and
horizontal antennas respectively at the destination node, H
v
(k) and H
h
(k) are the effective
channels at vertical and horizontal antennas respectively at the destination node given by
H
v
(k) =

h
v
sd
+ k
r
h
v
rd
h
sr
e
j2
k
L
(17)
and
H
h
(k) =

h
h
sd
+ k
r
h
h
rd
h
sr
e
j2
k
L
. (18)
The polarized frequency domain signals Y
d
v
(k) and Y
d
h
(k) are combined through MRC at the
destination node and the resultant signal spectrum Y(k) is
Y(k) = Y
d
v
(k)H
v
(k) + Y
d
h
(k)H
h
(k). (19)
The combined signal Y(k) is input to MMSE equalizer given by
W(k) = arg min
W
E
h
_
W(k) Y(k)
E
s
X(k)
2
_
(20)
where E
h
[.] denotes the expectation conditioned on the channel gains. For ease of notation
and without loss of generality, we drop the index k in the following derivation. Substituting
the value of Y from (15), (16), and (19) into the objective function of (20)
J =E
h
_
W
_
E
s
X|H
v
|
2
+ N
v
H
v
+
E
s
X|H
h
|
2
+ N
h
H
h
_
E
s
X
2
_
= E
h
_
_
W|H
v
|
2
+ W|H
h
|
2
1
_
E
s
X +WN
v
H
v
+ WN
h
H
2
_
. (21)
Solving the above equation for minimum value of W, we take the derivate of J w.r.t. W and
set it to 0, i.e.
dJ
dW
= 0
E
s
_
|H
v
|
4
W
+|H
h
|
4
W
+2 |H
v
H
h
|
2
W
|H
v
|
2
|H
h
|
2
_
+ N
v
0
|H
v
|
2
W
+ N
h
0
|H
h
|
2
W
= 0
_
E
s
|H
v
|
4
+ E
s
|H
h
|
4
+2 |H
v
H
h
|
2
+|H
v
|
2
N
v
0
+|H
h
|
2
N
h
0
_
W
= E
s
_
|H
v
|
2
+|H
h
|
2
_
(22)
Rearranging (22) we obtain,
W
=
E
s
_
|H
v
|
2
+|H
h
|
2
_
E
s
_
|H
v
|
4
+|H
h
|
4
+2 |H
v
H
h
|
2
_
+|H
v
|
2
N
v
0
+|H
h
|
2
N
h
0
(23)
Assuming H = |H
v
|
2
+|H
h
|
2
, (23) becomes,
W
=
H
|H|
2
+|H
v
|
2
N
v
0
E
s
+|H
h
|
2
N
h
0
E
s
. (24)
Taking the conjugate on both side and adding the index k, we obtain the nal form
W(k) =
H
(k)
|H(k)|
2
+|H
v
(k)|
2
N
v
0
E
s
+|H
h
(k)|
2
N
h
0
E
s
(25)
where
H(k) = |H
v
(k)|
2
+|H
h
(k)|
2
, (26)
N
v
0
= N
0
_
1 + k
2
r
E
_

h
v
rd
2
__
= N
0
_
1 +
k
2
r

PL
v
rd
_
(27)
and
N
h
0
= N
0
_
1 + k
2
r
E
_
h
h
rd
2
__
= N
0
_
1 +
k
2
r
PL
h
rd
_
. (28)
As the dual polarized antennas at the destination node are closely spaced, we can assume the
distance for the cross-polarizedchannels fromthe same node are the same, i.e., d
v
sd
= d
h
sd
= d
sd
and d
v
rd
= d
h
rd
= d
rd
. Therefore (27) and (28) becomes
N
v
0
= N
0
_
1 +
k
2
r
PL
rd
_
(29)
and
N
h
0
= N
0
_
1 +
k
2
r
PL
rd
_
. (30)
The detected data in frequency domain is then transformed back to time domain by using
inverse discrete Fourier transform (IDFT). Due to the full duplex nature of the relay, the
transmission time is reduced, which in turn increases the data rate as compared to the
conventional ANF protocol. Also no frame or symbol synchronization is required at the relay
node because of the use of FDE-MRC at the destination node.
3.4 Capacity analysis of asynchronous polarized cooperative system
In this section, the capacity of the APC scheme with one relay node will be presented. For
fairer comparison, we also present the capacity of ANF cooperative system which employs
dual polarized antenna at the destination node, where polarization diversity is also exploited.
3.4.1 Asynchronous polarized cooperative scheme
Given the channel information at the receiver, the ergodic capacity of the system in (15) and
(16) can be computed as
C = max
p(x)
I(x; y
d
) =
L
L +
E
_
log
2
_
1 +
E
s
G
_
E
h
_
|H
v
|
2
|N
v
|
2
+
|H
h
|
2
|N
h
|
2
____
= E
log
2
1 +
E
s
GL
.E
h
k=1
h
v
sd
+
k
r
h
v
rd
h
sr
e
j2
k
L
k
r
h
v
rd
N
r
(k)e
j2
k
L
+ N
d
v
(k)
2
+
L
k=1
h
h
sd
+
k
r
h
h
rd
h
sr
e
j2
k
L
k
r
h
h
rd
N
r
(k)e
j2
k
L
+ N
d
h
(k)
= E
log
2
1 +
E
s
GLN
0
k=1
h
v
sd
+
_
k
r
h
v
rd
h
sr
e
j2
k
L
2
1 + k
r

h
v
rd
2
+
L
k=1
h
h
sd
+
_
k
r
h
h
rd
h
sr
e
j2
k
L
2
1 + k
r
h
h
rd
(31)
where E
h
[.] denotes the expectation conditioned on the channel gains, G is a normalization
factor that is used to make sure that the transmission energy of the APC scheme is the same
as that of non-cooperative scheme, and is given by
G = 1 +
PL
rd
PL
sd
. (32)
Notice that the pre-log factor
L
L+
can be approximated to be one as the frame length L is
much larger than the delay . Hence the APC scheme will have a higher capacity than the
conventional scheme, which inevitably has the 1/2 pre-log factor.
3.4.2 Polarized ANF
As conventional ANF does not have the cross polarized channels, a polarized ANF system is
presented in this subsection for fairer comparison with the APC scheme. The system model of
polarized ANF with vertical polarized source antenna, vertical polarized relay antenna and
dual polarized destination antennas is given as
y
pa
= H
pa
E
s
x(t u) +n
pa
where H
pa
=
_
h
v
sd

h
h
sd
k
r
h
v
rd
h
sr
k
r
h
h
rd
h
sr
_
T
and
n
pa
=
1 0 0 0 0
0 1 0 0 0
0 0 1 0

k
r

h
v
rd
0 0 0 1

k
r

h
h
rd
. .
Q
n
d
v
(t)
n
d
h
(t)
n
d
v
(t + T)
n
d
h
(t + T)
n
r
(t)
where

h
v
sd
and

h
v
rd
are the co-polarized channels and

h
h
sd
and

h
h
rd
are the cross-polarized
channels. The ergodic capacity of the polarized ANF is thus given by
C
pa
=
1
2
E
_
log
2
det
_
I +
E
s
GN
0
H
pa
H
H
pa
_
QQ
H
_
1
__
(33)
where the normalization factor G is identical to (32). It can be noted that the 1/2 pre-log factor
in (33) shows that polarized ANF also requires two timeslots for one complete transmission.
3.5 Energy analysis of asynchronous polarized cooperative system
Cooperative communication achieves diversity through spatially separated cooperating
nodes. In most potential applications, these nodes are battery powered. Therefore energy
consumption must be minimized without compromising the transmission quality. As more
RF front ends are used by polarized antennas in the APC scheme, the total energy requirement
to achieve a required quality must be compared to the conventional ANF. In this section we
formulate the transmission energy consumption and total energy consumption of the APC
scheme.
In the following analysis, the energy consumption model developed by Cui et. al. is used (Cui
et al., 2004). The total energy consumption model that includes both the transmission energy
and the circuit energy consumption per bit is given by
E
bt
=
(P
PA
+ P
C
)
B R
b
(34)
where P
C
is the power consumption of all circuit blocks, B is the bandwidth, R
b
is the bit rate,
and P
PA
is the power consumption of all power ampliers, which depends on the transmit
power P
out
,
P
out
= E
T
R
b
B (35)
where E
T
is the sum of transmission energy from both the source and relay nodes. For the
APC scheme E
T
can be written as
E
T
= E
s
+ k
r
E
r
= E
s
1 +
E
_

h
v
sd
2
_
+E
_
h
h
sd
2
_
E
_

h
v
rd
2
_
+E
_
h
h
rd
2
_
= E
s
_
1 +
1/PL
sd
+ /PL
sd
/PL
rd
+1/PL
rd
_
= E
s
_
1 +
PL
rd
PL
sd
_
. (36)
The power consumption of the power amplies can be approximated as
P
PA
= (1 + ) P
out
(37)
where = (/) 1, with the drain efciency of the RF power amplier and the peak to
average ratio, which depends on the modulation scheme and the associated constellation size
M Cui et al. (2004)
= 3
M2
M +1
M1
. (38)
The power consumption of all circuit blocks along the signal path is given by
P
C
M
t
(P
DAC
+ P
MIX
+ P
FI LT
) +2P
SYN
+ M
r
(P
LNA
+ P
MIX
+ P
IFA
+ P
FI LR
+ P
ADC
) (39)
where P
DAC
, P
MIX
, P
FI LT
, P
SYN
, P
LNA
, P
IFA
, P
FI LR
, P
ADC
are the power consumption values
of the digital-to-analog converter (DAC), the mixer, the active lter at transmitter side,
the frequency synthesizer, the low-noise amplier, the intermediate frequency amplier,
the active lter at receiver side, and the analog-to-digital converter (ADC) respectively. M
t
and M
r
is the number of RF chains involved in one complete transmission at transmitter
and receiver side respectively. Although the APC scheme has two extra physical antennas
installed as compared to conventional ANF, both schemes effectively use the same number
of RF chains for one complete transmission. It is because conventional ANF takes two
timeslots for one complete transmission, which uses the RF chains again at the relay and the
destination. Simulation results for energy analysis are shown in the next section under the
same throughput and BER requirement.
4. Simulation results of asynchronous polarized cooperative system
Computer based Monte-Carlo simulations are carried out to illustrate the BER performance,
capacity and energy consumption of the APC system. In order to provide a fair comparison
among different schemes, spectral efciency is kept constant for all protocols and is set to be
2bps/Hz. The SISO and the APC scheme uses QPSK, whereas the ANF protocol uses 16QAM
for one relay network. This is because the SISO and the APC scheme takes approximately one
time-slot for complete transmission of one data frame, whereas conventional ANF protocol
takes two time-slots. For both the polarized ANF and the APC scheme, the cross-polarized
channel power () and receiver correlation coefcient (
r
) are set to be 0.4 and 0.5 respectively.
The time delay is assumed to be one symbol period. To obtain reasonable values of received
SNR, the transmitted signal from the source node is amplied by

PL
sd
to compensate the
path loss. The direct link SNR after this normalization is dened as
sd
. For the ANF and
APC scheme, normalization factor G in (32) is used to ensure the same total transmission
power as the SISO. Hence the normalization SNR
sd
can be used as a reference for all
schemes in capacity, and BER analysis. Table 1 summarizes the system parameters for all
simulations, which are mostly based on (Cui et al., 2004), and (Cui et al., 2003). The parameter
f
c
is the carrier frequency,

P
b
is the average probability of error for energy consumption
analysis, and M
L
is the link margin compensating the hardware process variations and other
background interference and noise. The number of transmit antennas M
t
and receive antennas
M
r
involved in one complete transmission are respectively 2 and 3 for conventional and
polarized ANF as well as the APC schemes, whereas they are both one for SISO scheme. Table
2 shows the parameters for the tapped delay line channel model derived by Matolak et. al for
vehicle to vehicle communication (Matolak et al., 2006).
P
DAC
= 15.4mW G
t
G
r
= 5dBi
P
MIX
= 30.3W = 3
P
FI LT
= 2.5mW f
c
= 5.12GHz
P
FI LR
= 2.5mW = 0.35
P
SYN
= 50mW

P
b
= 10
4
P
LNA
= 20mW M
L
= 40dB
P
IFA
= 3mW B = 10MHz
P
ADC
= 6.7mW
Table 1. System Parameters.
For capacity and BER analysis, the source to destination node distance d
sd
is set to be 200m.
The relay node is set at the midpoint between the source and destination node, i.e, d
sr
= d
rd
=
100m. For energy analysis, various positions of the relay node are considered.
Tap Index Fractional Tap Energy Weibull Shape Factor (b) Weibull Scale Factor (a)
1 0.7018 2.49 0.8676
2 0.1158 1.75 0.3291
3 0.0543 1.68 0.2226
4 0.0391 1.72 0.1903
5 0.0259 1.65 0.1528
6 0.0198 1.60 0.1322
7 0.0118 1.69 0.1040
Table 2. Vehicle to vehicle channel model (Matolak et al., 2006).
The increase in capacity of the APC scheme as compared to the conventional ANF scheme
is demonstrated in Fig. 8. The capacity of the APC scheme signicantly outperforms the
conventional ANF protocols due to the relays full duplex capability. For polarized ANF, the
use of dual polarized antenna at the destination node provides a marginal increase in capacity.
Therefore, even if polarized antennas are also used, the APC scheme has a signicant capacity
advantage over the polarized ANF scheme. The APC scheme without cross-polarization has
slightly less capacity than the APC systemwith cross-polarization but still it is higher than the
ANF systems.
The BER performance comparison among the SISO, ANF protocol, and the APC system is
presented in Fig. 9. The APC system without cross-polarization has a gain of about 4.5dB
over the conventional ANF protocol at BER 10
3
. Thus the cost of using dual polarized
antennas and separate RF chains at the relay node is justied by the signicantly lowered
BER. With the presence of cross-polarization, the performance further improves because
0 5 10 15 20 25 30 35 40
0
2
4
6
8
10
12
14
Normalized
sd
(dB)
E
r
g
o
d
i
c

C
a
p
a
c
i
t
y

(
b
p
s
/
H
z
)

ANF scheme
Polarized ANF scheme with crosspol
APC scheme without crosspol
APC scheme with crosspol
Fig. 8. Capacity comparison of one relay APC scheme.
polarization diversity can be achieved. The polarized ANF also has a marked improvement,
but is approximately 3dB worse than the APC scheme. Another observation is the differences
in the asymptotic slope of SISO to the APC scheme. It veries that diversity is achieved for
cooperative schemes with and without cross-polarization.
0 5 10 15 20
10
6
10
5
10
4
10
3
10
2
10
1
10
0
Normalized
sd
(dB)
B
E
R

Noncooperative scheme
ANF scheme without crosspol
ANF scheme with crosspol
APC scheme without crosspol
APC scheme with crosspol
Fig. 9. BER performance of APC system.
As more RF front ends are installed in the APC scheme, the total energy required to achieve
a particular quality is compared with the conventional ANF and SISO schemes in Fig. 10. The
total energy consumption is calculated using (36), where E
s
is obtained using direct link SNR
sd
observed at BER=10
4
, where
sd
= E
s
/N
0
. The direct link SNR is obtained by evaluating
the BER over 10,000 randomly generated channel samples at each transmission distance. It can
be observed that the APC scheme becomes more energy-efcient than both the ANF and SISO
protocols when d
sd
23m. The crossover point indicates the distance where the transmission
energy saving exceeds the extra circuit energy consumption in the APC scheme comparing
to the SISO and ANF scheme. In addition, for practical applications, the source to destination
node speration will be mostly larger than 20m. Hence, the APC scheme will consume less
energy in realistic scenario.
10 20 30 40 50 60 70 80 90 100
10
5
10
4
10
3
10
2
10
1
d
sd
(m)
T
o
t
a
l

e
n
e
r
g
y

c
o
n
s
u
m
p
t
i
o
n

p
e
r

b
i
t

i
n

J

Noncooperative scheme
ANF scheme
PolANF scheme
APC scheme
Fig. 10. Total energy consumption per bit over d
sd
when the relay node is located midway
between source and destination nodes.
5. Conclusion
In this chapter, we discuss some of the major asynchronous cooperative communication
protocols that can be used in vehicle-to-vehicle cooperative communications. The APCscheme
with full duplex relay that completes the data transmission between the source and the
destination in approximately same time duration as non-cooperative scheme is discussed in
detail. The performance improvement of APC scheme is demonstrated by the BER and the
capacity simulation results, which show its superiority over non-cooperative, conventional
and polarized ANF protocol. Even with the use of more RF front ends, the APC scheme has
less total energy consumption than ANF and non-cooperative schemes over more practical
distances between the nodes. Thus, the APC scheme is both spectral and energy efcient, and
is suitable for inter-vehicle cooperative communication.
6. References
Brennan, D. (Feb 2003). Linear diversity combining techniques, Proceedings of the IEEE
91(2): 331356.
Cui, S., Goldsmith, A. & Bahai, A. (2003). Energy-constrained modulation optimization for
coded systems, pp. 372376.
Cui, S., Goldsmith, A. & Bahai, A. (2004). Energy-efciency of mimo and cooperative mimo
techniques in sensor networks, IEEE Journal on Selected Areas in Communications
22(6): 10891098.
Fitzek, F. & Katz, M. (2006). Cooperation in Wireless Networks: Principles and Applications,
Springers.
Haykin, S. & Moher, M. (2004). Modern Wireless Communications, Prentice Hall.
Laneman, J. N., Tse, D. & Wornell, G. (2004). Cooperative diversity in wireless networks:
Efcient protocols and outage behavior, IEEE Transactions on Information Theory
50(12): 30623080.
Matolak, D. W., Sen, I. & Xiong, W. (2006). Channel modeling for V2V communications, Proc.
Third Annual International Conference on Mobile and Ubiquitous Systems: Networking and
Services.
Peters, S. & Heath, R. W. (2008). Nonregenerative MIMO relaying with optimal transmit
antenna selection, IEEE Signal Processing Letters 15: 421 424.
Sohaib, S. & So, D. K. C. (2009). Asynchronous polarized cooperative MIMO communication,
Proc. IEEE 69th Vehicular Technology Conference pp. 15.
Sohaib, S. &So, D. K. C. (2010). Energy analysis of asynchronous polarized cooperative MIMO
protocol, Proc. IEEE 21st Personal, Indoor and Mobile Radio Communications Symposium.
Sohaib, S., So, D. K. C. & Ahmed, J. (2009). Power allocation for efcient cooperative
communication, Proc. IEEE 20th Personal, Indoor and Mobile Radio Communications
Symposium.
Tang, X. & Hua, Y. (2007). Optimal design of non-regenerative MIMO wireless relays, IEEE
Transactions on Wireless Communications 6(4): 1398 1407.
Wang, D. &Fu, S. (2007). Asynchronous cooperative communications with STBCcoded single
carrier block transmission, Proc. IEEE Global Telecommunications Conference pp. 2987
2991.
Wei, S., Goeckel, D. L. & Valenti, M. (2006). Asynchronous cooperative diversity, IEEE
Transactions on Wireless Communications 5(6): 1547 1557.
Boto Bako and Michael Weber
Institute of Media Informatics, Ulm University
Germany
1. Introduction
Vehicular ad-hoc networks (VANETs) enable promising new possibilities to enhance trafc
safety and efciency. The vision of VANETs is that vehicles communicate spontaneously,
in an ad-hoc manner over a wireless medium. Based on this inter-vehicle communication
(IVC), vehicles exchange important information, e.g., about road conditions and hazardous
situations. Moreover, such information can be propagated via multiple hops, thus making the
dissemination of important information possible over longer distances.
This is the key advantage of this kind of safety applications compared to conventional safety
systems. Whereas conventional safety systems only rely on information sensed in the direct
neighborhood by onboard sensors of a vehicle, active safety applications based on IVC can
utilize information generated by nodes multiple hops away. Moreover, such information can
be enriched on the way with information sensed by relaying cars. This greatly enhances the
potential of VANET applications. The advantage is twofold:
Having information about distant hazardous situations like an accident ahead or icy road,
the driver can be warned in-time, thus being able to completely avoid the dangerous
situation.
Aggregating information from multiple cars enables retaining information on a higher
semantic level. This way, applications like cooperative trafc jam warning and cooperative
parking place detection can be realized.
The enabling technology for such applications is the wireless ad-hoc communication between
vehicles. Especially the dissemination of messages in a specic geographic region represents a
fundamental service in VANETs to which we refer to as geographic broadcast (GeoCast). This
communication paradigmis used by many applications to enhance trafc safety and efciency
but it can also serve as a basic mechanism for other routing protocols. Because of its relevance
in the domain of vehicular networks, it is of key importance that the communication protocol
enables efcient message dissemination.
The realization of a robust and efcient broadcast mechanism is a challenging task due to
the wide range of applications envisioned to build upon this communication technology,
the rigorous requirements of safety applications, and the special network characteristics
of vehicular networks. Therefore, the main focus of this chapter is the efcient broadcast
of information for VANET applications. We want to give a broad and in-depth review of
recent research in this topic and present simulation results of efcient dissemination protocols
designed for such applications.
This chapter is organized as follows: In Section 2 we discuss briey different types of
VANET applications, followed by an overview of different communication mechanisms
Efficient Information Dissemination in VANETs
3
used by these applications. After that, the special network characteristics of VANETs are
discussed and the requirements of broadcast protocols are summarized. We conclude that a
geographically limited broadcast is one of the most important communication paradigms for
VANET applications. Therefore, this communication mechanism is surveyed and classied in
Section 3. This is followed by an evaluation of selected protocols by simulations in Section 4.
Finally, Section 5 concludes the results and presents possible future works on this topic.
2. Inter-vehicle communication
The main objective of this section is to dene the key requirements for IVC (with the
focus on broadcast mechanisms) in VANETs. Therefore we rst give an overview over
different application types with their characteristics, followed by a short description of
communication paradigms used to realize such applications. The identied properties of these
applications together with the special network characteristics of VANETs allow us to perform
a requirements analysis for the dissemination protocols.
2.1 VANET applications
There are many applications envisioned for VAENTs, in (Vehicle Safety Communications
Project [VSCP], 2005) e.g., more than 75 application scenarios were identied. A successful
deployment of such applications would result in a high benet. According to this benet the
VANET applications can be classied as follows (Bai et al., 2006):
Safety applications
Active safety applications represent the most important group of VANET applications. The
goal of these applications is to reduce the number of injuries and fatalities of road accidents.
In the European Union (EU27) e.g., more than 1.2 million trafc accidents involved injury
of passengers in 2007 and more than 42,000 accidents ended fatal (EURF, 2009). Hence,
there is a high potential benet in the implementation of such applications (VSCP, 2005).
To achieve this goal, safety applications disseminate information about hazardous situations
(e.g. about abnormal road conditions or post-crash warning) to vehicles which can benet
from such information to avoid an accident. Thus, this kind of applications rely mostly on
the dissemination of information into a specic geographic region, therefore, a geographically
limited broadcast or so called GeoCast is used (Bai et al., 2006; Bako et al., 2008; Slavik
& Mahgoub, 2010). It is also important to note, that safety applications are delay critical, i.e.
it is essential that the information is disseminated immediately without any delay. A special
case of this broadcast is the one-hop MAC-layer broadcast also called beaconing which is
periodically sent to neighbors in communication range to exchange information (e.g. position,
velocity) sensed by own sensors. This information can be used for safety applications like
cooperative collision warning.
Convenience (trafc management) applications
This kind of applications intends to improve the driving efciency and comfort on roads
by means of communications. Driving efciency applications intend to optimize the trafc
ow on roads, i.e. minimizing the travel time by disseminating information about trafc ow
conditions on roads. Therefore, cars periodically exchange information, combine information
received from neighboring vehicles with information sensed by own sensors, aggregate them,
and disseminate the newly gathered information for other vehicles. This way, applications like
a cooperative trafc jam detection or travel time estimation for road segments can be realized.
A driver having e.g. information about a trafc jam in advance, can choose an alternative
route (with the assistance of the car navigation system), thus reducing greatly the travel
time. Therefore, a successful deployment of such applications could greatly reduce the fuel
consumption of vehicles, which would have a great impact on cost reduction as well as on
the reduction of CO
2
emission. Comfort applications assist the driver in many situations, e.g.
for nding free parking spots or merging into the ow trafc and so on. These applications
are delay-tolerant, i.e. they dont impose such tight time constraints as safety applications, but
mostly need to exchange information periodically into the direct neighborhood.
Commercial applications
Commercial applications provide communication services like entertainment, web access and
advertisement. Examples are remote vehicle diagnostics, video streaming, and map download
for the navigation system. In contrast to the previously discussed types of applications,
commercial applications mostly rely on unicast communication and require a much higher
bandwidth than the two other application groups.
A more detailed classication with application examples can be found in (Schoch et al., 2008).
In the next subsection we briey overview the different communication mechanisms used to
implement the presented application types.
2.2 Communication paradigms
Considering the three type of application classes from a network perspective, it can be
stated that the broadcast/GeoCast, beaconing and unicast communication paradigms are the
building blocks for these applications. According to (Bai et al., 2006), broadcast can be further
divided into the event-driven, scheduled, and on-demand sub-classes.
Event-driven broadcast is used by delay-critical applications like road hazard condition
warning, and the information is disseminated over multiple hops into a specic geographic
region. Scheduled broadcast is used for applications like cooperative collision warning and
most of the trafc management applications. Information needed by such applications is
exchanged periodically and sent as MAC-layer broadcast only to neighbors in communication
range. Applications which require a multi-hop dissemination need to apply some efcient
aggregation techniques to overcome the limited bandwidth problem (c.f. (Dietzel et al., 2009a;
Dietzel et al., 2009b)). Unicast is important especially for commercial and entertainment
applications.
!!
Fig. 1. Example for GeoCast communication.
Figure 1 shows the GeoCast communication paradigm. A broken down vehicle (marked red
on the left side of the image) initiates a broadcast message about the hazardous situation. This
message is disseminated via multiple hops to inform all vehicles in the specied destination
region. Beaconing is shown in Figure 2. The vehicle marked red in the gure sends a
MAC-layer broadcast message with data about the own vehicle, like position, heading, and
47 Efficient Information Dissemination in VANETs
Fig. 2. Example for Beaconing.
S
D
Fig. 3. Example for unicast communication.
velocity. This message is received by all vehicles in communication range and is not further
forwarded. The last communication paradigm, unicast communication, is shown in Figure 3.
A message is routed hop by hop from a sender (S) to a destination (D).
As already stated, active safety applications are the most important class of envisioned
applications and they offer a high potential benet. Because the majority of these applications
rely on broadcast, we can conclude that the broadcast/GeoCast communication paradigm is
of eminent importance for a successful application of VANETs. Moreover, broadcast is a basic
service used also for route discovery in many reactive unicast protocols like DSR (Johnson
& Maltz, 1996), AODV (Perkins & Belding-Royer, 1999), and LAR (Ko & Vaidya, 2000).
Therefore, the objective of this chapter is the evaluation of this type of vehicle-to-vehicle
communication.
2.3 VANET characteristics
Before going into the details of broadcast protocols, we briey summarize the VANET
characteristics. These characteristics, together with the discussed application classes, reveal
some important requirements onto the communication protocols.
High mobility
Vehicles on highways potentially travel at very high speeds. Thus, the communication period
between these vehicles can be very short. Moreover, high node velocities cause more frequent
topology changes which result in outdated neighbor tables. Thus, when a protocol relies on
such a table, the forwarding decision may be incorrect due to old or nonexistent entries in this
table.
Dynamic topology
Characteristic for VANETs is a very high dynamic network topology. The reason therefore
is twofold: First, the node density ranges from very sparse and partitioned networks e.g. on
rural freeways or late night hours to very dense networks at rush hours and trafc jams. Thus,
the number of neighbor vehicles in transmission range can vary from zero, up to hundreds
of nodes. Second, node mobility can range from static nodes in trafc jams up to very high
velocities on free highways. This implies that a routing protocol has to overcome with sparse
and partitioned as well as dense scenarios, which are subject to rapid changes over time due
to node mobility.
Wireless communication
The dissemination of information in a VANET is based on a wireless mediumwhich represents
an error-prone and scarce resource in the network. Especially in dense scenarios where many
cars compete for the wireless medium, the limited bandwidth constitutes a severe problemfor
the routing protocol. Therefore, an efcient broadcast protocol is of eminent importance for a
successful deployment of VANET applications.
Delay constraints
As outlined in the previous subsection, most of the safety applications are delay-critical. This
means, they rely on broadcast mechanisms which allow the dissemination of information
with minimal delay. Thus, a broadcast mechanism designed for such applications has to
forward safety critical information immediately, without introducing any delay e.g. for
routing purposes.
Geographical addressing
In VANETs a geographical addressing method is used and it ts for most envisioned
applications. This means that a broadcast is not performed network wide (which is simply not
possible due to the potential size of several million nodes in the network), but is limited by
a geographic region (GeoCast). Similarly, unicast protocols make use of position information
available via GPS receiver for route decisions.
Mobility patterns
Another important aspect of VANETs is that vehicle movements are constrained by the road
topology. This means, node movements obey mobility patterns imposed by the road network.
Thus, node mobility is predictable and can therefore be utilized by routing protocols to
enhance the dissemination performance. Roughly three main classes of movement patterns
can be distinguished, which directly inuence the degree of predictability of node movements:
inner city roads, rural roads, and highways.
Beaconing
The presence of up-to-date neighborhood information is a prerequisite for many VANET
applications (e.g. for cooperative collision warning). This information is exchanged by
periodical one-hop MAC-layer broadcast messages, so called beacons. This information can
be used by a broadcast protocol to enhance the rebroadcast decision, without introducing
additional communication costs.
Pseudonym change
By communicating information, vehicles reveal personal information which results in a
sever privacy problem. To solve this problem, vehicles are supposed to communicate using
pseudonyms which they change at a given frequency. By changing its pseudonym, a node
may be inserted into the neighbor table multiple times under different pseudonyms. Having
such incorrect neighbor entries, routing decision cannot be met correctly anymore. Thus,
pseudonym changes may heavily affect the underlying protocol if it uses neighborhood
information.
Characteristics Implications
High Mobility Outdated neighborhood information and short
communication periods.
Dynamic Topology High variance in network density and node velocity:
partitioned networks vs. trafc jams.
Wireless Communication Limited bandwidth and error-prone wireless
communication.
Delay Constraints Messages need to be broadcast immediately, without
introducing any delay.
Geographical Addressing Position information of vehicles are needed and GeoCast
is an important communication mechanism for safety
applications.
Mobility Patterns Protocols can benet from predictable mobility patterns to
enhance their routing decisions.
Beaconing More network load, but protocols can benet from
information exchanged by this basic service.
Pseudonym Change Pseudonym changes result in incorrect neighbor tables.
Thus, pseudonym changes introduce a new challenge to
VANET protocols which rely on neighborhood information.
Table 1. VANET characteristics and their implications.
Table 1 summarizes the discussed VANET characteristics together with their implications.
Based on these implications, a more exhaustive requirements analysis can be done in the next
subsection.
2.4 Requirements analysis
The diversity of VANET applications and the special network characteristics impose several
requirements to the broadcast protocols. These requirements are deduced from the previous
subsections and summarized in the following.
Scalability
The broadcast protocol has to cope with very dense networks like trafc jams in order to
enable correct operation of safety applications in such scenarios.
Effectiveness
The broadcast protocol has to assure that all nodes (or a percentage of nodes, dened by the
application) in the destination region receive the disseminated information.
Efciency
Due to the limited available bandwidth, the broadcast protocol needs to eliminate message
redundancy. This is achieved by minimizing the forwarding rate, but still achieving a
reception of a message by all nodes in a specic geographic region. This helps to avoid the
broadcast storm problem (Ni et al., 1999) and enables the coexistence of multiple VANET
applications.
Dissemination delay
Safety applications require the immediate relaying of information, without the introduction
of any delay.
Delay-tolerant dissemination
Because vehicular networks are subject to frequent partitioning, it is desirable to cache
information in such scenarios and propagate them later when new vehicles are available in
the vicinity. Otherwise important information can be lost when the network in the destination
region is not fully connected.
Robustness
The communication over the wireless medium is error-prone, nevertheless, the broadcast has
to cope with packet losses in order to assure the correct function of vital safety applications.
It has to be noted that not all requirements can be met to a full extent because some
requirements are contrary. So, for example, when minimizing the forwarding ratio to achieve
a high efciency, the requirement robustness cannot be fullled anymore because relaying
nodes represent a single point of failure in this case. Thus, when a relaying node fails to
forward a message (which is probable due to the wireless nature of the communication
channel) the overall reception rate can also drop signicantly. Therefore, in most cases an
elaborate tradeoff between such requirements is needed.
3. Review of broadcast protocols for VANETs
As we have seen, due to the diversity of VANET applications and the special network
characteristics, the design of an efcient broadcast protocol is a challenging task. The simplest
way to implement a broadcast mechanism is the use of nave ooding. In ooding every node
rebroadcasts a message exactly once (given that it is located inside the destination region),
thus the message is ooded into the whole region. The downside of simple ooding is that
this mechanism is very inefcient. Given the limited bandwidth of the wireless medium,
an inefcient information dissemination scheme like nave ooding leads to redundancy,
contention, and collision, to which is referred to as the broadcast storm problem (Ni et al.,
1999). To overcome these problems (and the ones identied in the previous sections), many
improved broadcast protocols were proposed by the research community.
In this section we rst introduce a classication of different broadcast approaches and discuss
them. After that, we review novel broadcast mechanisms which were designed to perform
well in highly dynamic environments like VANETs. We not only consider VANET protocols
here, but also cover protocols for mobile ad-hoc networks. Thus, we present an up-to-date and
broad discussion of broadcast protocols from multiple domains.
3.1 Classication of broadcast protocols
One of the rst in-depth classication of broadcast approaches was done by Williams and
Camp in (Williams & Camp, 2002). They identied four main classes: Simple Flooding,
Probability Based Methods, Area Based Methods, and Neighbor Knowledge Methods. More
recent works on this topic overtake this thorough analysis and rene it with new properties
(cf. e.g. (Heissenbttel et al., 2006; Khelil, 2007; Yi et al., 2003)). Although such a categorization
is useful to evaluate and discuss the properties of broadcast protocols belonging to different
classes, it has a signicant drawback. Such an exclusive differentiation into rigid classes is not
practicable for many broadcast protocols. We argue that many protocols are not belonging
to one xed class, but combine the properties from different classes. This was also stated by
(Slavik & Mahgoub, 2010) and we call them therefore Hybrid Broadcast Protocols.
In the following we give an overview over basic attributes of protocols, which dene key
characteristics. Knowing such attributes together with their implications, it allows a more
thorough analysis of the properties of a protocol. For example, we dont consider area based
methods as an attribute class (in contrary to many other classications in the literature)
because it only tells how the rebroadcast decision is calculated (based on the additional
coverage), but gives no information about the protocols properties. To compute the additional
coverage, atomic information like position and distance are needed, but it can be also
deduced fromtopology information. If a protocol uses such information, then exact properties
can be determined like complexity, weaknesses and strengths. Therefore we consider such
information as key attributes which are used in our classication.
Probabilistic
In this scheme, a node rebroadcasts a message with a certain probability. This probability can
be xed a priori (Static Gossip) or adapted dynamically (Adaptive Gossip). In their pure form,
probabilistic schemes are very simple and stateless (no need for neighborhood information).
They have moderate efciency but are robust to packet losses due to their probabilistic nature.
Topology based
Topology based protocols use neighborhood information (e.g. 1-hop or 2-hop) to calculate
the rebroadcast decision. Such information needs to be exchanged periodically (by so called
beacon messages) at a frequency depending on nodes velocity. This results in higher
communication overhead due to the periodical exchange of beacon messages but allows
on the other hand very efcient rebroadcast decisions. In dynamic networks this kind
of protocols may degrade in performance with increasing node velocity due to outdated
neighbor information.
Position/distance based
By using position information, the rebroadcast decision can be calculated more accurately in
some cases. E.g. the rebroadcast probability could be adjusted based on the distance to the
sender or relays can be selected in a VANET based on their positions.
Local decision
In local decision protocols, a node decides itself on reception to rebroadcast the message or
not. This is the contrary of imposed decision and is a desired property of protocols especially
in highly dynamic environments like VANETs, because this way rebroadcasts can be decided
locally, thus decoupling sender from receiver, which results in a more robust protocol.
Delayed rebroadcast based
This class of protocols introduces a delay before rebroadcasting a message dened by a
delay function (randomly or according to some property of the node like distance to the
sender). The delayed rebroadcast is useful when nodes overhear the communication channel
and gather information about rebroadcasts from other nodes, upon that a more efcient
rebroadcast decision can be taken. An example for this type of protocols is the Dynamic
Delayed Broadcasting (DDB), introduced by (Heissenbttel et al., 2006). We consider this
mechanism to improve the broadcast performance as orthogonal to other techniques. Thus, it
can be combined with other mechanisms, and therefore, we dont consider them separately
in this work.
Clustering, in contrary to other classications is not considered as a basic attribute of
protocols, but is more an aggregation of other properties. A standard clustering scheme
utilizes normally topology information to build the clusters and clusterheads utilize the
imposed decision scheme to designate the relays. This holds also for more advanced clustering
schemes, thus they utilize a combination of the key protocol classes dened above.
3.2 Deterministic broadcast approaches
A subclass of topology based broadcast protocols are the imposed decision protocols, where a
sender species in the broadcast message which neighbors have to perform a rebroadcast. We
refer to this type protocols as deterministic broadcast approaches. Deterministic approaches
explicitly select a small subset of neighbors as forwarding nodes which are sufcient to reach
the same destinations as all nodes together. Therefore, a relaying node has to know at least its
1-hop neighbors. As nding an optimal subset (i.e. with minimal size) is NP-hard, heuristics
are used to nd not necessarily optimal but still sufcient relaying nodes.
These type of protocols were one of the rst ones suggested by the research community
to minimize the broadcast overhead, thus to overcome the broadcast storm problem.
Characteristically these protocols achieve a very high efciency, because based mostly on
2-hop neighborhood information, very accurate rebroadcast decisions can be calculated.
Therefore, many variants of deterministic broadcast protocols can be found in the literature.
Examples of deterministic approaches are dominant pruning (Lim & Kim, 2000), multipoint
relaying (MPR) (Qayyum et al., 2002), total dominant pruning (Lou & Wu, 2002), and many
cluster based approaches (see e.g. (Wu & Lou, 2003) and (Mitton & Fleury, 2005)).
Despite the high efciency they offer, deterministic broadcast has a signicant disadvantage:
relaying nodes represent a single point of failure. If a relay fails to forward a message (e.g. due
to wireless losses, node failure, or not being in transmission range due to mobility) then the
overall reception rate of the message may drop signicantly. Thus, these kind of protocols lack
robustness and perform poorly in dynamic environments like VANETs. Therefore, they cant
be used for safety critical applications in VANETs and more robust but at the same time also
efcient broadcast schemes are needed.
3.3 Probabilistic broadcast approaches
One of the early probabilistic approaches to improve ooding is static gossiping, which uses a
globally dened probability to forward messages (Chandra et al., 2001; Haas et al., 2006; Miller
et al., 2005). All these variants work best if the network characteristics are static, homogeneous,
and known in advance. Otherwise they result in a low delivery ratio or a high number of
redundant messages. To overcome these problems, adaptive gossiping schemes have been
developed.
Haas et al. (Haas et al., 2006) introduced the so called two-threshold scheme, an improvement
for static gossiping based on neighbor count. A node forwards a message with probability
p1 if it has more than n neighbors. If the number of neighbors of a node drops below this
threshold n then messages are forwarded with a higher probability p2. The obvious advantage
of this improvement is that in regions of the network with sparse connectivity messages are
prevented to die out because the forwarding probability is higher than in dense regions.
(Haas et al., 2006) also describes a second improvement which tries to determine if a message
is dying out". Assuming a node has n neighbors and the gossiping probability is p then this
node should receive every message about p n times from its neighbors. If this node receives a
message signicantly fewer, the node will forward the message unless it has not already done
so.
In (Ni et al., 1999), Ni et al. introduced the Counter-Based Scheme. Whenever a node receives
a new message, it sets a randomly chosen timeout. During the timeout period a counter is
incremented for every duplicate message received. After the timeout has expired, the message
is only forwarded if the counter is still below a certain threshold value.
Although all these adaptations improve the broadcast performance, they still face problems in
randomnetwork topologies. For example, if a node has a very large number of neighbors, this
results in a small forwarding probability in all of these schemes. Despite this, there could e.g.
still be an isolated neighbor which can only receive the message from this node. An example
of such a situation is shown in Figure 4 (example taken from (Kyasanur et al., 2006)).
A
B D
C F
E G H
Fig. 4. Sample topology where static gossiping fails
When node A sends a message, all nodes in its neighborhood receive it. In this example
scenario only node E should forward it with the probability of 1 since E is the only node
that can propagate the message to node G. If the gossiping probability is only based on the
neighbors count, node E will be assigned a low probability since it has many neighbors. So
the broadcast message will die out" with a high probability and never reach G and all later
nodes. If the part of the network connected only via G is very large, the overall delivery ratio
will drop dramatically. Such situations can occur quite regularly in dynamic networks of a
certain density.
3.4 Hybrid broadcast approaches
As we have seen, deterministic broadcast approaches achieve a very high efciency but
they lack robustness. On the other hand, probabilistic approaches behave much better in the
presence of wireless losses and node failures, but have also other limiting disadvantages. E.g.
the adaptation of the forwarding probability to actual network condition is a challenging task
and is not solved adequately with simple heuristics. Therefore, recently novel probabilistic
broadcast approaches were proposed, which combine the strength of both protocol types,
becoming this way highly adaptive to the present network conditions. We call this type of
protocols hybrid broadcast approaches.
One of the rst hybrid broadcast approaches is the so called Smart Gossip protocol,
introduced by (Kyasanur et al., 2006). In smart gossip every node in the network uses
neighborhood information from overheard messages to build a dependency graph. Based
on this dependency graph, efcient forwarding probabilities are calculated at every node.
To ensure building up a stable directed graph, the authors make the assumption that there is
only one message originator in the whole network. This assumption may be sufcient in a few
scenarios, but especially in the case of VANETs this is not applicable, and therefore, as shown
in (Bako et al., 2008a; Bako et al., 2007) the performance of the protocol degrades massively in
such environments.
To overcome these problems, a novel hybrid probabilistic broadcast was introduced by
(Bako et al., 2007). In this so called Position based Gossip (PbG) 1-hop neighborhood
information are used together with position information of neighboring vehicles to build
a local, directed dependency graph. Based on this dependency graph efcient forwarding
probabilities can be calculated which adapts to current network conditions. PbGwas designed
for message dissemination only into one direction, e.g. for a highway trafc jam scenario,
where approaching vehicles have to be informed about the trafc jam. Thus, messages are
propagated only against the driving direction. This way only one dependency graph has to be
built, and therefore this protocol is denoted as the 1-Table version of PbG.
It is obvious that most VANET applications need to disseminate information in both directions
of a road and cannot be restricted only to one direction. For example at an intersection, we face
four road segments and therefore a message can be distributed in four directions. Therefore,
in (Bako et al., 2008b) a 2-Table version of the protocol was introduced, which ts much better
for general highway and intersections scenarios.
Furthermore, in (Bako et al., 2008) two more extension of the PbG protocol was introduced: a
network density based probability reduction and a fallback mechanism. The rst mechanism
reduces the forwarding probability in dense networks, thus reducing the broadcast overhead,
at the same time achieving similar reception rates as the original protocol. The second
extension aims to prevent message losses: Acommon problemin wireless networks represents
the so called hidden station problem. Because MAC layer broadcast frames are used,
techniques like RTS/CTS cannot be used to avoid this problem. Especially in very dense
networks the hidden station problem has a signicant impact on the performance of the
protocol. In such cases, the packet loss rate increases and application level requirements
for the delivery ratio cannot be fullled any more. To overcome this problem, the second
enhancement tries to determine if a message is dying out". The enhancement works as
follows. Each node receiving a new message initializes a counter which is incremented every
time it overhears the same message being forwarded by some other node. If the counter
is below a certain threshold after a xed period, the message is rebroadcast with the same
probability as if it was received for the rst time.
A more general gossip protocol similar to PbG was introduced in (Bako et al., 2008a). In this
so called Advanced Adaptive Gossip (AAG) protocol two-hop neighborhood information are
used to calculate forwarding probabilities similar to PbG. Thus, no position information are
needed, which may be imprecise or even not available in some cases. Moreover, this protocol
is not limited to any road topology. Furthermore, this protocol was enhanced by a message
loss avoidance mechanism in (Schoch et al., 2010), which is similar to the fallback mechanism
from (Bako et al., 2008). With this extension the protocol becomes much more robust and is
therefore called robust AAG, or short RAAG. In the mentioned work also benecial properties
of RAAG considering security are discussed and evaluated.
4. Evaluation
In this section we evaluate the performance of selected protocols in different scenarios.
Because the simulation of all protocols is very time consuming, we selected one representative
protocol for each protocol type discussed in Section 3 and evaluate the impact of mobility,
node density, and high broadcast trafc on these schemes. Therefore, we rst introduce the
simulation parameters and describe the two evaluated scenarios: city and highway. After that,
we showthat deterministic broadcast schemes are heavily affected by node mobility, thus they
are inapplicable for VANETs. The remaining subsections present the results of the selected
hybrid broadcast schemes in a highway and city scenario. For comparison we include also the
results of nave ooding and static gossiping. Results of the following protocols are presented:
Multipoint Relaying (Qayyum et al., 2002)
Flooding
Static Gossiping (Chandra et al., 2001; Haas et al., 2006)
Advanced Adaptive Gossiping (AAG) (Bako et al., 2008a)
Robust Advanced Adaptive Gossiping (RAAG) (Schoch et al., 2010)
4.1 Simulation setup
For the evaluation of the broadcast protocols we use the JiST/SWANS (Barr et al., 2005)
network simulator, including own extensions. JiST/SWANS provides a radio and MAC-layer
according to IEEE 802.11b. This is close to the IEEE 802.11p variant, which is planned for
vehicular communication. On the physical layer the two-ray ground model is used together
with the additive noise model, thus, the effect of packet collisions can be investigated. The
radio transmission power is set to achieve a wireless transmission range of 280 meters. For the
city scenario a eld size of 1000m x 1000m is used, whereas the simulations for the highway
scenario are run on a 25m x 3000m eld. Node density is varied from 10 up to 300 nodes, thus
comparing sparse as well as dense scenarios.
Parameter Value
Field City: 1000m x 1000m, Highway: 3000m x 25m
Simulation Duration 120s
Broadcast Start 5s
Pathloss Tworay
Noise Model Additive
Transmission Range 280m
Beaconing Interval 1s
Number Messages 3 Messages per node, max 150
MlA Acknowledges 1
MlA Replay Delay 2.5s
MlA Last Replay Offset 100s
Placement Random
Static Node Speed: 0
Random Waypoint Node Speed City: 3 20 m/s, Highway: 22 41 m/s
Highway Mobility Node Speed Highway: 0 30 m/s
Table 2. Simulation setup parameters.
The number of broadcast messages depends on the node density: Every node generates one
broadcast message per second (with a minimal payload), limited by a maximumcount of three
messages per node. The absolute number of broadcast messages is limited by 150. Thus, in a
scenario with 10 nodes 30 messages are initiated, whereas in scenarios with 50 or more nodes
150 messages are created (if not otherwise specied). This way we evaluate the protocols
under low as well as under heavy network load. To hold the neighbor tables up-to-date
beacons are used which are exchanged with a rate of 1 beacon per second. The beacon size
depends on the information required by the broadcast protocol. Thus with AAG and MPR the
entire neighbor list is sent in a beacon, whereas in Flooding only a message with minimal size
is sent (we assume this is required by the VANET applications).
A setup is simulated over 120s, where the broadcast of messages starts at 5 seconds. For the
RAAG protocol, the message loss avoidance (MlA) mechanism is congured to await at least
one acknowledge for a sent message, otherwise the message is rebroadcast again once (if new
nodes are present in the neighborhood), with a delay of 2.5 seconds. Messages have a timeout
of 100s and if a message was not yet acknowledged at least once, the message is rebroadcast
one more time.
To evaluate the impact of node mobility on the performance of the broadcast protocols we use
three different mobility models:
Static
Random Waypoint (RW)
Highway Mobility (HM)
The static model is used to measure the protocols performance in a best-case scenario, i.e.,
nodes didnt move at all, thus all neighborhood information are up-to-date. With the Random
Waypoint mobility model a worst-case scenario is investigated where nodes move in arbitrary
directions. A more realistic scenario is provided by the Highway Mobility model, which is an
own extension inside the JiST/SWANS framework. With this mobility model cars move in the
same direction on a 4-lanes highway with randomspeeds. They hold a safety distance to other
cars, change lanes and pass slower cars if necessary. At the end of the simulated highway the
lanes are blocked by 4 cars, thus trafc congestion is simulated here. The exact parameters
used for our simulations can be found in Table 2.
According to (Bani Yassein & Papanastasiou, 2005), the optimal xed probability for static
gossip is 0.7. Therefore, we use this value for the static gossip protocol in our evaluations. For
each simulation setup 20 simulation runs are done and the results averaged.
4.2 Effect of node mobility on deterministic broadcast
Multipoint Relaying (MPR) was selected as a representative for deterministic protocols to
evaluate the impact of node mobility onto this protocol type. Therefore, a highway scenario
with three different mobility models is used: static, random waypoint, and highway mobility.
Because MPR lacks robustness, and therefore the number of broadcast messages heavily
inuences the performance of the protocol, we also simulated a scenario where only one
broadcast message is initiated (Static 1). The other three simulation congurations (Static 2,
RW, and HM) use the normal parameters described in 4.1.
Figure 5 shows the results of this evaluation. As we can see, in sparse networks (10 and 25
nodes) the reception rates in all four simulation setups are very low. These results are as
expected, because the network is partitioned and therefore not all nodes can be reached by a
broadcast without additional mechanisms. With higher node densities and only one broadcast
message per simulation (Static 1), MPR achieves quite good reception rates. With 100 and 150
the reception rate is almost 100%and drops slightly with increasing nodes, but stays over 90%
which is an acceptable ratio. This slightly decline is due to the higher overhead introduced by
the beacon messages.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
Static 1
Static 2
RW
HM
0
0.05
0.1
0.15
0.2
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
Fig. 5. Performance of MPR in a highway scenario with different mobility models and
number of messages.
On the other hand, with a high number of broadcast messages (Static 2), the reception rate
drops signicantly in higher node densities. With 300 nodes MPR achieves only around 70%
reception rate with is clearly unacceptable for safety critical VANET applications. Thus, these
results show that heavy network load has a signicant inuence onto deterministic protocols.
Now considering mobility, we can see that with the random waypoint and highway mobility
model the reception rate drops even more drastically. With both mobility models in almost
all node densities the reception rates are around 50%. Thus, deterministic approaches are
inapplicable for dynamic environments like VANETs.
Regarding the forwarding rates, we can see that MPR is highly efcient, needing only around
3% or less rebroadcasts with 300 nodes. Thus, we can conclude that deterministic broadcast
approaches are highly efcient but cant meet VANET requirements in the presence of mobility
and high network load.
4.3 Hybrid broadcast approaches in a highway scenario
In this subsection we evaluate two hybrid broadcast protocols (AAGand RAAG) in a highway
scenario and compare the results with ooding and static gossip (SG). Figure 6 shows the
results for this scenario with static nodes. As we can see, in a partitioned network like with 10
nodes in these results, the reception rates of all four protocols are almost identical. Whereas
with 25 nodes (here the network is also not completely connected), static gossip already has a
signicant lower reception rate of around 10%. This gap is even bigger with 50 nodes, where
static gossip has a reception rate of around 57% compared with 83% of RAAG. This is because
the static gossip probability of 70%, which is too low for sparse networks.
With higher densities, AAGsignicantly drops regarding the reception rate, reaching not even
70% of other vehicles for the 300 node setup. Here static gossip and ooding achieve better
reception rates, both protocols are slightly under 90%. However, RAAG clearly outperforms
the other protocols, reaching almost 100% reception rates.
Regarding the forwarding rates, we can see that ooding has the highest forwarding rates
except for the scenario with 10 nodes. Here the message loss avoidance mechanism of RAAG
generates more overhead, but has not much impact onto the reception rate because the nodes
are static. The rebroadcast rate of ooding is way too high in higher densities, and that is a
serious problem causing the so called broadcast storm. We will discuss this effect later in a
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 6. Performance of hybrid broadcast approaches in a static highway scenario.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 7. Performance of hybrid broadcast approaches in a highway scenario using the random
waypoint mobility model.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 8. Performance of hybrid broadcast approaches in a highway scenario using the highway
mobility model.
scenario with higher network load. AAG achieves the best forwarding rate, but as we saw,
the performance is insufcient for this scenario. Static gossip has a lower forwarding rate as
RAAG with few nodes, but remains constant slightly about 60% with higher node densities.
Thus, static gossip doesnt scale well with increasing node density. On the other hand, the
forwarding rate of RAAG decreases constantly with increasing density and is constantly
around 10% higher as AAG due to the message loss avoidance mechanism.
Figure 7 and 8 show the same scenario with random waypoint and highway mobility
models. As we can see, there is almost no difference in the reception and forwarding rates
compared with the static scenario. This means, that all these protocols are not affected at
all by node mobility. This is a very important property which makes these protocols well
suited for VANETs. The only difference compared with the static scenario is the reception and
forwarding rates of the RAAG protocol in low densities. Due to node mobility, the cached
messages are here physically transported and rebroadcast later. Thus, RAAG manages to
overcome network partitions and achieves a much higher (at a cost of more rebroadcasts)
reception rate.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 9. Performance of hybrid broadcast approaches in a highway scenario under high
message load using the highway mobility model.
In the next simulation setup we evaluate the performance of these protocols under high
network load. Therefore, we increased the payload of broadcast messages to 512 bytes and
raised the limit of the absolute number of messages to 300. This means, every node creates
exactly 3 messages, with a rate of one message per second. The results for this simulation
setup are shown in Figure 9. As we can see, AAG and ooding cant cope with increasing
network load, thus the reception rate is dropping signicantly, reaching almost only 50% of
the nodes in the 300 node setup. The reception ratio of static gossip also declines constantly
with increasing node densities. Thus, these protocols are not scalable and cant be used for
VANET applications in such scenarios. Only RAAG manages to reach good reception ratios
in the tested setup, and as can be seen, it clearly outperforms the other protocols. Thus we can
conclude, that RAAG allows an efcient and effective dissemination also in scenarios with
extreme high network load. The forwarding rates can be compared with the other results.
AAG, ooding, and static gossip have lower forwarding ratios due to the packet losses.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 10. Performance of hybrid broadcast approaches in a static city scenario.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
R
e
c
e
p
t
i
o
n

R
a
t
e
Nodes
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 50 100 150 200 250 300
F
o
r
w
a
r
d
i
n
g

R
a
t
e
Nodes
AAG
Flooding
RAAG
SG
Fig. 11. Performance of hybrid broadcast approaches in a city scenario using the random
waypoint mobility model.
4.4 Hybrid broadcast approaches in a city scenario
For the city scenario we simulate a eld of 1000m x 1000m with static and random waypoint
mobility. Figure 10 shows the results for the static scenario. As we can see, the results are
similar to the static highway scenario. RAAG achieves the best reception rates for all node
densities, reaching almost 100% with 50 and more nodes. The reception rates of the other
protocols drop constantly with increasing nodes, and reach only around 80% with 300 nodes.
This is clearly not sufcient for critical safety applications in VANETs. The forwarding rates
are also similar to the previous scenario: ooding and static gossip have very high forwarding
rates and these rates dont scale well in contrary to RAAG and AAG.
Considering the mobile city scenario shown in Figure 11, we can here also conclude that
mobility has almost no effect on these protocols. Except for the RAAG protocol, where the
message loss avoidance mechanism positively benets from nodes movements. In highly
partitioned networks, like with 10 nodes in this gure, RAAG manages to achieve a reception
rate of around 30% higher than the other protocols, or RAAG itself in a static scenario. This is
a signicant gain and these results underline the need of a message loss avoidance mechanism
for partitioned networks.
5. Summary and outlook
In this chapter we gave an overview over possible VANET applications and showed different
communication paradigms used for such applications. We also pointed out the importance
of broadcast mechanisms for active safety applications. This was followed by an overview of
the special network characteristics of VANETs. From that, we deduced a set of requirements
for broadcast protocols which have to be fullled for a successful deployment of VANET
applications.
Also a classication of broadcast protocols was introduced which enables a more systematic
analysis of broadcast mechanisms. Based on this, we have reviewed state-of-the-art broadcast
protocols designed for inter vehicle communication. The main focus here was on hybrid
protocols, which combine positive properties of more protocol classes and offer thereby
promising characteristics for broadcast applications in vehicular networks.
The theoretical evaluations were conrmed by extensive simulations. We have shown that
deterministic protocols are heavily affected by node mobility and network load, and they
are therefore not suitable for VANET applications. Furthermore, we have shown that pure
ooding, as well as static gossip, is not scalable, i.e. they cause the so called broadcast
storm problem. Thus, with increasing node density and network load their performance drop
signicantly and they are therefore unfeasible for VANETs.
On the other hand, the RAAG achieves very promising results in sparse as well as in dense
networks. We have shown that the message loss avoidance mechanism yields a signicant
performance gain in sparse scenarios and increases the robustness of the protocol also in
dense networks. Moreover, RAAG is not affected by node mobility which is a very desirable
property of VANET protocols. Thus we can conclude that RAAGis predestinated for dynamic
networks like VANETs and satises the requirements in such networks also in the presence of
critical safety applications.
Although the presented results are very promising, there are some issues we want to address
in future work. First of all, RAAG requires 2-hop neighborhood information which generates
more overhead. We aim to reduce this required knowledge to 1-hop neighbors, similar to the
PbG protocol but in a more general way. Moreover, we have to evaluate the performance
of RAAG in the presence of pseudonym changes, which may have a signicant effect on
broadcast protocols. Also a detailed evaluation of the message loss avoidance mechanism in
partitioned networks and its optimization could result in a signicant gain in delay-tolerant
networking.
6. References
Bai, F., Krishnan, H., Sadekar, V., Holl, G. & Elbatt, T. (2006). Towards characterizing
and classifying communication-based automotive applications from a wireless
networking perspective, In Proceedings of IEEE Workshop on Automotive Networking
and Applications (AutoNet), San Francisco, USA.
Bako, B., Kargl, F., Schoch, E. & Weber, M. (2008a). Advanced Adaptive Gossiping
Using 2-Hop Neighborhood Information, IEEE Globecom 2008 Wireless Networking
Symposium (GC08 WN), New Orleans, USA.
Bako, B., Kargl, F., Schoch, E. & Weber, M. (2008b). Evaluation of Position Based Gossiping
for VANETs in an Intersection Scenario, 4th International Conference on Networked
Computing and Advanced Information, Gyeongju, Korea.
Bako, B., Rikanovic, I., Kargl, F. & Schoch, E. (2007). Adaptive Topology Based Gossiping in
VANETs Using Position Information, 3rd International Conference on Mobile Ad-hoc and
Sensor Networks (MSN 2007), Beijing, China.
Bako, B., Schoch, E., Kargl, F. & Weber, M. (2008). Optimized Position Based Gossiping
in VANETs, 2nd IEEE International Symposium on Wireless Vehicular Communications,
Calgary, Canada.
Bani Yassein, Ould Khaoua, M. M. & Papanastasiou, S. (2005). Improving the performance
of probabilistic ooding in manets, Proceedings of International Workshop on Wireless
Ad-hoc Networks (IWWAN-2005), London UK.
Barr, R., Haas, Z. J. & van Renesse, R. (2005). JiST: an efcient approach to simulation using
virtual machines: Research Articles, Softw. Pract. Exper. 35(6): 539576.
Chandra, R., Ramasubramanian, V. & Birman, K. (2001). Anonymous Gossip: Improving
Multicast Reliability in Mobile Ad-Hoc Networks, Technical report, Ithaca, NY, USA.
Dietzel, S., Bako, B., Schoch, E. & Kargl, F. (2009). A fuzzy logic based approach for
structure-free aggregation in vehicular ad-hoc networks, VANET 09: Proceedings of
the sixth ACM international workshop on VehiculAr InterNETworking, ACM, New York,
NY, USA, pp. 7988.
Dietzel, S., Schoch, E., Bako, B. & Kargl, F. (2009). A structure-free aggregation framework
for vehicular ad hoc networks, 6th International Workshop on Intelligent Transportation
(WIT 2009), Hamburg, Germany.
URL: http://www.kargl.net/docs/mypapers/2009-03-wit.pdf
EURF (2009). European road statistics 2009, Technical report, The European Union Road
Federation.
URL: http://www.irfnet.eu
Haas, Z. J., Halpern, J. Y. & Li, L. (2006). Gossip-based ad hoc routing, IEEE/ACM Trans. Netw.
14(3): 479491.
Heissenbttel, M., Braun, T., Wlchli, M. & Bernoulli, T. (2006). Optimized Stateless
Broadcasting in Wireless Multi-hop Networks, Proceedings of the 25th Conference on
Computer Communications (IEEE Infocom 2006), Barcelona, Spain.
Johnson, D. B. & Maltz, D. A. (1996). Dynamic source routing in ad hoc wireless networks,
Mobile Computing, Kluwer Academic Publishers, pp. 153181.
Khelil, A. (2007). A Generalized Broadcasting Technique for Mobile Ad Hoc Networks, PhD thesis,
Universitt Stuttgart.
Ko, Y.-B. &Vaidya, N. H. (2000). Location-aided routing (lar) in mobile ad hoc networks, Wirel.
Netw. 6(4): 307321.
Kyasanur, P., Choudhury, R. R. & Gupta, I. (2006). Smart Gossip: An Adaptive Gossip-based
Broadcasting Service for Sensor Networks, pp. 91100.
Lim, H. & Kim, C. (2000). Multicast tree construction and ooding in wireless ad hoc
networks, MSWIM 00: Proceedings of the 3rd ACM international workshop on Modeling,
analysis and simulation of wireless and mobile systems, ACM, New York, NY, USA,
pp. 6168.
Lou, W. &Wu, J. (2002). On reducing broadcast redundancy in ad hoc wireless networks, IEEE
Transactions on Mobile Computing 1(2): 111123.
Miller, M. J., Sengul, C. & Gupta, I. (2005). Exploring the Energy-Latency Trade-Off for
Broadcasts in Energy-Saving Sensor Networks, icdcs pp. 1726.
Mitton, N. & Fleury, E. (2005). Efcient broadcasting in self-organizing multi-hop wireless
networks, ADHOC-NOW, pp. 192206.
Ni, S.-Y., Tseng, Y.-C., Chen, Y.-S. &Sheu, J.-P. (1999). The Broadcast StormProblemin a Mobile
ad hoc Network., Proceedings of the 5th annual ACM/IEEE international conference on
Mobile computing and networking, MobiCom 99, ACM, New York, USA, pp. 151162.
Perkins, C. E. &Belding-Royer, E. M. (1999). Ad-hoc On-Demand Distance Vector Routing, 2nd
Workshop on Mobile Computing Systems and Applications (WMCSA 99), New Orleans,
USA, IEEE Computer Society, pp. 90100.
Qayyum, A., Viennot, L. & Laouiti, A. (2002). Multipoint Relaying for Flooding Broadcast
Messages in Mobile Wireless Networks, p. 298.
Schoch, E., Bako, B., Dietzel, S. & Kargl, F. (2010). Dependable and secure geocast in
vehicular networks, VANET 10: Proceedings of the seventh ACM international workshop
on VehiculAr InterNETworking, ACM, New York, NY, USA, pp. 6168.
Schoch, E., Kargl, F., Leinmller, T. & Weber, M. (2008). Communication Patterns in VANETs,
IEEE Communications Magazine 46(11): 28.
Slavik, M. & Mahgoub, I. (2010). Stochastic broadcast for vanet, CCNC10: Proceedings of the
7th IEEE conference on Consumer communications and networking conference, IEEE Press,
Piscataway, NJ, USA, pp. 205209.
VSCP (2005). Task 3 nal report: Identify intelligent vehicle safety applications enabled by
dsrc, Technical report, Vehicle Safety Communications Project, U.S. Department of
Transportation.
Williams, B. & Camp, T. (2002). Comparison of Broadcasting Techniques for Mobile Ad Hoc
Networks, MobiHoc 02: Proceedings of the 3rd ACM international symposium on
Mobile ad hoc networking & computing, ACM Press, New York, USA, pp. 194205.
Wu, J. & Lou, W. (2003). Forward-node-set-based broadcast in clustered mobile ad hoc
networks, Wireless Communication and Mobile Computing 3: 155173.
Yi, Y., Gerla, M. & Kwon, T. J. (2003). Efcient ooding in ad hoc networks: a comparative
performance study, Proc. IEEE International Conference on Communications ICC 03, Vol.
2, pp. 10591063.
4
Reference Measurement Platforms for
Localisation in Ground Transportation
Uwe Becker
University Braunschweig, Institute for Traffic Safety and Automation Engineering
Germany
1. Introduction
Although the importance of satellite based localisation has been rising in the last years to a
significant level in safety related applications like transportation systems, it seems obvious
that saturation of sales volume for those systems has not been reached yet.
Due to the development and set up of the European satellite based localisation system
GALILEO, new services and accuracy in localization open new market sectors. GALILEO
will provide five different services with different performances and characteristics that will
be suitable for different ranges of applications. These services will be: Open Services, Safety
of Life Services, Commercial Services, Public Regulated Service and Rescue Service. The
Safety of Life Service is the key service for most safety-related applications due to its
guaranteed characteristics integrity, availability and accuracy.
A new important aspect for the usage of these services is their formal approval by a safety
authority. Regarding this approval some intensive analysis has to be undertaken regarded
from the angle of safety and reliability.
Considering this analysis, following leading points are of eminent importance
- possibilities for the (practical) analysis of the reliability and availability of the
localisation information provided by a receiver
- set up of a regulatory framework in combination with specific procedures for the
evaluation according to the requirements of the safety authorities.
For the field of transportation it is important to know, that there exist multiple safety
authorities: (nearly) each mode of transportation (road, rail, water, air) has at least one of
them. In the past it was not necessary to set up a domain spanning approval for components
because it was not of economic interest. In the future this interest will emerge from new
technological possibilities. According to this, the present situation may change with the
upcoming GALILEO system because of its new functionalities and accuracy: at least the key
component of all applications will have to be used in all modes of transportation this will
be a receiver for the satellite signals.
The two aspects mentioned above will be discussed. In the section Reference Localization
Platforms, the systems for the analysis for the availability and accuracy of the satellite
signals by means of a special reference localization platform will be discussed. Figure 1
shows the generally concept. The first results on the topic of the fusion/structural
comparison of regulatory frameworks will be subject of the second section Regulatory
Frameworks.

66

Vehicle
Measurements platform
Referece measurement system
System for qualification
Control and
supervision centre
Evaluation unit
GNSS signal generator
Test-Parcours
Measurements area

Fig. 1. Generally concept for the measurements area
2. Reference localisation platforms
Satellite-based positioning systems are used in many applications for air, maritime and land
transport. In either case, at least four distinct satellites are used to obtain a four dimensional
position, consisting of three coordinates in space and one in time (fourth coordinate). This
position can be (almost) anywhere on the surface of the earth or in the airspace above it
(Grewal et.al., 2001).
When used on the surface of the earth, the reception of the necessary amount of satellites
(namely four) can be difficult due to environmental barriers (e.g. buildings, trees, etc.) in
close range of the object to be localised. This problem arises, because quasi-optical wave
propagation occurs in the frequency range used for satellite positioning. If an object
obstructs the necessary direct line of sight to the satellite, no signal can be received (Hnsel
et al., 2005). This fact reduces the availability of satellite based positioning in places
shadowed by other objects, which cannot be avoided in railway environment.
For the usage in railway systems, a high reliability is necessary for the use in safety related
functions (e.g. the train control system). On the other hand, the implementation of the
reference positioning system is simplified by the special domain inherent constraint that the
vehicle cannot leave the track.
For the experimental evaluation of the availability and accuracy, two generic reference
localisation platforms have been set up (Becker et.al., 2008a). Because of the focus to ground
based transportation (i.e. road and rail transportation) this section is confined to two generic
platforms CarLa and CarRail which were developed at the Institute for Traffic Safety
and Automation Engineering (iVA)
2.1 Road based traffic
CarLa (Car Laboratory) is a test bed for several sensors and control algorithms. The basic
hardware setup of CarLa is depicted in Figure 2.
The sensors used in the platform can be classified into the ones for the GNSS localization
system and the ones for environmental perception which are used in the control algorithms.
Reference Measurement Platforms for Localisation in Ground Transportation

67

Fig. 2. Measurement Platform for Road Application
The GNSS localization system consists of a GNSS receiver which is coupled with a rubidium
oscillator as an extremely precise clock.
The environmental perception is represented by a special tracking concept and additional
sensors for the measurement of the vehicles orientation.
The central component of the system is a small but powerful computer, running the Linux
operating system. All the sensors are connected to this by appropriate hardware interfaces
(e.g. CAN-bus, Ethernet, serial interface).
The computer offers two displays for different purposes. One is placed near the drivers seat
(where usually the navigation system is located) and the other one is placed in such a way
that an operator, sitting on the back seat, can see it. The first one is a touch screen, so that
input can easily be made by the driver (if it is necessary for the application); the second
display is equipped with a standard keyboard and mouse.
Because of the use of a classical operation system and hardware that is near to a standard
desktop PC, the development of the software is quite easy and fast.
The tracking concept is based on a reference track which is installed into the ground via
equally spaced magnets. These sensor signals are used to control the vehicles dynamics in
lateral and longitudinal direction.
The intention of the CarLa reference platform is to guarantee a position and velocity control
of high accuracy to the reference track. The attained accuracy of the lateral control is +/- 1
cm whereas the one of the longitudinal control is +/- 0.1 m/s inside the range from 0 up to
100 kph.
Therefore, ensuring high accurate positions of the track magnets, the GNSS measured
position can be compared to the reference position, which is estimated via the recognition of
the track magnets and the vehicles complete state vector.
Figure 3 shows an example of a measurement run. Shown is the error of measurement of
the GNSS antenna position from the calculated reference antenna position as lateral
deviation e
y
in metre.

68

Fig. 3. Error measurement of lateral deviation in driving direction
2.2 Guided traffic
For the evaluation of satellite based applications in rail transport a generic reference
measurement platform has been set up as well, CarRail (Figure 4).
The platform uses two different sensor systems together with an precise map of the track.
The first sensor used is a Doppler radar sensor for a continuous measuring of the relative
position along the track. This sensor has the disadvantage of a relatively high drift (0.2%)
that has to be stabilised.
The RFID-based absolute positioning is used as second sensor to stabilise the drift of the
radar sensor. For the RFID system, transponders are located along the track, which
represent absolute marks. The positions of the transponders (or at least some) are known
with high accuracy (about 0,5 cm) and they are unambiguously identifiable. Their positions
are stored in an XML based electronic track map to obtain information about the matching
between the topology of the track and the (geographic) positions of the transponders.
The first test track has been equipped with this reference measurement system. It is situated
near the Braunschweig main railway station. This track is 3 km long and includes 8
switches, straight sections as well as curved sections. This topographical constellation is
suitable for tests of track selectivity. The topographically constellation with a bridge and a
big manufacturer building near the track provides areas with critical reception of satellite
signals. Hence, it is well suited to test the accuracy and availability for safety related
applications in rail transportation.

69
RFID transponder
doppler radar
sensor
RFID receiver
read antennas
sensor data fusion
ethernet
RS485
position-informations
RFID transponder
electronic track map
ethernet
satellite based
application
data processing
display com. unit
GNSS antenna

Fig. 4. Measurement Platform for Rail Application
Currently, the accuracy of the reference measurements platform is 50 cm. With a modified
arrangement of the transponders, i.e. a modified distance between two transponders, it will
be possible to achieve an accuracy of about 15 cm at maximum speeds of up to 200 kph. By
the combination of these two sensors (together with the map), a continuous positioning with
a very accurate absolute position information is set up.
3. Regulatory framework
The spectrum of possible applications for the vehicle and the rail reference platform is quite
wide. We are focusing on questions on the accuracy and the availability of positioning
systems for safety related applications in road traffic. In road transport these will become
important in the field of advanced driver or traffic assistance systems that have the
possibility to intervene in traffic maneuvers or even to outvote the driver. Hence a
certification will be obligatory for the systems (Becker et.al., 2008b). The reference platforms
can be used for the investigation of the positioning systems.
For the transportation domain(s), the field of systems and applications regarding safety
responsibility will use the GALILEO based localization. A complete set of regulations exist
for the certification of safety related applications due to the European focus for safety
improvement in transportation. For the upcoming satellite based systems, adequate
regulations have to be set up as well. The existing ones have to be kept or adopted to keep
them interoperable with existing systems and for the purpose of migration.
Instead of a simple approach by setting up GALILEO concerning requirements for each
domain of transportation, the authors propose an intermodal (or domain spanning)
approach, so that those requirements occurring in all domains may be separated and taken
as a generic kernel of requirements to be the basis of a first step of a certification system,
containing two stages. The first stage will concern the domain independent certification as
stated above. The second one is domain dependant and contains those special requirements
which are specific for each domain.

70

A
v
i
a
t
i
o
n
R
a
i
l
R
o
a
d
Galileo
(ref. Receiver)
Interface
M
a
r
i
t
i
m
e

Fig. 5. Galileo and its Application Domains
For the transportation domain(s), the field of systems and applications regarding safety
responsibility will use the GALILEO based localization. A complete set of regulations exist
for the certification of safety related applications due to the European focus for safety
improvement in transportation. For the upcoming satellite based systems, adequate
regulations have to be set up as well. The existing ones have to be kept or adopted to keep
them interoperable with existing systems and for the purpose of migration.
Instead of a simple approach by setting up GALILEO concerning requirements for each
domain of transportation, the authors propose an intermodal (or domain spanning)
approach, so that those requirements occurring in all domains may be separated and taken
as a generic kernel of requirements to be the basis of a first step of a certification system,
containing two stages. The first stage will concern the domain independent certification as
stated above. The second one is domain dependant and contains those special requirements
which are specific for each domain.
This splitting into two stages avoids the tests of the first stage to be done multiple times
when certifying equipment for more than one domain. This opens the opportunity to
suppliers to broaden their market by having the stage one done and being able to undergo
domain specific tests for lower costs.
Within the approach, formal techniques from the field of internet technology (semantic web
resp. ontology) are used to develop models of the different domain languages and the
processes required for the certification. This allows the building of relationships between the
(same) processes and different terminologies from the different domains. By using this
relationship, experts from different domains can view the processes under the eyeglasses
for their respective domain, which helps them to understand the processes. As a result, the
communication between the experts from different domains can be improved and
discussions can be freed from terminological discussions (due to misunderstandings) and
accelerated to bring results. Another advantage is the increase in confidence in the
certification from different domains, because the processes can be seen and compared with
those still understood by the experts (DIN 1319-1).

71

Fig. 6. Methodological approach for the integration of different domain specific
terminologies with process descriptions
This methodological approach is implemented in new projects funded by the European
Commission, European space agency and German federal ministry of research and
education. The promising framework for certification processes shown here is still under
development and the paper will show the current status of the project and it will be show on
one GNSS based application for low density secondary lines.
Todays standards and regulations are formulated as texts in natural language. This is
adequate for documents being used in one domain (having its own domain specific
language e.g. terminology) and one country (having its own natural language). Different
domains often have languages (terminologies) that are incompatible with each other
because they use identical terms for different concepts. This leads to misunderstanding
between incorporated persons from different domains and makes a joined work very hard.
To overcome this problem, the terminologies used have to be described in a way known by
all incorporated persons to be understood easily. If this common language is a formal one,
formal methods and techniques can be applied to check the results for consistency and for
correctness. Different possibilities exist for formal descriptions. In Figure 7, a petrinet is
shown, that describes several processes defined in the standard IEC 17000 for the
conformity assessment (DIN EN ISO/IEC 17000).
For the approach to be used in certification, the terminologies used as well as the processes
to be applied have to be described formally. After having identified this, new standards or
advancements can be specified, by using inheritance like mechanisms (as speaking in
terms of object oriented methods). By that formal description, the concepts and structures
contained by the documents are formulated in an explicit and unambiguously way.

72

accreditation
5.6
testing
4.2
inspection
4.3
work
2.10
participation
2.11
sampling
4.1
Peer assessment
4.5
attestation
5.2
suspension
6.2
appeal
6.4
complaint
6.5
withdrawal
6.3
recognition
7.5
acceptance
7.6
audit
4.4
konformity
assessment
system
2.7
access
2.9
conformity
assessment
results
konformity assessment
2.1
participant
2.10
specified
requirement
3.1
decision
candidates for an
aggreement
group
Validity of
conformity
assessment
results
unilateral
arrangement
7.7
bilateral
arrangement
7.8
multilaterale
Vereinbarung
7.9
applied rules
member
2.11
management of
systems or
programs
product
3.3
procedure
3.2
general
requirements
conformity
assessment
body
2.5
accreditation
body
2.6
statement of
conformity
review
5.1
designating
authority
7.3
designation
7.2
scope of
attestation
5.3
system
person
proces position
transaction
3.3
sample
characte-
ristics
product design system
records
statements of
fact or
relevant
information
object of
conformity
assessment
conformity
suitability,
adequacy and
effectiveness of
selection and
deternination
Supplier of an
object of
conformity
assessment
organisation
representatives
of other bodies in
an aggreement
group
specified
requirements
third-party
claimant

Fig. 7. Formal description of processes defined in the standard IEC 17000 by means of a petrinet
From modelling existing documents (standards, regulations ) from different domains, the
common core can be identified more easily to be extracted for the use in the mode
independent certification.
An additional advantage of having the formal description is, that the translations to natural
languages can be realised in an easy way and avoids the usage of different terms for the
same concept which can lead to confusion or misunderstanding.
3.1 Modelling the contend of standards
The contents of normative documents can be modelled according to the approach of
ontological modelling described above. Two methods were developed (Hnsel, 2008). One
describes the retrospective modelling of existing normative documents to clarify the
contents. The starting point for this method is an existing technical standard. A second
method describes the modelling of new standards to avoid ambiguities right from the
beginning of the development of standards.
Both methods are similar in the setting up of the different parts required for the overall
model according to the ontological modelling as described above. The means of description
has to be selected in advance. The resulting ontology of the application domain will be
modelled as a network of concepts with their terms and relations, e.g. a taxonomy.
As an example for the modelling of a new normative document, Figure 8 shows a part of a
taxonomy that was set up for the certification approach of GNSS-receiver for different
transportation modes.

73

Fig. 8. Part of a taxonomy describing the certification of GNSS-receivers

Fig. 9. Part of a process for the certification approach of GNSS-receiver

74
With the application of the ontological modelling, the clarity of the definitions is improved
compared to the classical method of just writing natural language text. In Figure 8 the
results of the OntoClean analysis are included in the diagram in brackets.
A part of the process using the terms from the application domain is shown in Figure 9.
Here the means of Petrinets is used for the process modelling. This explicit modelling of the
processes helps the reader to understand the meaning or the intention of the author and
improves clarity.
4. Concepts of metrology
Two aspects are describes that will become important as soon as safety related assistance
The GNSS-System is a complex dynamic system. Its satellites and onboard receivers move
continuously. Many different sources of influences in terms of metrology can be identified,
which result in several GNSS system errors. Especially, a lot of principles of terminology
have to be analysed in detail to be compliant with the existing standard documents.
Besides the exact understanding of the standards in metrology, i.e. ISO 5725 and GUM, for
an intended certification, the metrology domain has to be linked to the requirements from
the application domain, for which the certification is intended. Such requirements are the
framework, in which the observed values have to fit in.
To understand the concepts of metrology and to be able to communicate with other persons
about this, a formal modelling, as described in the previous chapters, is very helpful. One
aspect needing clarification is the existence of multiple concepts termed as result of
measurement in the standards. In natural language texts it is common, that the same term
is associated with different concepts. However, the formal description makes the meaning of
the (intended) concepts clearer.
Figure 10 shows a preliminary result of a conceptual and formalised analysis of the
standards ISO 5725 and GUM based on UML class diagrams and explains the relationship
between the individual terms.
As can be seen, each measurement has a true value. The true value is by nature
indeterminate. Uncorrected observations P
IND,k
are obtained by measurement. An
uncorrected arithmetic mean of observations includes uncorrected arithmetic mean of
observations P
IND
and standard uncertainty of the uncorrected mean
PIND
. A corrected
result is a result of a measurement after correction for systematic error. The correction P
and the uncertainty of the correction
P
are determined by a calibration.
5. Conclusion
Two aspects are describes that will become important as soon as safety related assistance
systems in rail or road based transportation will integrate satellite based localization (esp.
GALILEO). These are, on the one hand, adequate reference localization systems for the
realization of measurements of accuracy and availability of the location information
provided by satellite systems. On the other hand, the regulatory framework will be
discussed with the focus on a domain spanning approach, where the regulations from
multiple domains have to be integrated. In addition to the description of standards, the
approach is applied to the field of metrology, where, in some cases, it is also of high
importance to exactly specify what is meant when doing measurements and evaluating the
results especially for the safety cases.

75

systematic error
-the corrected result
combined standard uncertainty
-indeterminate by nature
true value
error (of measurement)
-the uncorrected result
-result of a measurement
uncorrected arithmetic mean of observations
standard deviation
-the indication
uncorrected observations
random error
-predetermined
-result from calibration
uncertainty of correction
-predetermined
-result from calibration
correction
1
1
1
1
-the corrected result
corrected arithmetic mean of observations
-particular quantity
measurand
-the uncorrected result
standard uncertainty
1
1
1
1..*
recognized systematic error
unrecognized systematic error
1
1
1
1
A complete statement of the result of a measurement
k IND
P
,
) (
, k IND
P S
IND
P
IND
p
C
P
P
P
C
P
U S
E
,
R S
E
,
S
E
R
E
E
t
P

Fig. 10. Formalised taxonomy of the standards ISO 5725 and GUM by UML class diagram
representation
6. Acknowledgement
This contribution results from a project, which has been supported by the German Federal
Ministry of Economics and Technology (BMWi) under grant no. 50NA0614 and 50NA0615
which is gratefully acknowledged.

76
7. References
Becker, U.; Hnsel, F.; Hbner, M.; Poliak, J.; Schnieder, E.; Weng, Y.; Zhou, Y.(2008a).
Dynamic validation of satellite based positioning systems using reference
measurement platforms, Tagungsband der POSITIONs 2008, Gesamtzentrum fr
Verkehr Braunschweig e. V., Dresden
Becker, U.; Beisel, D.; Hnsel, F.; Poliak, J.; Schnieder, E. (2008b). Approach for the
Certification of safety related Satellite Based Transport Application across modes
ITS World Congress 2008, ITS Congress Association, New York
DIN 1319-1. Grundlagen der Messtechnik Teil 1: Allgemeine Grundbergriffe. Beuth-Verlag,
Berlin.
DIN EN ISO/IEC 17000: Conformity assessment Vocabulary and general principles, Beuth-
Verlag, Berlin
Grewal M. S., Weill L. R., Andrews A. P. (2001). Global Positioning Systems, Inertial Navigation,
and Integration, John Wiley & Sons, ISBN 0-471-35032-X, New York
Hnsel, F. (2008). Towards formalization of technical standards Technical University of
Braunschweig, Institute for Traffic Safety and Automation Engineering, PhD-
dissertation.
0
Coupling Activity and Performance Management
with Mobility in Vehicular Networks
Miguel Almeida
1
and Susana Sargento
2
1
Nokia Siemens Networks, Universidade de Aveiro
2
Instituto de Telecomunicaes, Universidade de Aveiro
Portugal
1. Introduction
We live in a mobile, fast paced world, where users are constantly on the move. Transportation
plays a major role in this matter. Users thrive for services being promptly delivered anytime
and anywhere. Nevertheless, business models still focus around Content Service Providers
(CSP) and Network Service Providers (NSP), who, as trusted entities, provide more than the
connection, further focusing, as time evolves, on the service delivery capitalization. As the
trends start to position these providers as the relay points for the information to be conveyed
into 3rd party cloud services, the delegation of management functions is also outsourced to
the 3rd party entities. It is in this view that the remote management of vehicles becomes of the
utmost importance, since connectivity allows the delivery of novel services built around the
monitoring of the vehicles conditions, location and user preferences. The immediate benets
would result in presence/location awareness for retrieval of additional information of the
surroundings, or even mechanical support, mechanical failure prediction or detection, based
on the continuous monitoring of the vehicles hardware sensors, as well as a whole plethora of
new advantages, propelled by the collection of performance and behavior information.
Vehicular networks are inherently associated with high mobility scenarios and this
fact introduces new requirements. Usually associated with high velocity patterns, the
requirements to support these networks are mainly positioned around the enabling of fast
mobility management protocols, and hence interfaces, gifted with the extensibility potential
for the exchange of additional information. Furthermore, when considering vehicular
scenarios, network mobility and efciency are two crucial features which need to be kept
in mind at all times. They have special inuence over the choice of the protocol used to gather
information from the vehicles towards the network. These requirements lead us to consider a
framework that was originally designed for the management of the mobility of the terminals,
and which therefore supports mobility with a high efciency ratio in terms of resource
consumption. This framework, the IEEE 802.21 Media Independent Handovers (802.21-2008,
2009), contains functionalities and elements that can be extended with advanced reporting
capabilities to provide seamless reporting in heterogeneous technologies and environments.
Using IEEE 802.21, it is also possible to integrate the actions of reporting with the actions
of network decisions enforcement. We show that this approach provides a signicant set
of functionalities not achieved with current approaches, while reducing the overhead on
cross-layer reporting. The typical approach is to perform such procedures above the IP layer.
5
Besides reducing the overhead, gathering performance and action reports at lower layers also
saves on signaling and simplies the protocol stack. When bringing mobility into the picture,
these concerns become even more crucial.
Knowing that different types of devices have different groups of requirements in terms of
network, hardware and applicational capabilities, the primitives with which all of them
interface should be the same: a common Application Programming Interface (API) which is
mobility driven and that cleanly exposes management functions (already under evaluation in
current research) for seamless mobility and reporting. Management frameworks today also
introduce, as a requirement, the denition of interfaces to 3rd party entities. We consider the
central management entity to be a cloud of functionalities and of centralized intelligence,
which allows interfaces for other cloud services, thus empowering the CSPs with new
advanced services and new ways of capitalizing the management functions. By considering
the several existing approaches, we derive a solution which combines the most commonly
used web services in order to provide a Cloud viewof the performance of the several vehicles.
In this chapter we present a solution to collect performance and behavior related information
from different communication layers of the vehicles, while keeping in mind the major
requirements associated with the inherent properties of the technology which will be
discussed along with relevant use cases. Performance management represents a topic which
is not widely covered when dealing with vehicular networks, and very little information
can be found regarding this subject in the literature. Most research is being conducted in
topics related with Vehicular networks focus on mobility management and communication
techniques. It is our main goal to evaluate the performance penalties introduced by several
layers: hardware, network, session and application. Service performance can be evaluated at
any layer without depending on a specic technology, while enabling media independent
service reports. It is in this context that vehicles, connected via multiple network access
technologies, will report user activities (user behavior related), performance metrics of the
mechanical hardware, of the network and of the applications running.
The chapter is organized as follows. Section 2 describes the relevant approaches to deal
with the collection of performance information and user behavior activities. Section 3
presents the architecture and the main functional entities to support the mobile user vehicle
reports integrated with network reconguration trigering. Section 4 depicts the performance
comparison of the reporting approach against existent mechanisms, both on a qualitative and
quantitave basis. Finally, section 5 presents the most important conclusions from the chapter.
2. Background
This section details the relevant approaches to deal with the collection of performance
information and user behavior activities. It details several means to gather information and to
present it in a cloud oriented solution (Voas & Zhang, 2009). While considering that vehicles
are moving and exposed to different environments, different contexts and different conditions,
information can be extracted and conveyed into a platform which, Extracts, Transforms and
Loads (ETL), processes it according to predened metrics, or Key Performance Indicators
(KPI), and allows the management parties to evaluate the performance and to take actions.
The proposals presented bellow represent efforts in trying to bring the devices closer to the
cloud in terms of performance management features. The following subsections detail the
technologies which are employed (Section 2.1), compare and contextualize the approaches
to provide a clear view of their adequacy in terms of offered features and usage drawbacks
(Section 2.2), and present the current reporting architecture and KPIs (Section 2.3).
There is already some work related to the interconnection of the devices with cloud services
for monitoring and management purposes. These studies are, however, not extensive. Most
of the efforts were related to sensors as the main analysis use cases, which are not necessarily
mobile. Mobile devices introduce additional concerns, since mobility requires maintaining
connectivity upon movement. Even in those scenarios, management related issues are still
little explored, as stated in Gurgen & Honiden (2009). In Gurgen & Honiden (2009), the
authors provide the major requirements for the denition of a platform to manage such
devices. In Jung et al. (2007), the work is more focused on a security aware, technology
agnostic framework, using Simple Network Management Protocol (SNMP) (J. Case, 1990)
to gather information into a command center. This last work is more connected with our
proposal, but our list of requirements goes beyond security, having mobility on the top of
the list.
Another major concern is the management of the devices in the Cloud, i.e., in an online
distributed tool, which appears to the end users as a centralized Graphical User Interface
(GUI). This vision on the management of devices is more related to the concept of the
Internet of Things (IoT), in which each vehicle can be seen as a thing. In Mohinisudhan
et al. (2006), SNMP is used to incorporate hybrid automobiles with a performance monitoring
system. Johansson et al. (2005) underlines the usage of a Controller Area Network (CAN)
in the automotive market. CAN is a serial bus communications protocol with the purpose
of interconnecting sensors, actuators, controllers and other elements. It denes the physical
and data link layers for an efcient and reliable communication between the entities. In
Johansson et al. (2005) it is presented an integration example with a passenger car, a truck,
a navy boat and a spacecraft. In this work the authors also describe the concept of CAN
gateways, which provide a way to integrate CAN-based networks with other networks
and protocols. This approach is useful in the context of coupling vehicular devices with a
performance management platform, since it allows the integration of industry deployed lower
layer mechanisms (very oriented to specic vehicle parts sensors) with network management
solutions such as the one here presented.
Extensible Messaging and Presence Protocol (XMPP) (Peter Saint-Andre, 2009) was created
for user communication purposes and has already been used for device integration with
the cloud, even if only as a protocol capable of interconnecting sensors in an asynchronous
wireless environment (Hornsby et al., 2009) (without yet being used for the IoT potential it
carries). More recent work, (namely Miguel Almeida (2010a)), takes into account requirements
for remote management and its procedures. This approach will be further evaluated in the
upcoming sections, since it employs a mechanism to easily integrate devices into the cloud.
Although Miguel Almeida (2010b) work does not focus on mobility, since it is merely the
denition of a framework and of the required extensions to support enhanced reporting
capabilities, it denes extensions and allows us to use them to couple reporting and vehicular
device management along with mobility. It takes a more lower layer approach to deal with
the problem we are solving and, because of that, it will also be detailed in the sections bellow.
Next we detail the technological solutions that are used in the aforementioned proposals.
79 Coupling Activity and Performance Management with Mobility in Vehicular Networks
2.1 Description of the involved technologies
In this subsection we describe the technologies involved in the process of collection
information from devices, rst detailing an array of data collection mechanisms which are
considered the most relevant. Then, we present the trend in the protocols used in the web
environments.
2.1.1 Data collection mechanisms
Simple Network Management Protocol (SNMP) (J. Case, 1990) is one of the most relevant
data collection mechanisms with wider acceptance, and which is represented by a large
scale adoption in a multitude of scenarios. In SNMP the information is collected from the
agents in the managed devices according to the meta-data detailed in the Management
Information Bases (MIBs), from where the information can be polled. MIBs group parameters
that are accessible via SNMP. The SNMP agents and stations use a request/reply
protocol to communicate which supports standard messages (Get-Request, Get-Response,
Get-Next-Request, Set-Request and Trap). The SNMP station uses Get-Request to solicit
information from the SNMP agent, which answers with a Get-Response message. SNMP has
been evolving over time with increased security and efciency. Also, one important aspect is
the addition of an unsolicited mechanism via Traps. SNMP-Trap is an unsolicited message
sent by SNMP agents to the manager. These messages inform about the occurrence of a
specic event, and can be used to inform that a link is down or that the agent is reinitializing
itself. Traps allow for reactiveness and simplify scenarios where polling is not the best option.
Remote Monitoring (RMON) (Waldbusser, 1995) extends this concept by introducing probes
and, instead of measuring Network Elements (NE), it focuses more on the analysis of trafc
ows. This approach is particularly useful for the identication of third party services or
servers, troubleshooting the network problems, security breaches or simply keeping logs of
user activities for accounting or proling.
The Common Management Information Protocol (CMIP) (J. Case, 1990) provides a complete
network management framework over many, diverse network machines and computer
architectures. CMIPs mode of operation differs from SNMPs, in the sense that the latest was
designed for simplicity and ease of implementation. Besides the same functions provided by
SNMP, CMIP contains more functionalities, thus allowing a wider range of operation sets. In
this framework, any relevant information can be requested from the managed object and can
be interpreted according to the managing system. Amain drawback of CMIP is its complexity,
and therefore, its adoption did not fall in the networking enviornment.
Call Detailed Records (CDR) (Breda & Mendes, 2006) include information of the call duration
and failure causes, and are generally used with some lightweight data mining processes to
withdrawimmediate conclusions. They are largely used by cellular operators to performsome
minimal proling computation in their business intelligence solutions. CDRs are typically
generated on a per-call basis: each call can originate a CDR. Although originally the CDR
was designed to describe call details for billing purposes, it can be used to trace the call
at the business level and retrieve service assurance relevant information. This information
complements the Performance Management information by extending the network behavior
analysis to the service/subscriber scope, providing the ability to propose new analysis
scenarios (e.g. to assess if network is accurate in the service delivery, or which services
are more suitable for that network considering the trafc model and user behavior). In
order to support our requirements of supporting seamless reporting through inter-technology
environments, CDRs would need to be extensively expanded leading to a high increase in
the overhead. The three most common procedures to collect CDR comprise: (1) the real-time
transfer of a single CDR each time a call occurs; (2) the near-to-real-time transfer which
takes place after several CDRs have been grouped into a single Blocked Generation Log and
subsequently sent via Event Forwarding Discriminator (EFD); (3) and the collection of CDR
records being stored in File Generating Log.
Other approaches, such as Mobile QoS Agents (Soldani, 2006), installed in the mobile devices,
are also being implemented to gather end-user related information. The agents are executed
remotely and gather a limited set of performance metrics that provide information to derive
the Quality of Experience (QoE), as they are physically near the users. Other proprietary
protocols over IP are also being implemented, where the data structure is XML, and the
meta-data is dened by the hardware manufactures and interpreted by the performance
monitoring solutions with knowledge on the specications. The agents can gather very
specic metrics depending on the device or on the analysis use case, and thus these solutions
should be seen as very implementation dependent and customizable. Their usage can be
applied to functions such as measure the user feedback from an application (via a pop
up questionnaire), substitute road-tests by measuring signal power and SNR to evaluate
coverage, analyze network metrics (throughput, delay, bit rates over time), estimate location
and deduce trends.
The IEEE 802.21 Media Independent Handovers (MIH) framework deals with the exchange
of information for mobility management in heterogeneous environments. This information
includes: events, which typically deal with the changes on the link layer level and which may
prompt for handover; commands, which serve the management purpose by indicating control
information about handovers; and information messages, which provide details on the status
of the extended services of the network and information on the available networks.
Fig. 1. IS Server Location
Media Independent Information Services (MIIS) were dened to support various Information
Elements (IEs), which can be used to provide further information for handover decisions. Fig.
1 shows how the MIH Users can interact with the MIH Function (MIHF), and how the Service
Access Points (SAPs) are implemented in order to allowcommunication with the lower layers.
2.1.2 Integrating with the webcloud
Up to now, the presented approaches are focused on ways to exchange information
using existent technologies that were primarily created for that particular purpose. The
requirements now invoke wider concerns, namely those related with the gathering of
information into a cloud of services. In this sense, from lower layer related protocols we now
move into solutions which were dened to integrate computers connected to the Internet,
which is the de-facto environment for the cloud based applications, and the substratum for all
presentation layers.
The typical model for communicating with the cloud, or web service guideline, is based on
Representational State Transfer (REST) interfaces over HTTP (Fielding, 2000). REST provides a
clear interaction model that enables a powerful and exible solution through simple interfaces
in a scalable environment. REST and Simple Object Access Protocol (SOAP) (Don Box, 2000)
are associated to HTTP as means to convey information understandable by the cloud. SOAP
provides the support of objects over HTTP; however, SOAP faces a scalability issue because it
usually requires a large amount of technology to establish bidirectional invocation: it usually
requires an HTTP web server, coupled with an application server to enable the web service
environment. Moreover, current trends resort to REST over HTTP due to its simplicity and
ease of usage given the mapping with of the HTTP methods (GET, PUT, DELETE and POST).
It also provides a long-lasting interface that is not coupled with the business logic behind the
interface. When deploying these mechanisms, manufacturers are concerned about assuring a
future proof solution, and hence look at the choices which grant them a more secure bet on
the long term.
XMPP (P. Saint-Andre, 2004) offers good conditions as a transport protocol for applications
within the web services (WS) scope since it offers reliability, synchronous and asynchronous
delivery of messages, and does not require a complex set of features such as WS-Routing and
WS-Referral to ensure identity trace back (Fabio Forno, 2005) within private domains, since
addressing is not only IP based. XMPP was conceived as an alternative Instant Messaging
protocol, but has been evolving to a broader concept. Given the fact that it is open and XML
based, it became easy extensible and became an IETF standard.
2.2 Taking advantage of the existing approaches
In this section we evaluate the several approaches to deal with the problem of collecting
performance management and behavior related data, and presenting it in the cloud - a place
which is spatially and software distributed, but which creates an abstraction to present a
centralized logic to the users accessing it. Users access a GUI which hides all the hardware and
software complexity behind it. Bellow are two major contributions that provide an answer to
this problem and which will be evaluated and used. The rst handles performance collection
on higher layers and takes advantage of existing solutions at the applicational layer, namely
using web oriented protocols, that are user oriented. The second provides a MAC layer
solution for the gathering of information using a protocol which was originally conceived for
the management of the devices mobility. We will take advantage of both solutions as inputs
for the denition of the architecture presented in this chapter. The way both are integrated is
explained in Chapter 3.
2.2.1 Using XMPP and REST
As stated, one good approach for collection of performance related information is to perform
it in a web oriented environment. Using XMPP and REST (Miguel Almeida, 2010a) brings
advantages in terms of collecting user information. According to (P. Saint-Andre, 2004), there
are three Core Stanza types dened by XMPP: The <message/>, <presence/> and <iq/>. The
rst works as a push mechanism to immediately send messages if the destination is online.
Presence relies on publish-subscribe mechanisms through which nodes inform the server of
their availability (e.g.: online, away, do not disturb), and is usually distributed among the
other nodes in the roster. The last one is a stanza responsible for entities making requests and
receiving responses (hence Info/Query) from each other for management, feature negotiation
and remote procedure call invocation.
One of the biggest advantages of XMPP is the fact that the addresses can be associated
with people or devices such as computers, mobile phones, sensors, routers or cellular
network elements (3GPP RAN and Core Network Elements). This is achieved by the
use of a Jabber ID (JID), a uniquely addressable ID, which is a valid Uniform Resource
Indicator (URI) (Berners-Lee & Masinter, 1998), created according to the following
format: person@domain/resource, where person usually represents the user entity; domain
represents the network gateway or "primary" server to which other entities connect for XML
routing and data management capabilities; and resource, which is of special interest since it
allows to identify a specic device associated with the person. Security can be achieved by
using Transport Layer Security (TLS) for channel encryption, while authentication is achieved
through Simple Authentication and Security Layer (SASL). Regarding the portability and
interoperability requirements, XMPP uses the "over-IP" approach and allows the binding of
resources to streams for network-addressing purposes. This feature also allows to perform
Identity Management via the relationships of the user and of the resource.
One of the requirement of our vehicle scenario is the communication across multiple domains
(e.g. across two operators). XMPP (P. Saint-Andre, 2008) allows multi domain management
that can be achieved while making use of server-to-server communication. It also allows the
capabilities exchange and location awareness features via the presence stanzas. Regarding
the efciency of the protocol, several activities are being conducted to improve XMPP
performance, namely new lightweight version such as (Hornsby & Bail, 2009); however, the
major performance issues drive from the presence signaling which can be optimized. This
concern can be overcome with proposals like SOAP over XMPP (Fabio Forno, 2005), that
would even enrich the performance concerns, since XMPP and SOAP are two XML based
protocols, running one on top of the other.
2.2.2 Using media independent handovers
(Miguel Almeida, 2010b) focus on the possibility to merge reporting with mobility intrinsic
protocols. Since IEEE 802.21 was developed and is used to provide a lower layer
communication framework to deal with the exchange of information in heterogeneous
environments, our aim is to further extend it to enable the exchange of end user reports
independent from the underlying technologies. Moreover, this extension will allow the
seamless integration and activation of network reconguration procedures.
By extending the IEEE 802.21 MIIS, end user reporting can be performed at lower layers,
using one single protocol to carry all user, vehicle and network related information, which will
increase the efciency of resource consumption. The MIIS is expected to provide mainly static
information but, for the envisioned approach, real-time and dynamic information is required.
The IEEE 802.21 standard also mentions that dynamic information such as available resource
levels, state parameters and dynamic statistics, can be obtained directly from the respective
access networks. However, this information usually does not provide a clear view on the
end-to-end service performance. Also, the gathering of user behavior related information
from the network requires a means to access this information: this can be supported through
the IEEE 802.21, by loosening the concept of the MIIS and supporting new features and
functionalities.
To determine the QoS and QoE, it is required to assess the impact of the lower layer
information on higher layers at the core side. This information can be related via the cross
relation of PoA (Points of Access) with the terminal identication via the SAP (Service Access
Points). Therefore, it allows the collection of most of the information locally (either from
lower or higher layers), pre-evaluate it and then send it to the network. This view is aligned
with IEEE 802.21 which, through the MIIS, can provide an indication of higher layer services
supported by different access networks and other relevant information that may aid in making
handover decisions. Such information may not be available (or could not be made available)
directly from MAC/PHY layers of specic access.
Finally, the support of user and device (vehicular) reports through IEEE 802.21 allows the
seamless activation of network reconguration procedures, such as session and terminals
handover to networks that better match the user/device requirements, to jointly optimize
network resources and user experience. In sections 3.1 - 3.3, we further detail and propose
extensions to the MIIS to support a more detailed communication of application level
parameters, and introduce more intelligence upon handover decision.
2.3 Performance and behavior management
As dened by ITU-T, the Telecommunications Management Network model (TMN) (ITU,
1996), used for managing open systems in a communications network, establishes four
management layers comprising: (1) the element management, which entities are hierarchically
above and gather the information which is collected by each Network Element; (2) the
Network management system, which evaluates these metrics (after a Transform and Load
process); (3) the Service Management, which is in charge of taking into account the previous
layer and extrapolate conclusions that can lead to active changes in the network; (4) and
the business management layer, which introduces the agreement levels that need to be
accomplished. The Network Elements (NE) are typically the network nodes which interact
with the delivery systems. Operations, Administration and Maintenance (OAM) describes a
set of management levels and their interactions. The concept has more recently evolved to
include Provisioning and Troubleshooting. It ideally would imply the cross view of the TMN
model with the Fault, Conguration, Accounting, and Performance and Security (FCAPS)
functionalities.
To provide a clear view on the performance of the vehicles, indicators need be dened
according to the relevant metrics in the vehicle network. Key Performance Indicators (KPIs)
are a set of selected indicators used for measuring the current performance and trends.
KPIs highlight the key factors of the current performance and warn of potential problems.
Considering a counter as the most elementary value which is collected from a vehicle, a KPI
can simply be equal to a counter or to an arithmetic abstraction of counters that can be applied
to monitor a certain part of the network, functionality or protocol. KPIs play a major role in
creating immediate and relevant feedback on the performance of a certain element (may it be
network, hardware, or behavior).
2.3.1 Generic reporting tool architecture
Since we are proposing a remote management platform, the whole system would not be
complete without the inclusion of an architecture to evaluate the performance of the vehicles
in the cloud. This Reporting Tool receives the information from the devices and allows an
online verication of their performance by the end users. Fig. 2 shows the main components
which are typically included in a generic architecture for a reporting tool. Bellow we explain
the major functionalities of each component and their relevance to the architecture. The
Reporting Engine is the mind behind the Reporting Tool (Fig. 2). It is responsible for the
database queries, it processes the results and displays them in a dened format. It provides
all the data visualization capabilities, offering different pre-dened models and allowing
the user to create their own. These pre-dened visualization models are important, because
they allow the manipulation of data in different dimensions, providing different reports for
different types of end users, and even for different type of analysis, starting from a unique
data set. Related to these models, there is an important reporting component, which is the
KPI set. KPIs are dened in conguration components and can be either calculated on the y
by reporting engine, or pre-calculated and stored in the Reporter Database. The Automated
Knowledge Discovery model is another important part of this reporting engine, and provides
the very important feature of automatic data monitoring, searching for patterns in the network
behavior for the purpose of forecasting upcoming events, such as the Operation, Maintenance
and Optimization needs.
Fig. 2. Generic Reporting Tool Architecture
The Reporter Database is a data warehouse designed for the coherent integration of diverse
data sources, dimensioned to optimize the data discovery and reporting. This data repository
modulates all the network topology into a hierarchal object structure, which provides the
capability to analyze the entire network. This analysis can focus on the correlation of different
parameters that can be Conguration, Performance or Fault Management related. A possible
use case would be to assess what kind of conguration optimizes better the performance
of the network, by improving network capacity and reducing its faults. This analysis can
be extended in time and from different network perspectives, as historical and object data
aggregation is possible. Moreover, the database primitives allow for the storage management
and provide all the data access information to other layers. This database is modulated based
on NE specic metadata, which denes the Object Class (OC) structure, and for each OC all
the PM Measurements and related list of PM Counters and PM counters aggregation rules.
The FM metadata is generic for all the OC, dening a list of failures that can occur.
The Statistics Post-Processor is a software component that plays a decisive part on the
reporting process. It is responsible for the entire object and time aggregations, which enhances
the analysis capabilities, allowing the time trend analysis and drilling through the network
objects, enabling a great diversity of network analysis. The aggregation rules are all dened
through metadata specic for each NE, and provide information on howPMcounters must be
aggregated. The statistics Processor is responsible for converting all the diverse data gathered
from the NEs according to a structured and generic meta-model. This particular module
processes Conguration, Performance and Failure Management Information.
The Raw Database loader is responsible for providing interfaces for the access relating to data
storage management features. It is another function of an ETL procedure which uploads the
gathered information into the rawdatabases. This module includes interfaces for mediation of
the interactions between the EM and the analysis tools that evaluate the collected data. These
interfaces answer to the Reporting Tool for requests related to Performance Management
(PM), Conguration Management (CM) and Fault Management (FM) data.
The NE Mediations manage the interactions between the Element Manager Module and the
several Network Elements in the network. They are responsible for the collection of the
Performance and Fault Management functions existing in each of the elements of the network.
The NE Mediation Module implements the Extraction part of an ETL procedure. Each network
element monitors its performance through the Performance Mediation. A subset of that
module is responsible for the communication with the Element Manager. That interface is
divided into three types of primitives relating to the type of data which is to be transported:
PM, CM and FM. The rst presents metrics related with the continuous operation of the
equipment, while the second indicates the conguration setup, including information such
as topology and capabilities. FM is a more urgent type of data as it indicates critical issues to
be evaluated.
2.3.2 Types of metadata
As stated, the main objective of this chapter is to focus on the end-to-end reporting capabilities
between the devices and the cloud, while providing mechanisms and information so that
decisions can be made and measures can be taken if problems occur. However, the decision
making and acting components are not discussed. When considering reporting functionalities,
the typical supported metadata types are: CM, PMand FM. The work presented in this chapter
also considers Behavior Management (BM), in the sense that it allows the gathering of metrics
associated with the behavior of inhabitants of the vehicles and their interactions with the
vehicles.
Conguration Management metadata is responsible for the mapping between the different
NEs present in the network and their components into a coherent and structured Object
Class model. This way, CM metadata is used to identify objects with the same properties
and to maintain possible occurrences of an object in the object class hierarchy. Two types of
objects can de dened for this model, Managed Objects (MO) and Reference Objects (RO).
Managed Objects refers to objects that are directly related elements present in the network
that can be managed, congured, manipulated, which are obviously the NE elements and
their components e.g. a Node B and its Cells. Reference Objects refers to virtual reporting
dimensions, i.e. elements that are virtually created to ease the network analysis by dividing
and grouping the network into smaller segments thus reducing the analysis complexity.
This Reference Objects are created and stored in the Reporter Database by the Statistics
PostProcessor module, using the CM metadata information.
Performance Measurement metadata is responsible for dening, for each network element, all
the PMmeasurements and Counters and relating themto the CMdata, i.e. to the OCstructure.
As network element represents a specic role in the network, there will be a different set of
measurements/counters for each NE. The number of measurements and counters needed
to monitor a specic NE is dependent on the NE complexity, ranging with the number
of functionalities. A PM measurement is a logic representation of a NE functionality that
denes a set of counters that monitors the network performance behaviour. A PM counter
is the fundamental element of the performance monitoring process, as it provides detailed
information ranging from specic procedures up to group functions. As counters are the basis
of PM, they are used to develop different kinds of aggregations such as KPIs and Reports. This
way, different kind of users and analysis can be satised with only single tool.
Fault Management metadata denes the mapping between all the NE components and the
fault events that describe system failures, either hardware or software driven. FM metadata
thus relates OC with incoming network failure notications. These failures are categorized
and ranked by severity, which can range from debug to emergency state. The special
characteristic of this type of data is the fact that it typically has an unsolicited behavior and
requires near real time functionalities.
3. Architecture description
The following section depicts the architecture and the main functional entities that need to
be included to sustain the previously dened requirements, namely, the inter-technology
scenarios and the support of end user terminal reports integrated with network
reconguration triggering. Fig. 3 presents the vision explained in this chapter. Vehicles are
moving freely and through a wireless connection, which can be WiFi, WiMAX or 3GPP
based (UMTS, iHSPA or LTE), and are connected to a CSP. By using a mobility management
protocol like the Media Independent Handover with extended reporting capabilities, the
performance measurements of the vehicles and the behavior of the users can be gathered,
stored and evaluated within a cloud. This allows early problem detection, location and
context awareness, remote assistance, and a plethora of services fromwhich eet management
functionalities should be underlined.
Fig. 3. Interactions between the vehicles and the cloud
Each vehicle has an agent installed which sends information to the cloud, and is handled
by entities connected to a logic bus, the Media Independent Handover Function (MIHF).
On the other side, the Performance and Behavior Management module (PBM), collects that
information and stores it according to the type of data (User Behavior (UB) or Performance
Management (PM), which will be detailed in the next subsection). 3rd Party Cloud services
access this information and apply analysis algorithms (data mining procedures can be applied
but are not within the scope of this work), to present graphics explaining occurrences of
problems in certain vehicle models or in certain zones.
3.1 Architecture specication
Fig. 4 shows the main required entities for the proposed reporting architecture. End user
Behavior reports are communicated by the Multimedia Application to the Behavior Manager
at an agent (BM@Agent) installed in the vehicle. The Performance Manager (PM@Agent) is a
MIH User as well, and collects information from both the lower layers (QoS information) and
upper layers (QoE related information).The PM also interacts with the running applications
and the vehicles mechanical parts as well as the software/rmware for monitoring purposes.
The agent is a MIH entity that is responsible for gathering information and communicating
performance and behavior metrics. Lower layers report metrics that are extracted from the
technology drivers, including link and network layer values, such as throughput, bit rate
and SNR. From the upper layers, the PM will receive the information regarding service
performance, mainly related with end-to-end performance and QoE feedback. To achieve this,
these modules have open interfaces, which can be used through specied primitives, as will be
explained in the next subsection. Lower layer information can also be retrieved directly by the
Media Independent Information Service (MIIS). The MIIS (see Fig. 1) collects the data on the
network side and feeds the relevant user behavior and performance values to the Performance
and Behavior Manager (PBM@network).
Fig. 4. Interactions between the vehicles and the cloud
The PBM@network is responsible for the interaction with the multiple terminals to perform
proling analysis for individual and group behaviors. It comprises a database and interacts
with the API Management Agent, which is responsible for allowing access to the reporting
tool and 3rd party services. The typical approach to evaluate a users opinion on a service, and
hence depict his prole, is to collect end user reports after a service has been delivered, and
converge the opinion with the provided services characteristics. However, other methods can
provide an equally efcient evaluation of the users acceptance to performance trade offs. The
QoE and expected experience forma couple of properties that cannot be considered separately.
A user may be willing to accept lower performances if the contract fee is lighter. This
conclusion and consequent proling can be drawn from the user behavior. After applying the
proling analysis algorithm, the PBM formats the information to feed the Mobility Manager
(MM) and stores the results.
The MM receives the inputs from terminals and decides if an action is required. The MM can
use this input to take decisions, activating events in the terminal or events in the network
for optimization purposes. This process will make use and extend the IEEE 802.21 signaling.
Other proposals (Chung et al., 2008), (Jesus et al., 2007) deal with the mechanisms involving
mobility decisions and mobility signaling more deeply. To better understand this process, the
core network should be seen as a mediator of information. The vehicles will send information
to the CSP infrastructure to be handled by the PBM, and this information will be made
available to the in-cloud applications through the Performance and Behavior API Agent (see
subsection 3.4). It is also through that agent that the in-cloud applications can interact with
the vehicles. Fig. 4 shows an application in the cloud performing the remote management of
the performance of the devices. This scenario shows how the mediated data collected from
the devices can be outsourced to other services.
3.2 Signaling
To better understand the message ow, we will consider the scenario where a user contains
a multi-homed terminal connected to two wireless networks (e.g. WiFi and UMTS), but is
using the WiFi one (Fig. 5). Periodically, the multimedia applications report activity updates
and performance metrics. These messages are not IEEE 802.21 messages (Action messages
in Fig. 5), but are internal primitives. The same application will periodically issue another
message (Performance Report) informing about the relevant performance metrics for that
particular service. As shown in Fig. 5, an Action message is sent from the user application
to the BM@agent indicating that the user wants to receive a video and has issued for a
VoD request. The message should be according to the type: Action (Application ID, Type of
Request, Timestamp), thus specifying the application type which is being used (Video, Audio,
Gaming, Browsing, etc.) and the type of request (VoD, Streaming, Conference, Starting Game,
etc.). Following that procedure, different Performance messages will also be sent regularly.
The application issues a message containing the following structure: Performance (SourceID,
MetricType, Value, Timestamp), thus depicting the type of metric being reported and the value
for that metric. These messages are issued locally and then mapped to IEEE 802.21 by the MIH
Users (both BMand PM) for transmission to the network side (in the formof the messages and
procedures depicted in Section 3.3. Moreover, the PM@Agent receives Performance updates
from the hardware adaptation as explained in Section 3.3.2.
Ideally, the agent@terminal has well dened interfaces with the applications, and each
application can be responsible for reporting the users activities and performance feedback.
However, we introduce these entities as functional blocks for a better comprehension and
easier compliance with legacy applications. Hence, both message types Performance and
Action do not reect any type of signaling protocol but are instead internal primitives. The BM
and the PM are now ready to report this information to the network using the MIH Function.
When the Information Service requests for a UE update (MIH_Get_Info.request), it gets a
Fig. 5. Signaling diagram. URI is constructed from through the vehicle ID as explained in Fig.
7. (*Mechanical and Application Adaptations which are the Agent interfaces)
response containing the QoS performance from the application, from the lower layers, and
also the reported user action (via the Information Service Transport message). The IS receives
an update on the user status, and forwards this information to the PBM which evaluates the
QoS parameters, increments the user prole and communicates the changes to the MM. The
MM will evaluate the feasibility of this network for the desired service. Since it already knows
that the terminal has another interface with different properties (via standard IEEE 802.21
signaling: link_up message), it decides that a link with more bit rate would be better for the
services in use. It then issues a handover request to the terminal, so that it performs a handover
to LTE.
Whenever an action is taken by a user, the system needs to identify if the user is satised with
the current quality of the service (according to QoE parameters): the desired characteristics
and the used applications requirements should be taken into account to assess if they are
being met. If not, a change is required, by using another available interface (performing a
session handover) or switching to another PoA in order to enhance the terminal reception
conditions. This makes it possible to optimize the network or re-allocate users on different
PoAs. When Behavior and Performance Updates are collected by the PBM@Network, the
Cloud Agent is notied. It should then lookup the 3rd party entities which have interest
in receiving this update and use the POST method to transfer that information into the
destination web applications. This information can also be requested from the 3rd Party
services (which we designate as cloud in Fig. 5), using the Get method. Both GET and POST
methods use the URI, which is formulated via the identication of a user and the element of
a vehicle belonging to that user as well as the metric which is to be retrieved (this mapping
is done via the SAP ID and the metric type as explained n Section 3.4). To support this view,
it is required that both the Cloud Agent and the 3rd party entities (in Fig 5 the web service
is exemplied by a performance reporting tool) are running a web server, since the REST
methods are used via HTTP.
The way the mapping of web resources is done is detailed in subsection 3.4. In this section, we
demonstrate the signaling owto better explain howthe information is conveyed to and from
a web cloud based application. In general terms, the main concern is to cache the information
locally and send an update to the web applications with which the CSP has an SLA, and which
has expressed intention in receiving a particular devices instruction. It is assumed that this
process is preceded by a pre-enrollment phase, by means of which the owner of the vehicle
agrees on the availability of his data, as well as remote management functionalities being
sent to the vehicle. The dened signaling will allow both synchronous and asynchronous
reporting messages that are of relevance in order to support two types of information: a)
periodic performance or behavior data that has no crucial impact on the performance of the
vehicle; b) relevant and urgent information that might indicate a malfunction or a possible
problem and thus should be sent in an unsolicited way. Although the use case of sending a
command from the cloud is not expressed in Fig. 4, it is supported as explained in subsection
3.4, via the usage of a Rest command which is received at the API on a particular resource
which unequivocally identies the vehicle and the command to be remotely executed. The
API agent then conveys that command to the destination vehicle through the UBM.
3.3 Information elements
The IEEE 802.21 already denes general IEs, access network specic IEs, PoA specic IEs
and other IEs. Information Service elements are grouped into three categories: a) General
Information and Access Network Specic Information, which give an overview of the
different networks; b) PoA Specic Information that provides information about different
PoAs for each available access network; c) Other information that is access network specic,
service specic, or vendor/network specic. Next, we propose to include the Service
Performance IEs and the User Behavior IEs to be used in the previously presented architecture,
thus extending the ones dened in the standard, while taking (Miguel Almeida, 2010b). The
agent at the vehicles handles the following 3 types of messages:
Action (SourceID, Type of Request, Timestamp)
Performance (SourceID, MetricType, Value, Timestamp)
Alarm (SourceID, MetricType, Value, Timestamp)
The rst two can be easily handled by normal MIIS procedures, while the last one, given
its nature, would benet from a more unsolicited behavior. One way to solve this problem
is to use the MIH_Get_Information.request message carrying data that would indicate the
occurrence of an alarm situation. The MIH_Get_Information.request is sent to the MIHF in
the termina,l and then in the network side, a MIH_Get_Information.indication noties the
PBM@network (corresponding MIH User).
The MIH_Get_Information.response brings the conrmation that the alarmhas been received.
This provides a good workaround for the lack of unsolicited messages between MIH Users.
The Alarmreects the over crossing of a threshold value; the format of the message is the same
as the one of the Performance, but only indicates the urgency of the problem to the network.
For the purpose of supporting interfaces with the Controller Area Network (CAN) existing in
the vehicles (Johansson et al., 2005), the agent should also be able to act as a CAN gateway.
Since CAN and IEEE 802.21 are both lower layer based approaches, the overhead is minimal,
thus presenting a resource efcient solution. The way our framework handles this integration
is explained bellow in Section 3.3.2.
3.3.1 User behavior information elements
The type of active applications is usually communicated in the bootstrap of an application.
We dene it to be in the form TYPE_IE_UB_ACTIVE_APP_ID. It contains an index of the
application regarding the content a user is requesting. After reporting the active application,
several requests can be made by the user. The message type that should be used for this type
of report will be: TYPE_IE_UB_ACTION (ApplicationID, UserAction, Timestamp). The User
Action eld is dened in Table 1, which contains information on the actions that are required
to be supported for the previously mentioned services. The Timestamp is a time reference.
ApplicationOn ApplicationOff RequestChannel IdleMode
EndConversation InitiateConversation RequestURL ActionMode
SendMessage ReceiveMessage MovementMode RetrieveNeighboursList
JoinServer LeaveServer LeaveGame JoinGame
Table 1. User/Service Interactionsn
3.3.2 Performance information elements
Performance metrics from lower layers are already supported by the IEEE 802.21; the ones
being proposed relate to the application aware QoS and QoE values, which are directly
reported from the application layer to the MIH User. The agent collects information from
the several sensors in the vehicles as well as from the applications running. Vehicles become
multimedia oriented as time evolves, and now include displays for video and gaming
purposes, network connectivity for the retrieval of additional network based services, or
even simple internet access by itself. The QoE values will require additional metrics that are
relevant to characterize the service delivery quality.
QoS Performance Information Elements
As stated, it is required to have the complete view of the vehicles performance in the
management platform. Usually the metrics associated with the vehicles performance are
related with the components of each type of vehicle. Different sensors are applied to the
several parts and monitor temperature, pressure, speed, etc. Table 2 shows an example of
Distributed Control Architecture using CAN (Johansson et al., 2005).
Powertrain and Chassis Body electronics
Transmission control module Driver information module
Engine control module Steering wheel module
Brake control module Rear, Frontal and Central modules
Door module Climate control module
Steering angle sensor Steering wheel module
Suspension module Auxiliary electronic
Audio module Infotainment control
Table 2. example from (Johansson et al., 2005) for a particular automotive integration solution
The dened information elements do not need to be gathered via the CAN solution. They can
be collected via any customized solution as long as the agent receives them in the pre-dened
format and is able to send themto the network-side modules of management. For the purpose
of integrating the evaluation of the vehicles analysis in the cloud, we consider Information
Elements to be of the type: TYPE_IE_VP - Type Information Element Vehicle Performance.
QoE Performance Information Elements
The QoE Performance Information Elements are related with the services being accessed by
the users on the vehicle. The services can be bundled by the CSPs as part of the overall package
and can be evaluated individually. These parameters were rst described in Miguel Almeida
(2009), where a more extensive study is performed, employing a 3GPP viewwhile underlining
the relevant parameters at each layer of a 3GPP network. In fact, these parameters are a
rst glance at the performance view of a service, and could be narrowed down, for problem
identication purposes, until the lower layers. It denes the way in which the KPIs should
be constructed and how they can be evaluated in a Cross Layer View, while in Igor Pais
(2009), a more QoE centric analysis is performed. For each service we consider metrics such
as the total setup waiting time for a service to be received (TYPE_IE_SP_WT), the Mean
Opinion Score (TYPE_IE_SP_MOSQ), the Service Availability (TYPE_IE_SP_AVAILABILITY),
the Lost Packets (TYPE_IE_SP_LOSS), the Time Resolution (TYPE_IE_SP_TIMERES for voice
and TYPE_IE_SP_FPS for video), and, of course, the Bit Rate (TYPE_IE_SP_BR). All primitives
include the following elds: SourceID (or application ID), Application Type ID, Time and
Value.
3.4 Integrating with the cloud
In the previous sections we have been debating the collection mechanisms which allow to
gather and convey the information from the vehicles into the network. This would allow the
Mobile Network Operators (MNO) which own the gathering technology to access the data
and evaluate it. In order to make it publicly available and, in this way, further capitalize this
solution, the MNO would greatly benet from a seamless way to make it available to 3rd
party entities (e.g. 3rd Party CSPs), which could hire the access to the data. For instance, eet
management functions could be employed and then the vehicles performance information
might be outsourced. Subsection 2.2.1 shows a way to gather information using XMPP and
then relaying it into the cloud using REST. By extending that proposal in order to also convey
the information carried in the MIH messages, we create a compliant system. To do so, we
need to create an adaptation in the Cloud Bridge Server (Miguel Almeida, 2010a), which we
denote as Performance and Behavior API Agent. That agent interfaces with the MIH agents
which collect the messages from the vehicles generating REST alike messages which update
the 3rd party webservices accordingly. In this way, instead of using XMPP as the transport
protocol, we use 802.21 to convey the performance messages, which is a more efcient solution
for medium to high mobility scenarios (see Fig. 6). The web services expose an interface
that allows information to be asynchronously supplied, and commands requested, when
necessary, due to network management operations.
We gather the SAP ID and cross match it with the vehicles information within the operator.
There are two solutions to cope with the device identication. We couple the information of
the IDs of the Service Access Points and of the MIH Users that coexist in the same vehicle
and then translate them into a resource. By employing the translation shown in Fig. 7, we
can guarantee the required consistency between the adaptation and the CBS, enabling the
exposure of MIH resources to the REST interface, which can now be fully controlled on a
Fig. 6. Protocols Used for the interaction between the Vehicles and the Cloud
per-vehicle policy determined by the bridge. All messages are sent to the Cloud seamlessly,
given that any web like environment can support RESTful primitives.
Fig. 7. Creating the ID for the REST resource
As the entry point towards the Cloud, the Performance and Behavior API Agent enables the
communication between the devices and the Software as a Service, taking on a vital role for
authentication and authorization. Each service must register, dene an SLA, and authenticate
to gain access to information pertaining to the vehicles. Given the control over the information,
the Performance and Behavior API Agent is able to dene a granular access control to the
information exposed. The Performance and Behavior API Agent supports HTTP compliant
messages (GET, PUT, UPDATE, DELETE), which are also the core of REST functionality,
thus assuring the integration with minimal effort on the Cloud Bridge Server. This secure
setup even allows customizing the devices towards a specic web service. If the devices are
congured accordingly, they can opt to send reports to the custom Web Service URLs. The
802.21 signaling will transport the report, and then the Performance and Behavior API Agent
will performa HTTP POST on the destination Web Service. The Web Services need to be aware
of the data model being communicated.
4. Performance comparison
The main advantage of the described approach is to save on signaling and overhead while
simultaneously allowing end-to-end heterogeneous reports, and seamlessly integrating both
reporting and network enforcement processes. The proposed architecture aims to provide an
accurate proling of users for an enhanced network optimization and resource consumption
prediction. In this section, it is presented a quantitative evaluation in terms of trafc generated
by different approaches, and a qualitative evaluation in terms of supported functionalities.
4.1 Qualitative evaluation
In this section we include a qualitative evaluation of the functionalities provided by current
approaches and the IEEE 802.21 based approach (denoted as MIHR). A summary of the
supported features is presented in Table 3. We rst start by comparing the different approaches
in terms of features support. We can see that for the purposes of integrating the devices with
a web environment, SNMP is the most inadequate, since HTTP based solutions are already
web based, and XMPP is XML based which enables a direct transformation of the objects. The
802.21 approach requires a special adaptation on the network side. Since it only denes a way
for the devices to intercommunicate with the network on a lower level, a MIH user is required
at the network side to behave as a proxy towards the web cloud.
Regarding the security features, XMPP creates a secure unique channel using TLS. Earlier
versions of SNMP present serious security constraints, and this has been addressed in SNMP
v3; still, the common applications use IPSec bellow the SNMP communication. Although this
feature does not reside within the main focus of the 802.21, it can use 802.1x for the IEEE
based networks and use the channel security procedures of the 3GPP networks. SNMP also
allows authentication to verify that the message is from a valid source, while XMPP with
REST supports authentication of an Identity and its merging with accounting information.
With respect to Identity Management, XMPP is the best proposal to link several devices
to someones identication. Considering that the platform is to be deployed at a Network
Operators site, then it would be good to allow the cross matching of accountings with
the operators database, only feasible using the possibilities offered by XMPP. Since we are
focusing on the device management more than on the Identity management functions, it is
simpler to use a lower layer device management system, which takes into account the SAP ID
or simply an IMEI.
Feature: MIHR XMPP + REST HTTP Based SNMP
Security Yes TLS SSL IPSec
Reliability Yes Yes No Yes
Authentication No SASL No No
Web Cloud Integration High Easy Easy Medium
CSP Integration Medium Easy N/A Complex
Identity Management No Yes No No
Bi-Directionality Yes Yes No Yes*
Table 3. Management Features Comparison; *not when using traps
Being able to support asynchronous communication allows the deployment of a very relevant
feature for fault management functions, which is sending alarms in a near real-time way.
There are several applications that take advantage from this possibility, and XMPP is the best
approach to deal which this type of events. Moreover, it allows bidirectional communication
without the need to run a web server in the devices, which is the only way to support
bidirectional communication using HTTP with REST or with SOAP. The way 802.21 handles
this problem is through commands, which allow sending information to the devices. In this
sense, by gifting the network with the appropriate intelligence at the MIH user, it is possible
to enable bi-directionality of the management applications, in a very resource effective way.
IEEE 802.21 considers reliability to be a requirement from the underneath media, which
needs to have a reliable message delivery procedure in order to allow the MIH Protocol
Acknowledgment Service (802.21-2008, 2009). XMPP and HTTP run on top of TCP, which
ensures the reliability mechanisms. SNMP runs on top of UDP, and therefore it employs the
mechanism of request/response: if the response is not obtained, the request is again sent (and
can be customized). Obviously, this a problem for the trap-enabled version of the protocol.
CDRs are call oriented and transport the information of the sessions and of the users involved.
Their major advantage is the fact that they were created for an easy usage within the operators
domain. The MIHR requires an effort to integrate the solution with the Mobile Network
Operators (MNO) Home Location Register (HLR), while employing a mapping between
the devices and the owners. For the integration of the HTTP based solutions, this would be
difcult and would require a great deal of customizations, namely assuring that the agents
collect such information on the client side, and then, convey it to operator. The degree of
customization would be so high that we opt to dene it as non applicable, since it is a concept
too broad and too subjective. SNMP also would require an adaptation mechanism that is
aware of the identication of the device with which the network is communicating. The
network would then need to map this information with something previously known, i.e.,
it would need to previously be aware of the agent ID, and map it into a specic user.
Asummary of the performance metrics is presented in Table 4. Using the IEEE 802.21 method,
it is not required to exchange the full performance details, but only those which are required.
Since we are dealing with a media independent proposal, our reports can be applied in any
type of technology even in a switched domain. SNMP is technology oriented and MIBs are
dened for specic types of hardware. Typically, CDR analysis is very technology-oriented
and includes details relevant only for the source of the type of access. When using an approach
on top of HTTP, it also becomes depends on the network information, and the typical approach
is to perform QoS measurements and establish a TCP connection towards a collecting server.
In IEEE 802.21, the MIIS already handles that interface seamlessly. XMPP uses TCP and runs
on top of IP, so although it could be considered technology agnostic, it brings large amount of
of signaling, fromthe presence stanzas. For cellular networks, this would require optimization
procedures, and additionally, login/logout functionalities.
Metric MIHR XMPP + REST CDR HTTP Based SNMP
Overhead Good Bad Bad Bad Medium
Signaling Good Bad Bad Bad Medium
Heterogeneity Yes Yes Yes Yes Yes
Synchronous Yes Yes Yes Yes via Traps
Asynchronous Yes Yes Yes Yes Yes
Multilayer Yes Yes No N/A No
Table 4. Qualitative Performance Comparison
Multilayer analysis relates to the ability of the different procedures to evaluate the
performance at different layers, and in parallel, to deal with the end user behavior issues.
Using the IEEE 802.21-based approach, information can be collected from lower layers and
fromupper layers, simultaneously or separately depending on which information is required.
Information can be collected locally at the NEs and then be reported to a central server: an
approach with reporting over IP will still work, but will be less seamless for the intermediate
NEs. The same concept applies to SNMP; however, usually this procedure will not be applied
to upper layer analysis or behavior related parameters. On the other hand, CDRs will not
focus on the network information. XMPP is dependent on what the agent is collecting, but it
introduces no constraints at this level, relying only on the capabilities of the agent@terminal.
IEEE 802.21-based approach relies on the IEEE 802.21 mechanisms, which do not support
events from upper layers. Having that in mind, it is only possible to support reactive
events for some of the typical performance parameters. However, as explained before, we
overcome this problem for the support of alarms via the issuing of customized messages from
the terminals. Asynchronous events could be supported by implementing event triggering
from upper layers for end user behavior analysis. Since event triggering from lower layers
is still valid, this proposal partially supports synchronous reporting. XMPP was created
for instant messaging purposes, so it excels at both approaches, while SNMP can only be
considered to be asynchronous when using traps. CDRs are implemented on a per-call basis,
so it can be considered an asynchronous approach. The HTTP based approach can also
be considered proactive, since it requires inputs, but it supports both asynchronous and
synchronous methods (e.g. Get method will request for information, while the Post will add a
new resource). In order to employ a synchronous procedure with requests from the network
to the client, this would require the devices to have a webserver installed, which is clearly
a downside. When considering the heterogeneity support, one issue arises: assuming that
most of the approaches can run on top of the IP/TCP stack, we would need to state that
XMPP, HTTP and SNMP approaches that use TCP and UDP are heterogeneous as long as the
technologies use IP. This is a fairly accepted assumption today. However, the only technology
that was created having in mind heterogeneity support was the IEEE 802.21.
In terms of signaling, overhead and performance comparison, we evaluated how the access
links that connect the devices to the cloud would behave. As can be seen in Section 4.2,
SNMP outperforms the XMPP approaches. This happens mainly because in XMPP we dened
an Object Class and used the objects within an XML to make the transactions. The results
presented for the XMPP approach were obtained from experimental evaluation, while in
the SNMP related results, we computed the overhead induced by raw data transportation,
when conveying binary information. This information would require post processing at the
network management system. The exchange of performance records in CDRs is typically
xed size oriented and depends on the record detail. Moreover, the overhead is signicantly
high, since it includes information related to the user device for identication (e.g. IMEI),
which is not required with IEEE 802.21, since the SAP ID already matches the device ID.
The major inconvenience of using CDRs with additional inputs from the terminals is that it
requires additional overhead and signaling, since in 3GPP networks it is required to consider
the tunneling effects and the establishment of IP or GPRS connections for data transmission.
In the future 3GPP releases this problem may be mitigated; however, reporting at this level
will always introduce an overhead larger than the IP+UDP one. The best approach is to use
the IEEE 802.21 solution, which handles the problem on a lower layer basis, thus reducing
overhead and signaling.
4.2 Quantitative evaluation
Figure 8 presents the trafc generated by the several approaches with the size of the object
of information being reported. We compare the performance of different protocols on the
wireless link, i.e., the link between the vehicles and the networks infrastructure nodes, which
is the most problematic part of the network, when considering high mobility scenarios.
This is in fact the link which introduces more concerns, in terms of resource consumption
efciency. The evaluated protocols include: SNMP Polling or Trap-based methods, HTTP
using SOAP (Simple Object Access Protocol), IEEE 802.21 and XMPP transporting REST
methods. Regarding XMPP and REST, we used the results detailed in Miguel Almeida (2010a),
which were experimentally obtained. The other metrics were obtained as detailed bellow.
Typically, Mobile Web Services introduce high overheads in general. (M. Tian, 2003), (Pras
et al., 2004) include details on the overhead impact under various conditions. As Figure 8
shows, when using an object based transport like the one used in MQA using SOAP over
HTTP, one must consider the overhead introduced by the signaling and also the headers of the
IP, TCP (with timestamps), SOAP and SOAP envelope. The calculated overhead is presented
in Equation 1. As the object size increases, packet segmentation occurs at the TCP layer, thus
signicantly increasing the size of generated trafc.
length
Soap
= header
IP
+ header
TCP
+ header
HTTP
+ +header
Soap
+ envelope + object
size
(1)
When using SNMP, the overhead per object is decreased for both scenarios: we consider
both the best case scenario with unsolicited messages exchange (via the usage of traps),
and also in the case of polling-based approach. According to (de Lima et al., 2006), the
header size of SNMPv2c is approximately 25 octets. The overhead of SNMPv3 is given by
Equation 2, where the Header Data of the SNMP is given by Equation 3 as 17 octets. This
means that SNMPv3 adds a minimum of 17 octets to SNMPv2c. Considering the signaling
generated by both versions, the total generated trafc is given by Equation 4 and Equation
5, respectively for SNMPv2c and SNMPv3. In Figure 8, SNMPv3 trafc was generated using
ASN.1 (Abstract Syntax Notation One), and Get processes (Request + Response messages)
were taken into account for overhead measurement purposes. For the transmission of more
objects, the performance of SNMP is decreased, as shown in (de Lima et al., 2006) and in Figure
9. Using a bulk procedure would greatly reduce the overhead in this case, but this would be
true for all other proposals.
Tra f f ic
SNMPv3
= HeaderData + SecurityParameters + ScoopedPDUdata (2)
Header
SNMPv3
= Version + MSGID + MaxSize + Flags + SecurityModel = 17Octets (3)
Tra f f icSNMPv2c = 54 + n(10 +2.Objectlength + Value) (4)
Tra f f ic
SNMPv3
= 88 + n(10 +2.Object
length
+ Value) (5)
When considering the trap-based approach, we determine the trafc of SNMP in Equation 6
for SNMPv2. Then, we calculate the minimumtrafc generated by SNMPv2 traps in Equation
7 and add the additional minimum overhead (considering Communitysize=6 octets and
Traplength=1 octet) to get Equation 8 for the trafc generated by SNMPv3 traps (OID refers to
Object Identier). As can be seen in Figure 8 and Figure 9, Traps reduce the generated trafc,
especially when a large number of events are being generated.
Tra f f ic
TrapSNMPv2c
= 63 + Community
size
+ TrapOID
size
+ n.(3 +OID + Value) (6)
Tra f f ic
SNMPv2Trap
= 70 + (3 +OID +Object
size
).numObjects (7)
Tra f f ic
SNMPv3Trap
= 87 + (3 +OID +Object
size
).numObjects (8)
Fig. 8. Trafc Generated vs number of Objects
We also evaluated the minimum trafc generated by an approach which only uses UDP over
IP without any additional signaling. Besides the application port, the OID and the value of
the object, we only considered the values for an unsolicited procedure with the overhead of
the IP and UDP headers and a variable eld using one octet for the OID. The trafc generated
consequently decreases when compared to SNMP. Considering our IEEE 802.21 approach,
we can observe that it generates similar or lower trafc as an IP unicast transmission of the
raw information over UDP. This becomes more clear in Figure 6, since MIH reports do not
require the specication of a way to request for specic objects above the IP and UDP layer;
the major overhead saving comes from the lack of requirement for the usage of the IP header
and signaling, which would be required in order to support the request of an object and
consequent transport. (Melia et al., 2007) further details this aspect with emphasis on the low
overhead introduced by the protocol during mobility procedures.
Fig. 9. Trafc generated by number of Nodes
5. Conclusions
In this chapter, we presented an architecture based on IEEE 802.21 that is able to
gather information from vehicles and allows access via a web cloud. It provides
seamless support at different levels: reporting of cross-layer information, support of
inter-technology environment, and integration of the actions of reporting with those of
network reconguration. This approach combines the basic mechanisms of the IEEE 802.21
to gather information from the devices on the wireless link and the REST primitives to convey
that information into the Cloud. Using the IEEE 802.21 Information Service, we underlined
the required basic signaling to integrate both upper and lower layers information, regarding
both the QoE the user is experiencing and the QoS parameters of the services. Moreover,
using IEEE 802.21, the network can then act on the vehicles, basing its decision on proling
algorithms which estimate the future actions of a group of users, seamlessly supporting
both mechanisms of reporting and reaction. The results presented show that this approach
signicantly decreases the reporting overhead, while it introduces a set of functionalities not
present in current approaches.
Through the analysis of several transport technology (and as the core driver for interactions
within the core domains), we consider a XMPP based solution to be the best approach
when envisioning the integration of end users interacting with terminals, gaming consoles,
cell phones, IP enabled sensors, etc, with web environments and also with the operators
infrastructure. However, for more mobile environments, where wireless resources are the
major concern, and where fast connectivity maintenance procedures play a major role, its
performance decreases. The number of features supported allows the authentication using
the account identication of the devices owners within the operators Charging Gateways,
allowing the easy deployment of charging per usage.
The integration with the cloud environment provided through REST interfaces allows
the interaction with 3rd party web services, increasing the possibilities of applicability
and revenue. By providing common and consistent interfaces to act and report on
devices, we enable a new array of business relationships and opportunities that put the
telecommunication and infrastructure operator back in the driver seat of the network, while
enabling a clear interaction with the Cloud world, a feature which has been profoundly
lacking from the operators portfolio. We believe that these paradigms will be a key revenue
system where both operators and service providers can capitalize by using the adequate tools
to unite the common approaches. This view greatly facilitates the interactions with vehicles
on the move, via several technologies seamlessly reporting behavior and performance metrics
using a lightweight reporting mechanism coupled with mobility
6. References
802.21-2008, I. S. (2009). Ieee standard for local and metropolitan area networks- part 21: Media
independent handover, IEEE Std 802.21-2008 pp. c1 301.
Berners-Lee, T., F. R. & Masinter, L. (1998). Uniform Resource Identiers (URI): Generic Syntax,
IETF RFC 2396.
URL: http://www.ietf.org/rfc/rfc2396.txt
Breda, G. & Mendes, L. S. (2006). Qos monitoring and failure detection, Telecommunications
Symposium, 2006 International, pp. 243 248.
Chung, T.-Y., Yuan, F.-C., Chen, Y.-M., Liu, B.-J. & Hsu, C.-C. (2008). Extending always best
connected paradigm for voice communications in next generation wireless network,
Advanced Information Networking and Applications, 2008. AINA 2008. 22nd International
Conference on, pp. 803 810.
de Lima, W., Alves, R., Vianna, R., Almeida, M., Tarouco, L. & Granville, L. (2006). Evaluating
the performance of snmp and web services notications, Network Operations and
Management Symposium, 2006. NOMS 2006. 10th IEEE/IFIP, pp. 546 556.
Don Box, David Ehnebuske, G. K. A. L. N. M. H. F. N. S. T. D. W. (2000). Simple Object Access
Protocol (SOAP) 1.1, W3C Note.
Fabio Forno, P. S.-A. (2005). XEP-0072: SOAP Over XMPP, 1999 - 2010 XMPP Standards
Foundation.
URL: http://xmpp.org/extensions/xep-0072.html
Fielding, R. T. (2000). Architectural Styles and the Design of Network-based Software Architectures,
Doctor of Philosophy Dissertation, University of California, Irvine.
Gurgen, L. & Honiden, S. (2009). Management of networked sensing devices, Mobile Data
Management: Systems, Services and Middleware, 2009. MDM 09. Tenth International
Conference on, pp. 502 507.
Hornsby, A. & Bail, E. (2009). x03bc;xmpp: Lightweight implementation for low power
operating system contiki, Ultra Modern Telecommunications Workshops, 2009. ICUMT
09. International Conference on, pp. 1 5.
Hornsby, A., Belimpasakis, P. & Defee, I. (2009). Xmpp-based wireless sensor network and
its integration into the extended home environment, Consumer Electronics, 2009. ISCE
09. IEEE 13th International Symposium on, pp. 794 797.
Igor Pais, M. A. (2009). End user behavior and performance feedback for service analysis,
Intelligence in Next Generation Networks - ICIN.
ITU (1996). Telecommunications Management Network Reference Model: CCITT Recommendation
M.3010, International Telecommunication Union.
J. Case, E. A. (1990). A Simple Network management Protocol (SNMP), RFC 1157.
Jesus, V., Sargento, S., Corujo, D., Senica, N., Almeida, M. & Aguiar, R. (2007). Mobility with
qos support for multi-interface terminals: Combined user and network approach,
Computers and Communications, 2007. ISCC 2007. 12th IEEE Symposium on, pp. 325
332.
Johansson, K. H., T urngren, M. & Nielsen, L. (2005). Vehicle Applications of Controller Area
Network, Birkhduser.
Jung, S.-J., Lee, J.-H., Han, Y.-J., Kim, J.-H., Na, J.-C. & Chung, T.-M. (2007). Snmp-based
integrated wire/wireless device management system, Advanced Communication
Technology, The 9th International Conference on, Vol. 2, pp. 995 998.
M. Tian, e. A. (2003). Performance considerations for mobile web services, IEEE Communication
Society Workshop on Applications and Services in Wireless Networks, Vol. 2, pp. 741 746.
Melia, T., de la Oliva, A., Vidal, A., Soto, I., Corujo, D. & Aguiar, R. (2007). Toward ip
converged heterogeneous mobility: A network controlled approach, Comput. Netw.
51(17): 48494866.
Miguel Almeida, A. M. (2010a). Bridging the devices with the web cloud: A
restful management architecture over xmpp, 6th International Mobile Multimedia
Communications Conference.
Miguel Almeida, Rui Inacio, S. S. (2009). Cross layer design approach for performance
evaluation of multimedia contents, International Workshop on Cross Layer Design.
Miguel Almeida, S. S. (2010b). Media independent end user behavior and performance
reports, IEEE GLOBAL COMMUNICATIONS CONFERENCE.
Mohinisudhan, G., Bhosale, S. &Chaudhari, B. (2006). Reliable on-board and remote vehicular
network management for hybrid automobiles, Electric and Hybrid Vehicles, 2006.
ICEHV 06. IEEE Conference on, pp. 1 4.
P. Saint-Andre, E. (2004). Extensible Messaging and Presence Protocol (XMPP): Core, IETF RFC
3920.
URL: http://www.ietf.org/rfc/rfc3920.txt
P. Saint-Andre, E. (2008). XEP-0238: XMPP Protocol Flows for Inter-Domain Federation, 1999 -
2010 XMPP Standards Foundation.
URL: http://xmpp.org/extensions/xep-0238.html
Peter Saint-Andre, Kevin Smith, R. T. (2009). XMPP: The Denitive Guide Building Real-Time
Applications with Jabber Technologies, OReilly Media.
Pras, A., Drevers, T., van de Meent, R. & Quartel, D. (2004). Comparing the performance of
snmp and web services-based management, Network and Service Management, IEEE
Transactions on 1(2): 72 82.
Soldani, D. (2006). Means and methods for collecting and analyzing qoe measurements in
wireless networks, World of Wireless, Mobile and Multimedia Networks, 2006. WoWMoM
2006. International Symposium on a, pp. 5 pp. 535.
Voas, J. & Zhang, J. (2009). Cloud computing: New wine or just a new bottle?, IT Professional
11: 1517.
Waldbusser, S. (1995). Remote network Monitoring management Information Base, RFC 1757.
6
Ultra-Wideband Automotive Radar
Akihiro Kajiwara
The University of Kitakyushu
Japan
1. Introduction
A lot of progress has been made for automotive radar during the last years. There are two
types of automotive radar; long-range radar at 77GHz with a range capability up to 200m
for automatic cruise control (ACC) and short-range radar at 24/26 and 79GHz up to 30m
for anti-collision. Long radar with narrow radiation beam enables a automobile to maintain
a cruising distance, while short-range radar has recently attracted attention because of many
applications such as pre-crush warning, stop-and-go operation and lane change assist. The
short-range radar with a very broad lateral coverage has a few significant problems to be
overcome such as target detection and clutter suppression. This is because the widely
radiated radar echo contains not only automobile echo, but also unwanted echoes called
clutter. It is actually not easy to detect a target echo in increased clutter. Ultra-wideband
impulse-radio (UWB-IR) radar with high range-resolution has recently attracted much
attention for automotive use, because it offers many applications such as pre-crush warning
and lane change assist.
The followings provide an overview of this chapter;
1. Section 2 introduces various radar systems for automotive use. It begins with a
discussion of radar technologies such as Pulse Doppler, FM-CW and UWB-IR.
2. UWB-IR radar requires high speed A/D devices which can directly process the received
nanosecond pulse. For example, A/D devices of several GS/s or more should be
required for the UWB-IR radar with a bandwidth of 1GHz, which have not been
available yet. The use of wideband may also cause unacceptable interference on existing
narrowband systems. Therefore, some interference mitigation scheme may be required
for the radar emission in the future. To solve these problems, a stepped-FM radar
scheme is introduced in Section 3, which does not require high speed A/D device and
provides the co-existence with exisiting narrowband systems.
3. Short range radar is expected to provide a wide coverage in azimuth angle. Therefore,
increased clutter makes it difficult to detect automobile target accurately. The clutter
can be classified from automobile by the Doppler, but it will not be applicable to the
UWB-IR. In Section 4, a scheme is introduced which estimates the Doppler by using the
time-trajectory of radar echo and the measurement results are presented.
4. Automotive radar is required to detect automobile accurately, but not to detect clutters
falsely, even in complicated traffic conditions. In order to satisfy the requirement, a
target discrimination scheme with range profile matching is introduced in Section 5 and
the measurement results are presented. The results show that the automobile type can
be discriminated.

104
2. Fundamentals of radar technologies
2.1 Radar detection
The basic principle of UWB-IR automotive radar detection is illustrated in Fig.1 where the
received signal includes many echoes scattered from desired and undesired objects (Skolnik,
2001) (Taylor, 1995). The one-dimensional signal, which is referred to as range profile, is
generally presented by multiple impulses with gains {
k
} and propagation delays {
k
},
where k is the impulse index. Suppose a nanosecond pulse of s(t), the range profile, y( ,t), is
the time convolution of s(t) and the impulse echo response
k k
t ( ) as follows;

) ( ) ( =
k
k k
t s t , y
(1)
Fig.1 (b) shows an example of received power range profile for a bandwidth of 1GHz
(corresponding to 1 nanosecond pulse) on a roadway.
When a target echo exceeds a given threshold, it can be recognized. However if a clutter
echo exceeds the threshold, it is mistaken as a target. This is called a missed detection.
Consider the target detection in increased clutter as shown in Fig.1, it is not easy to
recognize the automobile target since the echoes over a threshold cant be classified usually
as target or clutter.

(a) Automobile radar image

(b) Power range profile
Fig. 1. Principle of radar detection
The typical radar utilizes a pulse waveform. The transmitted pulse is intercepted by some
objects and re-radiated in many directions. The re-radiation directed back towards the radar

105
is collected by the directional antenna. The received signal is then processed to detect the
presence of target on the threshold basis. The range to a target is estimated by measuring the
travelling time from the transmitter to the receiver. The range is given by d c = /2 where
is the time and c is the speed of light. The radar equation generally relates the detectable
range. It is useful not only for determining the maximum range at which the radar can
detect a target, but it can serve as a means for understanding the factors affecting radar
performance. For the transmit power P
t
in a particular direction, the maximum gain G of
antenna and the received signal P
r
at a range d is given by (Skolnik,2001)

t
r
P G
P
d

=
2 2
3 4
,
(4 )
(2)
where is the wavelength and denotes the reflectivity of the target which is also called
radar cross section (RCS).
The RCS depends on the target shape, and radar bandwidth. An example of the
measured RCS of a sedan type of automobile is shown in Fig.2, where we considered a CW
of 25.5GHz and UWB of 22-29GHz (Matsunami et al., 2008). Change in the RCS of the CW is
found to be significant to the azimuth angle, while the RCS forof the UWB is much less
sensitive to the azimuth angle relative to the CW.

Fig. 2. Azimuth variation of the RCS of a sedan typed automobile
2.2 Automotive radar system
There are several types of radar systems and a large number of different applications. In this
section, three types of radar system are discussed: Pulse Doppler, FM-CW and UWB-IR
radar systems.
Pulse Doppler radar
Pulse Doppler radar sends out a pulse train. When the transmit pulse is long enough and
the targets Doppler is large enough, it may be possible to detect the Doppler shift on the
basis of the frequency change within a single pulse. Fig.2-3 (b) shows that there is a
recognizable Doppler shift in frequency domain. To detect the Doppler on the basis of a

106
single pulse of width T
p
generally requires that there be at least one cycle of the Doppler
frequency f
d
within the pulse. Also, the transmit frequency source must have very good
phase stability and the system is required to be coherent. Fig. 3 illustrated the principle of
Pulse Doppler where the pulse repetition period is T. For a moving target, the received echo
experiences a Doppler f
d
as shown in Fig.2-3. Therefore the Doppler can be estimated by
processing the received echo in frequency domain.

Fig. 3. Principle of Pulse Doppler radar
FM-CW radar
FM-CW (Frequency-modulated continuous-wave) radar system varies the frequency of the
transmit signal and measures the range based on the frequency difference between the
instantaneous transmit signal and received echo. The typical modulations includes
triangular and saw-tooth. Current applications of FM-CW include high-resolution systems.
Fig. 4 (a) shows the block diagram of a typical triangular FM-CW radar where the super-
heterodyne-receiver is assumed. The frequency-analyzed beat signals for up-frequency and
down-frequency are passed through an A/D device, and digital processing of FFT is then
conducted. Please note that the beat signal exhibits a peak at which the intensity becomes
large corresponding to the target. And the peak frequency corresponding to this peak
carries information concerning the distance, and the peak frequency differs between the up
portion and down portion of the triangular FM-CW wave due to the Doppler associated
with the relative velocity of a target.
The up-beat is given by the followings;

d
m
f d
f - f f
T c
= =
1 2
2
(3)

m
f
f t
T
=
1
(4)

d
m
f d
f t f
T c

= +

2
2
, (5)
where f is the frequency excursion.
The down-beat is also given by

d
m
f d
f - f f
T c
= = +
2 1
2
(6)

107
( )
m
m
f
f t - T f
T
= +
1
(7)

m d
m
f d
f t T f f
T c

= + +

2
2
- (8)
From the above and , the Doppler f
d
and the distance d to a target are derived as
follows;

( )
( )
1
2
4
d
m
f
cT
d
f
=
= +
(9)

(a) Block diagram of triangular FM-CW radar

(b) Frequency-time relation for triangular FM-CW

108

(c) Frequency-time relation
Fig. 4. FM-CW radar
If there is more than one autobile traveling in front, one pair of peak frequencies, one in the
up portion and the other in the down portion, occurs for each automobile. The pairing of the
peak frequencies between the up and down portions is done on an automobile-by-
automobile basis. The pairing is performed based on the peak frequencies as shown in Fig. 4
(c), and the range and relative velocity of the corresponding automobile are determined.
However the selection requires complicated processing.
UWB-IR radar
UWB-IR radar system transmits signals across a much wider frequency than conventional
signals. The transmit signal is significant for its very light power spectrum which is typically
lower than -41.3dBm. The most common technique for generating a UWB-IR signal is to

Ultra-short pulse (b) Power spectrum
Fig. 5. Train of ultra-short pulses and its spectrum
transmit a very short pulse train (less than 1nanosecond). The spectrum of a very narrow-
width pulse has a very large frequency spectrum approaching that of white noise as the

109
pulse becomes narrower and narrower. These very short pulses need a wide bandwidth as
shown in Fig.5. The amount of spectrum is at least 25% of the center frequency. For the high
range-resolution radar with wider than 1GHz, each scattered echo should be separated.
Fig.6 shows the range profiles of a roadway where an automobile target was located at 10m.
The measurement was conducted for various bandwidths of 100MHz to 5GHz (Matsunami
et al., 2008). It is seen that the echoes for various objects can be separated using a shorter
pulse.

(a) 100MHz (b) 500MHz

(c) 1GHz (d) 5GHz
Fig. 6. Power delay profiles for automobile
3. Stepped-FM radar
UWB-IR radar provides multipath tolerability and high range-resolution, but it requires
high speed A/D devices which can directly process the received nanosecond pulse. For
example, A/D devices of several GS/s or more should be required for the UWB-IR radar
with a bandwidth of 1GHz, which have not been available yet. Please note that it is difficult
to realize analogue devices which can process such wideband pulse. The use of wideband
would also cause unacceptable interference on existing narrowband systems. Therefore,
some interference mitigation scheme may be required for the wideband radar emission in
the future. To solve these problems, stepped-FM radar is introduced which does not require
high speed A/D device and provides the co-existence with exisiting narrowband systems.

110
3.1 Stepped-FM scheme
The block diagram and each signal waveforms are shown in Fig.7. The stepped-FM radar
transmits a series of bursts of narrowband pulses, where each burst is a sequence consisting
of N pulses shifted in frequency from pulse to pulse with a fixed frequency step f . The
received echo from a target is phase-detected into a train of narrowband base-band pulses
and is then I-Q sampled by a relaxed speed of A/D. Each complex sample is applied to the
inverse discrete Fourier transformation (IDFT) device in order to obtain an N-element
synthetic range profile, which is called range spectrum, where the range resolution becomes
approximately N f 1 / (Nakamura et al., 2010) (Wehner, 1995). For example, suppose
f =34.5MHz and N=30, the resolution is approximately 30 cm which is equivalent to a
very short pulse with 1GHz. Therefore it does not require high speed A/D devices at the
receiver.
The received stepped pulses are phase-detected and the resulting video pulses are then I/Q
sampled. The n-th complex sample R
n
is given by

n n n
R A j = exp( ) (11)
( )
n c
d
f n f
c
= +
2
2 ( 1) , (12)
where A
n
is amplitude of n-th pulse (For a stationary target, A
n
should be approximated by
A), f
c
is fundamental frequency and c is velocity of light. Therefore received pulses (n = 1,
2,,N) are denoted by N f at frequency domain and the range-resolution R is given by

c
R
N f
=
2
(13)
Then, N complex sample is applied to the following IDFT device in order to obtain an N-
element range spectrum.

( ) ( )
N
n
n
R R j n
N
d
c N f
c
N A
d
c N f
N c

=

=

1
2
exp 1
2
sin
2
sin
(14)
It is clear from Eq. 14 that the peak of range spectrum appears at dN f c = 2 / , the
estimated distance d is given by

c
d
N f
= ,
2
(15)
where the maximum detectable range d
max
is given by c f /2 .

111

(a) Block diagram

(b) Signal waveforms
Fig. 7. Block diagram and signals
3.2 Spectrum hole
The stepped-FM scheme also offers spectrum hole (non-activated within a portion of the
radio spectrum) since it consists of independent pulses with different frequency. Thus it is
expected to coexist with the existing systems (Nakamura et al., 2011).
Fig.8 shows the power spectrum for f = 34.5MHz and N=30 where the stepped-FM pulses
with the spectrum-hole of 69MHz from 3.655 to 3.724GHz 6.6% (corresponding to two
consecutive stepped pulses) are not transmitted. The holed band is seen to be by
approximately 13dB suppressed relative to the signal spectrum. Consider the FCC
regulation of -41.3dBm/MHz, the normalized spectrum of the holed band is less than -
55dBm. Therefore it should not interfere with the existing systems unlike a UWB-IR. Fig. 9
shows the range spectrum for a point target at 2.2m where we assumed a spectrum-hole of
6.6% (69MHz band as 3.655~3.724GHz) and a zero-padding for the IDFT processing
(1,024points). Remarkable degradation of range-resolution is not seen, although the range
side-lobe is degraded.

112

Fig. 8. Power spectrum of transmit signal with spectrum-hole of 6.6%

Fig. 9. Range spectra with and without hole. A point target is placed at 2.2m
3.3 Effect of spectrum-hole
The effect of spectrum-hole on the range spectrum is presented where the measurement
specification is shown in Table 1. Two sphere targets with -9dBsm and -15dBsm are
measured in an RF anechoic chamber (Skolnik, 2001) (Nakamura et al., 2011). The
measurement was conducted in an RF absorber where theses targets on turn table were
placed at 2.2m and 3m from the antenna. Fig.10 shows the range spectrum with spectrum-
hole of 6.6%s, which is compared with that without hole. Please note that the other echoes at
0.8m and 1.6m are from the turn table. Fig.11 shows the range spectrum as a function of
rotation angle where the distane from thsese targets to the antenna are almost equal at the
rotation angle of 90 degree. These targets are found to be discriminated because of the range
resolution of approximately 15cm. The measurements were conducted for f = 34.5MHz
and N=30. Consider f = 7.5MHz and N=133, however, the maximum detectable range d
max

is 20m and the range-resolution is approximately 15cm which is applicable to the short-
range automotive radar.
From the measurement results, it can be concluded that the stepped-FM radar without high
speed A/D devices can be coexistent with other narrowband wireless applications.

113
Frequency 3~4GHz
Stepped width f 34.5MHz
Number of step N 30
Stepped cycle 10msec
A/D 10kS/sec
IDFT point 1024
Table 1. Measurement specifications

Fig. 10. Range spectra for two targets when the spectrum-holes is 6.6%

Fig. 11. Range spectrum as a function of for two sphere targets
4. Detection using trajectory estimation
Short-range automotive radar with high range-resolution should suffer from clutter because
of its very broad lateral coverage. It is therefore an important issue to detect moving
automobile in heavy clutter conditions. The clutter may be generally classified from

114
automobile by the Doppler, but it will be difficult for a very short-pulse of UWB-IR radar.
This is because a shorter pulse will have better range-resolution, but poorer Doppler
resolution. Observing the range profile during several micro-seconds, however, each object
echos trajectory is estimated using Hough transformation and the Doppler is then
calculated (Okamoto et al., 2011). When the speed of object is almost constant during the
time, for example, the trajectory is regarded as linear on the time-range coordinate (Hough
space). As a result, moving automobiles are separated from stationary clutter in the Hough
space and detected/tracked with high range. The field measurement results at 24GHz are
presented.
4.1 Time-range profile
Fig.12 shows an example of received range profile on a roadway for a bandwidth of 1GHz.
The profile includes many echoes distinguishable with different delay. Detection,
recognition and tracking of automobile in clutter are very important issues in automotive
radar. Traditionally, the received range profile for each transmit pulse is compared against a
given threshold and a detection decision is made. And once the decision is successfully
done, the range profile is discarded and the next one is considered. This is called threshold
detection. However it is not easy to detect some automobiles simutaneously in heavy clutter
because the automobiles cant be distinguished from clutter in frequency domain. A time-
range profile based detection is useful for the UWB-IR radar where moving automobiles are
classified from clutter by observing the range profile. Fig.13 shows the range profiles as a
function of transmit pulse number, which is called time-range profile. It is seen from Fig.13
that each echos trajectory may be estimated and the Doppler is then calculated.
4.2 Hough transform
Hough transform (HT) has been widely applied for detecting motions in the fields of image
processing and computer vision. Consider the time-range profile as shown in Fig.13, the
time trajectory of each object echo can be estimated by the HT, which is a computationally
efficient algorithm in order to detect the automobile on time-space data map. For example,
the trajectory would be linear for a short duration of 0.1 second or less, thereby the Doppler
can be calculated from the inclination of line.

Fig. 12. Power range profile for a roadway

115

Fig. 13. Time-range profiles for 50 nanosecond pulses
4.3 Automobile classification
A. Measurement set-up and procedure
The measurements were conducted on a roadway as shown in Fig.14. The detail
specification is shown in Table 2. The four automobiles were driven along the roadway and
the received signals were processed on board. A pulse repetition interval (PRI) of 15ms is
considered for the scenario of Fig.14. The antennas with a beam-width of 70 in horizontal
direction were placed 60cm above the ground. Please note the anti-collision radar is
designed for short-range/wide-angle object detection.

(a) Measurement scene (b) Measurement scenario
Fig. 14. Measurement scenario

116
Bandwidth 5GHz, 1GHzcentered at 24GHz
Polarization H-H. plane
Type Double-ridged Horn
Gain 12.5dBi24GHz
Antenna
Height 60cm
Sedan: #1 4.64m1.72m1.34m
SUV: #2 4.42m1.81m1.69m Target
Mini-van: #3 4.58m1.69m1.85m
Table 2. Measurement parameters
B. Measurement results
Fig.15 shows the flow of HT algorithm from time-range profile to trajectory line. The quasi-
images (8bits time-range image) for BW=300MHz and 500MHz are shown in Fig.15(a) and
(b) respectively. Many trajectories are plotted by the Hough space translation. The number
of trajectory lines depends on the signal-to-clutter ratio (SCR) and the window size to
observe the time-range profile. Some trajectory lines of a time-range profile would be
connected to the lines of the following profile. Therefore the trajectory of object echo can be
selected using the continuity between the consecutive time-range profiles, while the quasi
trajectory should be discarded. Fig.16 (a) shows the estimated trajectory lines for a BW of
500MHz. It is seen that many lines are depicted because of significant clutter. Fig.17(b)
shows the survived lines by the algorithm of Fig.15 where three time-range profiles for 20
pulses are used. It is seen that clutter can be estimated from the Doppler. Fig.18 also shows
the estimated lines for a BW of 300MHz. The results of Figs. 17and 18 are found to agree
with the scenarios. The measurements were also conducted for different scenarios of side-
looking and back-looking radar and the trajectory estimation scheme is found to be useful in
order to classify the automobile from heavy clutter.

Fig. 15. Signal flow for HT algorithm

117

(a) BW=500MHz (b) BW=300MHz
Fig. 16. Quasi-images of time-range profile

(a) Estimated trajectory lines by HT (b) Survived trajectory lines
Fig. 17. Estimated trajectory line (BW=500MHz)

Fig. 18. Estimated trajectory line (BW = 300MHz)
5. Target discrimination
Automotive radar is required to detect automobile accurately, but not to detect clutters
falsely, even in complicated traffic conditions. One-dimensional range profile of an

118
automobile target has dependence on the shape because it has some remarkable scattered
centers. Therefore the different types of automobile has different range profile feature which
can be used as a unique template for automobile target discrimination/identification purpose
in tracking mode. That is, the target is detected accurately by the correlation of received
signal with template. The scheme also offers real-time operations unlike two-dimensional
image processing (Overiez et al., 2003) (Sato et al., 2006). The measurement results are
presented for various types of automobile (Matsunami et al., 2009) (Matsunami et al., 2010).
5.1 Target discrimaination and identification
Figs.19(a)-(c) show the measured range profile for various bandwidths where a sedan typed
automobile was placed at approximately 10m. Please note that the profiles are expressed as
a function of range-bin corresponding to the range-resolution (=1/BW). Echoes from
various objects are found to be distinguished for wider bandwidth. It is seen that there exist
some remarkable scattered centers. However the feature is not so clear because of
scintillation and noise. Figs.20-22 show range profiles for various bandwidth where the non-
coherent integration of 50 pulses was conducted in order to reduce the scintillation and
noise. For the dedan, some strong echoes are seen from the side mirror and interior, and the
SUV shows a unique feature.

(a) BW=500MHz

(b) BW=1GHz

(c) BW=5GHz
Fig. 19. Power range profiles for various values of BW. A sedan was placed forward the
radar antenna where the antenna to target separation was approximately 10m

119

(a) Sedan (b) Mini-van

(c) SUV (d) Mini-truck
Fig. 20. Unique profiles of automobile (BW=5GHz)


Fig. 21. Unique profiles of automobile (BW=1GHz)

120


Fig. 22. Unique profiles of automobile (BW=500MHz)
5.2 Profile matching
Range profiles have been measured for four automobile #1~#4 (sedan, mini-van, SUV and
mini-truck) which have been processed as the template. And the profile matching rate is
calculated for various unknown automobiles. The matching rate is shown in Table.3-5. For
BW=500MHz or more, it is higher than 96% when the automobile is the same as the
template and each automobile can be detected. Assuming a correlation value of 0.6 for the
discrimination, each profile can be identified in clutter since it has unique feature with god
cross-correlation.

Subject vehicle
Template
Sedan Mini-van SUV Mini-truck
Sedan 99.1 22.5 26.9 19.9
Mini-van 13.1 96.9 14.9 15.5
SUV 8.6 4.6 98.6 21.6
Mini-truck 16.8 20.8 15.3 98.2
Table 3. Matching rate [%]BW=5GHz

121
Subject vehicle
Template
Sedan 99.4 30.4 53.7 6.2
Mini-van 19.5 96.0 24.8 28.3
SUV 45.1 19.3 99.3 18.4
Mini-truck 14.7 24.8 31.7 98.6
Table 4. Matching rate [%]BW=1GHz

Subject vehicle
Template
Sedan 99.9 38.0 76.6 33.5
Mini-van 38.0 98.9 19.2 31.4
SUV 55.3 25.3 98.2 31.7
Mini-truck 31.2 20.0 33.0 99.3
Table 5. Matching rate [%]BW=500MHz
6. Conclusion
UWB-IR short-range radar a5 24/26GHz will be used for various applications such as pre-
crash detection and blind spot surveillance. The short-range radar has a few significant
problems to be overcome such as multiple targets detection and clutter suppression.
This chapter has presented how to detect multiple automobile targets in clutter. The
presented results are as follows;
UWB-IR radar requires high speed A/D devices to synchronize and detect the received
nanosecond echo, thereby the system becomes very complicated and expensive. In
section 3, the use of stepped-FM scheme which does not require high speed A/D has
introduced for UWB-IR radar. In addition it offers spectrum hole to coexist with
existing wireless systems.
UWB-IR short-range radar is expected to provide a wide coverage in azimuth angle.
Therefore, increased clutter makes it difficult to detect multiple automobile targets.
Section 4 has introduced a multiple target detection scheme in heavy clutter using the
trajectory of radar echoes.
Section 5 has introduced a target identification scheme in order to improve the
detection performance where a power delay profile matching is employed and the
usefulness has been demonstrated by the measurement at 24GHz. The results have
showed that automobile targets can be recognized and identified.
7. References
Skolnik, M. (2001). Introduction to Radar systems, 3rd ed., McGraw-Hill, ISBN0-07-288138-0,
New York

122
Taylor, J. D. (1995). Introduction to Ultra-wideband Radar Systems, CRC Press LLC, ISBN0-
8493-4440-9, Wsshington, D.C..
Matsunami,I.; Nakahata, Y.; Ono, k. & Kajiwara,A. (2008). Empirical Study on Ultra-
wideband Vehicle Radar, Proc. of IEEE Vehicular Technology Conference, ISBN 978-1-
4244-1722-3, 8G-5, Calgary, Sept. 2008.
Nakamura,R.; Yokoyama,R. & Kajiwara,A. (2010), Short-Range Vehicular Radar Using
Stepped-FM Based UWB-IR, Proc. of IEEE Radio and Wireless Symposium, ISBN 978-
1-4244-4726-8, New Orleans, Jan. 2010.
Wehner, D. R. (1995). High-Resolution Radar, Artech House, ISBN978-0-89006-727-7, pp.197-
255, 1995.
Nakamura,R. & Kajiwara,A.(2011), Empirical Study on Spectrum-Hole Characteristics of
Stepped-FM UWB Microwave Sensor, to be appeared in Proc. of IEEE Radio and
Wireless Symposium, Jan. 2011.
Okamoto,Y.; Matsunami,I. & Kajiwara,A.(2011), Moving vehicle discrimination using
Hough, transformation, to be appeared in Proc. of IEEE Radio and Wireless
Symposium, Jan. 2011.
Ovariez,J.P.; Vignaud,L.; Castelli,J.C.; Tria, M., & Benidir,M.(2003). Analysis of SAR image
by multidimensional wavelet transform. IEE Proc. Radar Sonar Navig., pp.234-241,
Aug.2003.
Sato,T. & Sakamoto,T(2006). Reconstruction Algorithms for UWB Pulse Radar Systems,
IEICE Trans. Comm., ISBN1344-4697, vol.J88-B, pp.2311-2325, Dec.2006.
Matsunami,I. & Kajiwara,A.(2009). Power Delay Profile Matching for Vehicular Radar, Proc.
of IEEE Vehicular Technology Conference, ISBN 978-1-4244-2514-3, 5E-1, Anchorage,
Sept. 2009.
7
An Ultra-Wideband (UWB) Ad Hoc Sensor
Network for Real-time Indoor Localization of
Emergency Responders
Anthony Lo
1
, Alexander Yarovoy
1
, Timothy Bauge
2
, Mark Russell
2
,
Dave Harmer
2
and Birgit Kull
3

1
Delft University of Technology,
2
Thales Research & Technology Limited,
3
IMST GmbH,
1
The Netherlands
2
UK
3
Germany
1. Introduction
A localization system is a network of nodes, which is used by an unknown-location node to
determine its physical location. The Global Navigation Satellite System, GNSS (Hofmann-
Wellenhof, 2008) is an example of a widely used outdoor localization system. However,
outdoor localization systems perform poorly in indoor environments due to strong signal
attenuation and reflection by building materials, and no line-of-sight propagation. Thus,
Indoor Localization Systems (ILSs) are needed to provide similar localization inside
buildings. ILSs have many potential applications in the commercial, military and public
safety sectors. This chapter focuses on the public safety application. The considered ILS is
used to track emergency responders, e.g. fire-fighters and policemen, who carry out search
and rescue missions in the disaster zone such as building fires and collapsed tunnels. Such
an ILS was first crystallized in the EUROPCOM (Emergency Ultra wideband RadiO for
Positioning and COMmunications) project (Harmer, 2008; Harmer et al., 2008). The
EUROPCOM system is an ad hoc sensor network which comprises a small number of base
or reference nodes deployed outside surrounding a building, and the rest of the nodes are
unknown-location nodes which are worn and deployed by emergency responders entering
the hostile building. The unknown-location node is self-localized by collectively
determining its position relative to base nodes. Additionally, the unknown-location node is
also allowed to determine its position relative to neighboring unknown-location nodes. This
greatly enhances the accuracy and robustness of the ILS. It is fully autonomous and can be
rapidly deployed with little human intervention.
Ultra-WideBand (UWB) is the radio transmission technology used by the EUROPCOM
system. A UWB signal is defined to be one that possesses an absolute bandwidth of at least
500 MHz or a fractional bandwidth larger than 20% of the center frequency. Currently,
several UWB technologies exist, namely direct sequence UWB, impulse radio UWB, Multi-

124
band Orthogonal Frequency Division Multiplexing (MB-OFDM) UWB, Chaotic UWB, and
Frequency Hopping (FH-UWB). The EUROPCOM system selected FH-UWB because it
offers significantly better range and position accuracy than other technologies such as pulse
UWB (Frazer, 2004).
A great deal of effort has been expended on localization algorithms, but the Medium Access
Control (MAC) and routing protocols for ILS have received very little attention yet. Unlike
other ad hoc sensor networks, the considered ILS exhibits unique characteristics. Therefore,
it poses new technical challenges in the MAC and multi-hop routing protocol design. Firstly,
the ILS is heterogeneous in the sense it is composed of different types of nodes with varying
capability, processing power and battery energy. Secondly, the ILS operates in a highly
dynamic and hostile environment. Lastly, emergency applications require fast localization in
the order of seconds. In order to address these challenges, we propose a novel Self-
Organizing Composite MAC (SOC-MAC) protocol and a Lightweight and robust Anycast-
based Routing (LAR) protocol. Cross-layer approach is present in the design to attain highly
optimized, bandwidth- and energy-efficient protocols.
2. Network architecture of an Indoor Localization System (ILS)

CU
BU
DU
MU
MU
Incident Zone
n
o
n
-
U
W
B

l
i
n
k
(
I
m
p
l
e
m
e
n
t
a
t
i
o
n

d
e
p
e
n
d
e
n
t
)
U
W
B

L
i
n
k
BU
BU
BU
BU
MU
MU
DU
DU
Legend: BU Base Unit
CU Control Unit
DU Dropped Unit
MU Mobile Unit

Fig. 1. Network Architecture of an Indoor Localization System
The assumed ILS, which is an ad hoc sensor network, consists of four types of nodes: a
Control Unit (CU), Base Units (BUs), Dropped Units (DUs) and Mobile Units (MUs), as
shown in Fig. 1. The MU is a sensor that is worn by every emergency responder. The MU
has the capability to calculate its position which is in turn delivered to the CU. The BUs are
located outside and around the incident area, while maintaining wireless connectivity with
An Ultra-Wideband (UWB) Ad Hoc Sensor Network for
Real-time Indoor Localization of Emergency Responders

125
the emergency responders inside the building. Unlike other units, the position of BUs is
known and most likely to be acquired through GNSS. Furthermore, the BUs will remain
stationary throughout the entire mission. The DUs are strategically placed in the incident
area by emergency responders to serve as relay nodes once the MUs lose wireless
connectivity with the BUs. Similar to MUs, the DUs can determine their positions and relay
them to the CU. The CU provides the main visual display to the rescue coordinators,
showing the current position and direction of movement of individual emergency
responders with respect to the incident area topology, e.g. a building. As shown in Fig. 1,
the ILS is composed of a UWB subnetwork and a non-UWB subnetwork. The reason for two
separate subnetworks is that the CU is not involved in the localization process. Thus, more
radio resources are available for the UWB subnetwork, in particular, when the number of
MUs increases.
2.1 System assumptions
In this subsection, we state several assumptions made in the design of the MAC and routing
protocols. The MAC and routing protocol design assumes the FH-UWB technology is
employed by the Physical layer of the BU, the DU and the MU. The operating bandwidth of
the FH-UWB units is 1.25 GHz which consists of 125 carrier frequencies. This means, the
carrier spacing is 10 MHz. The center frequency is located at 5.1 GHz. Each unit follows a
fixed hop pattern. The pair CU-BU communicates over a non-UWB link. Similarly, the BU-
BU transmission is also over non-UWB links. The rationale for using a non-FH-UWB
technology is that more radio resources are available to the FH-UWB subnetwork. Since the
non-FH-UWB technology is implementation-dependent, we will not further deal with the
specifics of the non-UWB technology in the rest of the chapter. The design of the MAC and
routing protocols is described in subsections 2.2 and 2.3, respectively.
2.2 A Self-organizing Composite Medium Access Control (SOC-MAC) protocol
As each MU is mobile, it will determine and transmit its position information to the CU
periodically. For instance, in order to cope with user mobility in the order of 0.5 m/s
(walking speed), an MU needs to measure and transmit position information to CU at a rate
of one position packet per second. As a result, SOC-MAC is based on the Time Division
Multiple Access (TDMA) because such a MAC is particularly suited to the periodic nature of
localization process. Unlike traditional TDMA, SOC-MAC is designed for ad hoc networks
with no requirement for a central controller for allocating time slots as it is self-organizing.

RA-TDMA
A-TDMA I-TDMA
Reserved TDMA

Fig. 2. SOC-MAC

126
Using SOC-MAC, each unit can autonomously select and reserve time slots based only on
local network knowledge without the need of dedicated signaling messages.
As shown in Fig. 2, the SOC-MAC protocol operates in two phases: a Random Access
TDMA (RA-TDMA) phase and a Reserved TDMA phase. The former phase is invoked by a
unit prior to joining the network or when the unit has not used any time slots in the
previous superframe; thus, there are no reservations in the current frame. In this phase, a
unit acquires a time slot through a random access mechanism. Once a time slot has been
acquired, SOC-MAC will enter the Reserved TDMA phase. The RA-TDMA phase and the
Reserved TDMA phase are described in the following subsections.
2.2.1 Random Access TDMA (RA-TDMA) phase
The UWB medium is segmented into SOC-MAC superframes, each of which has a constant
period of T seconds. Each superframe is in turn partitioned into N orthogonal time slots of
duration T/N seconds. The start of the superframe is provided by one of the BUs, known as
the Master BU (MBU). Naturally, MBU will occupy the first time slot. Units unable to hear
the transmissions of MBU will synchronize to the TDMA frame by monitoring the
transmissions of other neighboring units, which will identify the time slot in which they are
transmitting. From this time slot number information, the start of the frame can be inferred.
Fig. 3 depicts the structure of the superframe and time slot. Each time slot can be used for
either data transmission (referred to as data slot) or ranging (referred to as ranging slot).
The latter time slot format is specifically used for determining the range between two units.
Thus, the ranging slot can accommodate a very limited payload. The MAC header includes
identifiers of up to five ranging units, denoted as R1 ID to R5 ID, which have been
selected to respond to ranging requests. The final part of the ranging slot is reserved for the
corresponding pong responses from R1 ID up to R5 ID. Unlike ranging-slot, data-slot
is purely utilized to transmit user data and can accommodate larger payload. The MAC
header of the two slot structures contains similar fields except the ranging-slot includes the
ranging unit identifiers and the position data of the transmitting unit (i.e., TX ID POS).
Hence, the MAC header of the ranging-slot is longer and has to be split into two parts
separated by two pilot tones as shown in Fig. 3.
In general a unit enters the RA-TDMA phase prior to joining the network. Fig. 4 contains the
flow chart of the RA-TDMA phase. Before the unit can transmit in a time slot, it must listen
to the physical channel for at least one complete TDMA superframe period. During this
period, the unit constructs a list of one-hop neighbors and a map of their time-slot usage.
Based on the time-slot usage map, the unit derives a list of vacant time slots in the
forthcoming superframe. The number of vacant slots in the list is denoted as candidate slot
counter (csc) in Fig. 4. When a first vacant time slot in the next superframe arrives, the p-
persistent algorithm is applied to determine if this vacant time slot can be used for
transmission. The p-persistent algorithm defines two parameters, namely P1 and P2. P2 is
inversely proportional to csc, and P1 is randomly selected from an interval [0 .. 1]. If P1 is
equal to, or less than P2, then the vacant time slot is reserved and transmission should occur
in the reserved time slot. If not, the number of vacant time slots csc in the list is decremented
by one and the same procedure is repeated for the next vacant time slot. The p-persistent
algorithm minimizes the chance that two or more units in the RA-TDMA phase are
contending for the same time slot. A low csc increases the probability of selecting the next
vacant time slot. The number of unsuccessful attempts in reserving a time slot is recorded in
attempt count (ac). Once a vacant time slot has been successfully reserved, the RA-TDMA
phase ends and the reserved TDMA phase sets in to complete the channel access procedure.

127

0 1 2 3 4 5 6 N 0 1 2 3 4 5 6 N
Pilot Pilot
MAC
Header
(Part I)
Pilot Pilot R0 R1 R2 R3 R4 R5
Responses of 5 Units Data from Transmitting Unit
MAC Header
(Part II)
and Payload
Superframe 0
T
Superframe 1
T
Pilot Pilot
MAC
Header
Pilot Pilot Pilot Pilot NWK Payload
Ranging-Slot Format
Data-Slot Format
FEC
Rate
MLEN SYNC
TX
ID
R0
ID
R2
ID
R1
ID
R4
ID
R3
ID
MAC
Info
(I)
CRC
FEC
Rate
MAC
Info
(II)
NWK
Payload
CRC
TX ID
POS
RX
ID
MAC
Reserved
ST
Msg
Type
Submsg
FEC
Rate
MLEN SYNC
TX
ID
Bit Sequence for identifying
Data Slot
MAC
Info
CRC
ST
Msg
Type
Submsg
(*)
RX
ID
MAC
Reserved
SLOTNUM Unused
SOFF Unused
SOFF LEN
(*) This field has the same structure as the submsg field in the Ranging Slot
(+) This field has the same structure as the NWK Header in the Ranging Slot
or
or
(Submessage Type 1)
(Submessage Type 2)
(Submessage Type 3)
For A-TDMA when ST > 0
For A-TDMA when ST = 0
For I-TDMA to reserve extra time slots
NWK
Header
NWK
Header
(+)
Hop
Count
Congestion
Level
Reserved

Fig. 3. SOC-MAC Superframe

128

Is slot vacant
?
Analyze next slot in
forthcoming superframe
Select random
number , p1, from [0,1]
P1 P2
?
A slot is selected ;
RA-TDMA phase ends ;
Reserved TDMA phase starts ;
ac = 0;
ac = ac + 1;
csc = csc 1;
P2 = 1/csc;
Yes
Yes
No
No
Start
csc = number of
vacant slots ;
P2 = 1/ csc;
ac = 0;
End
Scan one complete
superframe

Fig. 4. Flow Chart of RA-TDMA
2.2.2 Reserved TDMA phase
The Reserved TDMA mode comprises two operations, namely the Autonomous TDMA (A-
TDMA) and the Incremental TDMA (I-TDMA). The latter is used to acquire additional slots
in the same superframe in addition to the one acquired in the RA-TDMA phase. A-TDMA is
responsible for managing the acquired time slots.
A-TDMA
Once a slot has been acquired through RA-TDMA and/or I-TDMA, the same time slot is
automatically reserved for the next Slot_Timeout (ST) superframes, where ST is randomly
picked from an interval [1 .. MAX_TIMEOUT]; MAX_TIMEOUT is a MAC design parameter.
A-TDMA is responsible for keeping ST up-to-date. That is, ST is decremented by one in each
new superframe. ST is included in the MAC header so that other units can determine when the
time slot will be free. When ST > 0, the Submessage Type 1 is used, which contains the time-slot
number (SLOTNUM) of the currently reserved time slot as illustrated in Fig. 3. When a time
slot expires (i.e., ST = 0), A-TDMA randomly chooses a vacant time slot in the next superframe
from a list of vacant time slots in the time-slot usage map, and pre-announces to the other

129
units the offset between the present time slot and the newly selected time slot (SOFF
expressed in number of time slots) using the format Submessage Type 2 in the current
superframe as shown in Fig. 3. This allows other units to find this unit in the next superframe
without searching and to update the time-slot usage map. The new time slot will only be used
in the next superframe. The continuous change of time-slot positions ensures that if two or
more units had chosen the same time slot in the RA-TDMA phase, the collision can only
persist for a maximum of MAX_Timeout superframes before one or all involved units must
choose a different time slot. Thus, the collision is resolved through a probabilistic means.
Hence, MAX_TIMEOUT must be small in order to reduce the number of collisions that are
energy-wasting. On the other hand, if MAX_TIMEOUT is too small then neighbor units need
to perform frequent updates on the time-slot usage map, which in turn increases power
consumption. The new time slot is assigned a new ST value which is obtained using the same
process as described above. The A-TDMA algorithm is depicted in Fig. 5.

Is
Slot_Timeout
Expired ?
Check Slot _Timeout
Select a random
slot from a set of vacant slots
Yes
No
Pre-announce selected slot in
Current superframe
Start
End

Fig. 5. Reserved TDMA Phase: ATDMA Operation
I-TDMA
The time slot acquired during the RA-TDMA phase is the first and only one for each unit. If
a unit needs extra time slots, then I-TDMA is employed to reserve the extra time slots in the
same superframe to increase the data rates. I-TDMA calculates the number of required time
slots N
I
based on the actual queue length provided by the Network layer. It searches for a
block of N
I
successive vacant time slots in the time-slot usage map. If not available, N
I
is
reduced until the search is successful. The number of reserved time slots (LEN) and the
offset (SOFF) between the current and the first new time slot are advertised using the format
Submessage Type 3, refer to Fig. 3, so that all other units are informed about the new
reservations. In principle each new time slot can be used for another I-TDMA operation so
that the number of time-slot reservations can grow more rapidly. Hence, the usage of I-
TDMA needs to be restricted if the channel is busy and the number of vacant time slots is
small. Note that in almost all cases the A-TDMA operation is required, while the I-TDMA

130
operation is only sporadically needed to increase the data rates by reserving additional time
slots. In order to free reserved time slots, the time slots are simply not renewed by A-TDMA
after Slot_Timeout superframes.
2.3 A Lightweight and Robust Anycast-based Routing (LAR) protocol
The Lightweight and Robust Anycast-based Routing (LAR) protocol routes data packets
from MUs or DUs to the nearest BU. There is no exact destination BU for a data packet.
Thus, routing decisions must rely on routing parameters and packet types. LAR defines two
routing parameters, namely hop count and congestion level. Hop count indicates the distance
of a unit (in terms of the number of hops) to a reference BU. It increases monotonically at
each hop. Congestion level is used to indicate the buffer occupancy of a unit. These routing
parameters are not disseminated using dedicated routing packets but carried and
propagated in the Network (NWK) header of data packets. Thus, LAR does not incur
routing packet overheads. The format of the NWK header is depicted in Fig. 3. This means
that irrespective of the data type, the NWK header always contains the mandatory routing
parameters. The NWK header occupies 12 bits in a total of 1831 bits in one time slot of the
SOC-MAC superframe. Therefore, the overheads of the NWK header for routing are less
than 1%, which conserves bandwidth and energy.
Route establishment is initiated by BUs to form spanning trees rooted at each BU. This is a
natural choice because each BU periodically broadcasts its position which is known
beforehand, while DUs and MUs just listen to the BU broadcasts since they need to
determine their position. The BU sets the initial value for the hop count and congestion
level. From the BU broadcasts, the DUs/MUs create a new entry in the routing table if it
does not exist. The routing table entry contains the following fields: neighbor unit id, hop
count, congestion level, FEC level and the expiration time of the entry. The first field identifies the
address of the unit that broadcasts the data packet, which represent the next-hop unit for the
route towards a destination BU. The neighbor unit id is contained in the MAC header. Note
that the unit maintains only the next-hop routing state, which provides the routing protocol
with a high degree of scalability. The hop count in the routing table is incremented by one
with respect to the received hop count. For instance, if the incremented hop count is n+1
then the unit is n+1 hops away from the destination BU. The congestion level field is
extracted from the NWK header. FEC (Forward Error Correction) level determines the
channel bit rate for communicating with the next hop of the neighbor unit id. Four FEC
levels, viz., FEC-1 to FEC-4, are defined. FEC-4 provides the highest bit rate but no or the
lowest level of error protection. The FEC level is also contained in the MAC header. Once a
DU/MU has determined its position, it can broadcast its position. The hop count in the
NWK header is obtained from the selected route in the routing table entry, while the
congestion is set to the maximum of its congestion level and that in the routing table entry.
In the case of multiple entries in the routing table, a route selection algorithm with load
balancing is used to choose the next hop. The algorithm will be described in subsection 2.3.1.
So far, we have focused on route construction from an MU/DU to a destination BU, which is
referred to as forward route. A reverse route (from a BU to an MU or DU) can be
constructed using data packets sent on the forward route. One such data packet is position
reporting which is used to transport position data to the BU. Position reporting packets are
periodically sent by an MU and DU. The position reporting packets are transmitted using a
forward route selected by the unit to a destination BU. All units along the forward route
store the source and forwarding unit identifiers in their routing table. The latter identifier is

131
the address of the intermediate unit that forwards the data packet while the source identifier
is the address of the unit which generates the position reporting packets. No other routing
parameters are needed for the reverse route. Since the position broadcasting and position
reporting are periodic, the forward and reverse routes are always up-to-date. Therefore, no
specific route recovery or maintenance functions are required.
2.3.1 Load balancing
Load balancing is achieved using the congestion level parameter, which is based on the
occupancy of queues in a unit. The queues allocated by a unit are assumed to be fixed size.
The congestion level is then deduced from the queue occupancy as shown in Table 1.

Congestion Level Queue Occupancy Definition
0 - 2 20% - 40% full Not congested
3 - 4 50% - 60% full Slightly congested
5 - 6 70% - 80% full Congested
7 90% full Heavily congested
Table 1. Congestion Levels for Load Balancing
2.3.2 Route selection

next -hop candidate
Is candidate hop
count smallest ?
next -hop candidates with
smallest hop count
Is candidate
Congestion level
lowest?
Advertised selected
next hop
No
Yes
No
Yes
Yes
Start
End

Fig. 6. Next-hop Selection Algorithm with Load Balancing

132
Hop count is the primary routing metric, while congestion is the secondary metric due to the
delay at the MAC layer, which cannot be tolerated by real-time data packets. In the case of
multiple entries in the routing table, LAR must select the candidate route with the smallest
hop count. If there are several candidate routes with the same hop count then the candidate
with the lowest congestion level is picked. By selecting the candidate route with the smallest
hop count the selection algorithm can guarantee loop-free delivery as a data packet is
always forwarded from a unit with a higher hop count to a unit with lower hop count. The
selection algorithm is shown in Fig. 6.
3. Simulation set-up
The feasibility and performance of SOC-MAC and LAR are evaluated by means of
simulation. To this end, we extended the Mobility Framework (MF) (Mobility Framework)
module by incorporating a model for a UWB Physical layer, the SOC-MAC protocol, the LAR
protocol and the Application layer, and the ILS network entities. MF is an add-on package for
simulating mobile and wireless networks on the OMNeT++ platform (OMNeT++) which is a
powerful generic, object-oriented and discrete-event simulation tool. Naturally, MF can be
easily extended for simulating the ILS network. Thus, three new simulation nodes, namely
BUhost, DUhost, and MUhost, were defined. These nodes correspond to the units BU, DU
and MU, respectively. CU was not modeled because it is in the non-UWB subnetwork which
is implementation-specific. Fig. 7 depicts a sample of the simulation network, which consists
of four BUhosts, two DUhosts and four MUhosts. In each of the simulation nodes, three
protocol models, viz., the application, the network and the Network Interface Card (NIC)
were defined as extensions to the corresponding models in MF. The internal structure of the
node is shown in Fig. 8(a). The Blackboard and Mobility models were used without
extensions. Note that BUhost, DUhost and MUhost have the same internal node structure.
The application model, EuropAppLayer, the network model, EuropNetwLayer, the MAC model,
EuropMacLayer, and the Physical model are described in the next subsections.

Fig. 7. Simulation Network

133

EuropAppLayer
EuropNetwLayer
EuropNIC
BU/DU/MUhost
Blackboard
Mobility
EuropMacLayer
SnrDecider
EuropSnrEval
NIC
(a) (b)

Fig. 8. Node and NIC Structure
3.1 Physical layer model
The Physical model is divided into EuropSnrEval and SnrDecider as shown in Fig. 8(b). The
former was extended from SnrEval in MF while the latter was used as it is. EuropSnrEval is
used to calculate the Signal-to-Interference-plus-Noise (SINR) of a received MAC frame. The
SINR is defined as

, max
10log
r
n
P
SINR
P I
=
+
(1)
where P
r,max
is the strongest received signal power among the received signals, based on the
capture effect (Rappaport, 2001). P
n
is the Additive White Gaussian Noise (AWGN). I is the
interference power which is defined as the sum of all received signal power excluding P
r, max
.
The interference power I is expressed as

,max r
r
P
I P
= (2)
In case of collision-free transmission, the term I is null. Hence, Equation (1) is reduced to

, max
10log
r
n
P
SINR
P
= (3)
The computed SINR is passed to SnrDecider which determines whether the MAC frame is
correctly received or not. A MAC frame is considered to be correctly received, if SINR
SINR
th
, where SINR
th
is the SINR threshold. A correctly received frame is delivered to the
EuropMacLayer, otherwise it is discarded. SINR
th
was obtained through physical layer
simulation, which produces Bit Error Rate (BER) plots as a function of SINR. Given a target
BER, SINR
th
is deduced. The physical layer simulation was carried out separately using
another tool since OMNeT++ and MF lack the support for simulating physical layer
functions such as frequency hopping, channel coding, modulation, and signal processing.

134
The received power P
r
in Equation (1) is characterized by large-scale fading and small-scale
fading. Large-scale fading represents the average signal power attenuation when
transmitted through the medium. The attenuation or commonly known as Path Loss (PL) as
a function of distance is expressed as (Rappaport, 2001).

/ 10
0 0
0
( ) ( )( ) ,
10
x
d
PL d PL d d d
d

= (4)
where d
0
is the reference distance, is referred to the path loss exponent, and X
denotes the
log-normal shadowing effect with a zero-mean normal distribution (in dB) and standard
deviation (also in dB). PL(d
0
) is evaluated using the free-space path loss equation or by
conducting measurements. In our work, PL(d
0
) was determined using the free-space path
loss equation which is given by (Rappaport, 2001)

2
0
0
4
( )
c
f d
PL d
c

=

(5)
where
max min
.
2
c
f
f f
=
+
f
min
and f
max
are the lower and upper boundary of UWB transmission
frequency band, respectively. Substituting Equation (5) into Equation (4), and let d
0
= 1 m in
our case, we can rewrite Equation (4) as

2
/10
0
4
( ) 10 ,
x c
f
PL d d d d
c

=

(6)
Small-scale fading represents the wide variations in received signal strength caused by
interference between two or more versions of the transmitted signal arriving at the receiver
at slightly different times. It is typically modeled by the Ricean distribution or the Rayleigh
distribution when there is a line-of-sight or non-line-of-sight, respectively. In UWB systems,
the signal power variations due to small-scale fading are not severe due to the ultra-large
bandwidth of UWB signals and diversity techniques used in the physical layer. Thus, in our
physical channel model, we are only concerned with the large-scale fading. Hence, the
received power P
r,max
in Equation (1) and P
r
in Equation (2) can be calculated using

( )
t
r
P
P
PL d
= (7)
where PL(d) is given in Equation (6), and P
t
is the transmit power.
3.2 MAC layer model
The MAC model, EuropMacLayer, captures the complete functionality of SOC-MAC
described in Section 2.2. It was derived from the BasicMacLayer model of MF. The model
definition consists of three parts, referred to as a EuropMacLayer module definition, a
EuropMacLayer protocol data unit definition, and a EuropMacLayer module
implementation. The EuropMacLayer module definition, which is specified using the
OMNeT++ NED language. The EuropMacLayer protocol data unit definition, called
EuropMacPkt, was derived from the MacPkt definition of MF. The derived module contains

135
the fields of the EuropMacLayer protocol data unit only. The EuropMacLayer module
implementation contains the algorithms of the composite MAC. Unlike the EuropMacLayer
module definition and EuropMacPkt definition, this module was directly written in the C++
programming language. The EuropMacLayer module definition and EuropMacPkt are
translated into C++ code when an executable of the simulation program is built.
3.3 Network layer model
The Network model, EuropNetwLayer, implements the LAR protocol described in Section
2.3. It was derived from the SimpleNetwLayer model of MF. Similar to the MAC model, it
consists of three parts: a EuropNetwLayer module definition, a EuropNetwLayer protocol
data unit definition, and a EuropNetwLayer module implementation. The EuropNetwLayer
protocol data unit definition, called EuropNetwPkt, was derived from the NetwPkt
definition of MF.
3.4 Application layer model
The application traffic model generates dummy position packets of fixed size at regular
intervals. The dummy position packets carry no real position information and the simulated
nodes do not perform position estimation. This does not affect the performance of SOC-
MAC and LAR as long as the application model can mimic the traffic behavior of the real
system. The application traffic model, called EuropApplLayer, which was derived from
BasicApplLayer of MF. The application traffic model also consists of three parts: a
EuropApplLayer module definition, a EuropApplLayer protocol data unit definition, and a
EuropApplLayer module definition.
4. Simulation results
4.1 SOC-MAC performance
We analyze the performance of SOC-MAC. The performance measures for SOC-MAC are
the successful SOC-MAC packet reception rate and the network throughput. An SOC-MAC
packet consists of a header and payload for both the data- and ranging-slot as illustrated in
Fig. 3. Thus, in one time slot, only one SOC-MAC packet is transmitted. The successful
packet reception rate P in the network is defined as the total number of SOC-MAC packets
received by all units divided by the total number of SOC-MAC packets transmitted by all
units in the network. Hence, P is expressed as

1
1
( 1)( )
M
i
i
M
j
j
r
P
M b
=
=
=

(8)
where r
i
is the number of MAC packets received by the ith unit, and b
j
is the total number of
MAC packets transmitted by the jth unit. M is the total number of units in the network. The
scale factor in the denominator of Equation (8) is due to the fact that a packet transmitted by
jth unit is received by all the other M 1 units in the single hop case. Therefore, P is unity in
an ideal case.
Network throughput is defined as the total throughputs of all units, where the throughput
of a unit is the amount of successfully received MAC frames in bits per second. The network

136
throughput is normalized to the channel capacity. Thus, the normalized network
throughput S is defined as

1
M
i
i
x
S
C
=
= (9)
where x
i
is the throughput (in bits per second) of the ith unit, C is the channel capacity in bits
per second, and M is the total number of units in the network.

Parameter Value
Number of units M = (MU + DU + BU) From 10 to 360
Area X Y Z 40m 40m 3m
Superframe duration T 4s
Number of time slots N 160
Time slot duration T/N 25ms
Number of time slots
allocated per Superframe
1 and 4 time slots
MAC
MAX_TIMEOUT 2, 4, 6, 8, 10
Transmit Power P
t
0.11 mW
AWGN P
n
-115.1 dBm
Receiver Sensitivity -120 dBm
Path loss exponent 3.5
Shadowing standard
deviation
0dB, 2dB, 4dB,
8dB
SINR
th
-5 dB
Bandwidth 1.25 GHz
f
min
, f
max
6 GHz, 7.25 GHz
Propagation
model
Center frequency f
c
6.625 GHz
Application Traffic model
0.25 packet/s,
1 packet/s
Simulated time 600 s
Number of simulation runs 30
Table 1. Simulation Settings
The simulation parameter settings are given in Table 1. In the simulation, the total number
of units was varied from 10 to 360 units. Since we fixed the number of BUs to four units, the
number of MUs and DUs was varied but always at an equal quantity. The DUs and MUs
were randomly distributed in a square region with area X Y m
2
. The length of X and Y is
calculated such that signals can still be detected by the receiving units which are at the
maximum distance from the transmitting units using the transmit power, AWGN, path loss
exponent , receiver sensitivity, and SINR
th
given in the table. The BUs were placed around
the edge of the area in order to reflect the arrangement of the real system.

137
4.1.1 Effects of shadowing
Figs. 9 and 10 depict the successful MAC packet reception rate and network throughput as a
function of the number of units without the shadowing effect, i.e., = 0 dB. In an area of
40m 40m, all units are within the radio transmission range of each other. This means that
the transmission of a unit is heard by all the other M 1 units. In Figs. 9 and 10, two sets of
similar simulation runs were carried out. In the first set, each unit reserved only one time
slot in each SOC-MAC superframe, while in the second set, four time slots were reserved by
each unit per SOC-MAC superframe. In the second set of the simulation run, the RA-TDMA
phase is followed by I-TDMA to reserve the extra three time slots. The I-TDMA phase is not
triggered in the first run since only one time slot is required, which is already reserved in
the RA-TDMA phase. The plots for one time slot and four time slots, which are denoted as
1-timeslot (solid line) and 4-timeslot (dashed line), respectively in Figs. 9 and 10, exhibit
similar behavior except the roll-off of the frame reception rate and the maximum network
throughput occurs at different number of units. When the time slot occupancy is less than
the number of time slots in a superframe, i.e., N = 160, the network throughput increases
linearly and the successful frame reception rate is 100%. For N = 160, the network can
accommodate a maximum number of M = 160 or 40 units for 1-timeslot and 4-timeslot,
respectively. When M = 160 or 40, the successful reception rate drops to 90% for 1-timeslot
and 4-timeslot, respectively, and the network throughput reaches the peak, which is 90%, as
observed in Fig. 10 and then gradually falls as M increases. When M < 40 or 160, the effect of
time slot collisions during the RA-TDMA phase is negligible. Thus collisions are resolved
when the involved units enter the A-TDMA phase, which randomly picks a free time slot in
the next superframe.

0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
S
u
c
c
e
s
s
f
u
l

F
r
a
m
e

R
e
c
e
p
t
i
o
n

R
a
t
e
,

P

1-timeslot
4-timeslot
MAX_TIMEOUT = 4
= 0 dB

Fig. 9. Successful SOC-MAC Packet Reception Rate versus Number of Units for = 0 dB,
and MAX_TIMEOUT = 4, and = 0 dB

138
0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
N
o
r
m
a
l
i
z
e
d

M
A
C

N
e
t
w
o
r
k

T
h
r
o
u
g
h
p
u
t
,

S

1-timeslot
4-timeslot
MAX_TIMEOUT = 4
= 0 dB

Fig. 10. Normalized Average SOC-MAC Network Throughput versus Number of Units for
MAX_TIMEOUT = 4, and = 0 dB
When M > 40 or 160, both the successful frame reception rate and the network throughput
rapidly deteriorate due to a larger number of collisions occurred in the RA-TDMA phase.
Unlike in the case of M 40 or 160, the A-TDMA is unable to resolve all the collisions because
the number of time slots in the superframe is insufficient to accommodate the number of time
slots needed by all the units. Simulation traces reveal that the collisions persisted through the
entire simulation duration. In the case of collisions, the SINR of a received frame by all the
receiving units is given by Equation (1). If the SINR of a received MAC frame is less than
SINR
th
, then the information on newly reserved time slot is lost in the collided time slot and it
is considered to be free by other units. Hence, the set of free time slots built by each unit will
consist of spurious free time slots since all the actual free slots are occupied. Furthermore,
when the unit density is above the number of time slots in the superframe, the probability that
two or more units select the false time slot is reasonably high.
Next, the same sets of simulation as above were repeated with the shadowing effect of = 2
dB, 4 dB and 8 dB. These values were chosen based on UWB channel measurements for
indoor environment (Irahhauten et al., 2006). Figs. 11 and 12 show the successful SOC-MAC
packet reception rate and the normalized network throughput as a function of the number
of units. As shown in the figures, shadowing has detrimental effect on the performance. The
performance degrades as increases. For = 8 dB, the maximum achievable network
throughput is around 75% as compared to 90% for the case without shadowing. As
observed in Fig. 12, the shadowing effect is more pronounced when M = [35 .. 50] and [140 ..
200] for 4-timeslot and 1-timeslot, respectively. In this region of M, the channel capacity is
saturated. Therefore, SOC-MAC packet losses are due to both collisions and shadowing.
When M > 50 and 200 for 4-timeslot and 1-timeslot, respectively, the majority of the SOC-
MAC packet losses are due to collisions rather than shadowing. This is evidenced by the
convergence of the three plots.

139
0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
S
u
c
c
e
s
s
f
u
l

F
r
a
m
e

R
e
c
e
p
t
i
o
n

R
a
t
e
,

P

1-timeslot
4-timeslot
= 2 dB
MAX_TIMEOUT = 4
= 4 dB
= 8 dB

Fig. 11. Successful Packet Reception Rate versus Number of Units for MAX_TIMEOUT = 4,
and = 2 dB, 4 dB and 8 dB
0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
N
o
r
m
a
l
i
z
e
d

M
A
C

N
e
t
w
o
r
k

T
h
r
o
u
g
h
p
u
t
,

S

1-timeslot
4-timeslot = 2 dB
MAX_TIMEOUT = 4
= 4 dB
= 8 dB

Fig. 12. Normalized Average SOC-MAC Throughput versus Number of Units for
MAX_TIMEOUT = 4, and = 2 dB, 4 dB and 8 dB

140
4.1.2 Effects of timeout
In this subsection, we investigate the influence of the MAX_TIMEOUT parameter on the
performance. As mentioned in subsection 2.2, MAX_TIMEOUT is the maximum number of
superframes a unit may occupy a particular time slot. The same sets of simulation, as in
subsection 4.1.1, were repeated and the MAX_TIMEOUT value was varied from 2 to 10
superframes in a step of 2. Figs. 13 and 14 depict the plots for the successful MAC packet
reception rate and the network throughput. We can make two observations. Firstly, each of
the MAX_TIMEOUT values delivers the same performance at low unit density, which is for
M < 40 and 150 for 4-timeslot and 1-timeslot, respectively. Secondly, at high unit density,
larger MAX_TIMEOUT values give a slight performance advantage than smaller ones. The
performance of MAX_TIMEOUT = 2 is the worst, and MAX_TIMEOUT = 8 and 10 achieve
the best performance. For MAX_TIMEOUT = 2, it means that the frequency of time slot
renewal is the highest. Thus, at high unit density, a frequent renewal is not preferred
because all of the time slots are fully occupied. If there is an available time slot, two or more
units would select the same time slot, which results in collision as evidenced by the lowest
successful MAC frame reception rate in Fig. 13. Additionally, information on the newly
reserved time slot is lost in the collided time slot at high unit density. Other units would
have mistaken these time slots for vacant. At high unit density, the time-slot occupancy map
constructed by any unit would mainly consist of spurious time slots. As a result, selecting
any of these spurious time slots would prolong collisions.

0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
S
u
c
c
e
s
s
f
u
l

F
r
a
m
e

R
e
c
p
e
t
i
o
n

R
a
t
e
,

P

MAX_TIMEOUT=2 (1-timeslot)
1-timeslot
4-timeslot
= 0 dB

Fig. 13. Successful SOC-MAC Packet Reception Rate for = 0 dB, and MAX_TIMEOUT = [2,
4, 6, 8, 10]

141
0 50 100 150 200 250 300 350 400
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of Units, M
N
o
r
m
a
l
i
z
e
d

M
A
C

N
e
t
w
o
r
k

T
h
r
o
u
g
h
p
u
t
,

S
1-timeslot
4-timeslot
= 0 dB

Fig. 14. Normalized Average SOC-MAC Network Througput for = 0 dB, and
MAX_TIMEOUT = [2, 4, 6, 8, 10]
4.2 LAR performance
In this subsection, we investigate the performance of LAR. First, we analyze the
performance of LAR route establishment algorithm without the influence of route selection.
Thus, a chain topology was used. With the chain topology, a unit usually has two neighbors
except the unit at either end of the chain, which has only one. Each unit is stationary and
spaced at an equidistant of 50m, which is just below the maximum transmission range that
was determined using Equation (7) and the parameter values of Table 1. A single BU is
placed at one end of the chain. A chain topology, which comprises (h+1) units, consists of h
hops. In the chain topology, the BU is always assumed to be the 1st unit and the (h+1)th unit
is the last unit. The BU is responsible for initiating the route construction by broadcasting its
position packets. The rest of the units in the chain are either MUs or DUs. The composition
of MUs and DUs is irrelevant since they are functionally equivalent from the routing
protocol point of view. The route discovery and end-to-end packet delays were examined. Route
discovery delay is defined as the time in seconds when a unit (except the BU) found a route
(i.e., next-hop neighbor) on the forward path. End-to-end delay is defined as the time in
seconds taken by a data packet to traverse from an MU or a DU to the BU. Fig. 15 shows that
both the route discovery and end-to-end packet delays are linearly proportional to the
number of hops. The results prove that LAR became highly scalable. For a network diameter
of 50 hops, it took the last unit at the end of the chain less than 30 SOC-MAC superframes to

142
discover a route since the BU started broadcasting the routing information. End-to-end
delays are determined using position reporting packets which are sent by the last unit, i.e.
the (h+1)th unit, to the BU for an h-hop network where h varies from 1 to 50. Note that the
(h+1)th unit only starts transmitting the position reporting packets once a route is found to
remove the queuing effect due to route discovery.

0 10 20 30 40 50 60
0
5
10
15
20
25
30
35
40
45
Network Diameter (Number of hops)
N
u
m
b
e
r

o
f

S
u
p
e
r
f
r
a
m
e
s

Route Discovery Delay
End-to-End Packet Delay

Fig. 15. Route discovery and End-to-End Packet Delay
Next, we will analyze the performance of the LAR route selection algorithm. For the
analysis, we used a star topology as shown in Fig. 16. Each unit was stationary and spaced
at an equidistant of 50m from its adjacent neighbors. BU0, BU1 and BU2 initiated the route
construction simultaneously by broadcasting their position packets, and triggered
neighboring units to transmit their positioning packets. In the star configuration, the unit at
the center, MU0, received position packets from three different neighboring units, namely
DU1, DU2 and MU1, see Fig. 16. Consequently, MU0 created three forward routes in its
routing table. These routes are referred to as Route 1, Route 2 and Route 3, respectively, as
shown in Fig. 16. The hop count of Route 1 and Route 2 is three hops while Route 3 is four.
Since hop count is the primary routing metric, the routes with the least hop count would be
selected by MU0. In this case, Route 1 and Route 2 were picked by the route selection
algorithm of MU0. In the simulations, each unit broadcast position packets at a fixed
interval of 4s. Hence, the traffic load was uniformly distributed across the network. In other
words, none of the MUs or DUs were more congested than others. Therefore, the route

143
selection algorithm would arbitrarily choose between Route 1 and Route 2. The MU0 was set
to transmit position reporting packet at time t = 50s after the BUs started the route
construction. Simulation traces show that Route 2 was selected by MU0 for transporting its
position reporting packets to BU1. And the end-to-end packet delay is approximately 2
superframes, which conforms to the 3-hop delay in Fig. 15. At t=100s, MU2 was set to send
position reporting packets, which introduced extra traffic on to the network. MU2 used
Route 2 for transporting its position reporting packets since Route 2 was the shortest. Fig. 17
shows the congestion level seen by MU0.

BU0 DU0 DU1 MU0
BU1
MU1 MU2
DU2
DU4
DU3
BU2
Route 1 Route 2
R
o
u
t
e

3

Fig. 16. Star Network Topology
Fig. 17 depicts the position reporting packets received by BU0 and BU1. As shown in Fig. 17,
initially MU0 selected Route 2 for transporting its position reporting packets until the time
was approximately 110s, where it switched to Route 1. The switching occurred when MU0
detected the congestion level on Route 2 was increased to 3. The increase in congestion was
caused by MU2 when it started transmitting its position reporting packets at t=100s. Due to
congestion, some in-flight packets on Route 2 were experiencing excessive delays and
arrived at BU1 later than packets sent on Route 1. The congestion level of both Route 1 and
Route 2 continued to rise, and on Route 2, the congestion level reached the maximum at
about 150s. When both MU0 and MU2 stopped transmitting position reporting packets at
250s, the congestion level did not drop until t = 350s for Route 2 and t = 410s for Route 1
because of a large number of packets already in the queue. At t = 350s, the congestion level

144
of Route 2 dropped to 5, which was the same as Route 1. At this point a route change
occurred since MU0 selected Route 2 again. All the remaining packets in its queue were sent
on Route 2. After time t = 450s, the congestion level of both MU0 and MU2 dropped sharply.

0 100 200 300 400 500 600
0
1
2
3
4
5
6
7
Time (s)
C
o
n
g
e
s
t
i
o
n

L
e
v
e
l

Route 1
Route 2
Route 1
Selected
Route 2
Selected
Route 1
Selected
Route 2
Selected
Route Change
(A)
Route Change
(C)
Route Change
(B)

Fig. 17. Congestion Level
5. Related work
This section reviews the MAC and routing protocols developed for UWB-based ad hoc
sensor networks.
5.1 UWB-based MAC protocols for ad hoc sensor networks
In the past few years, a number of MAC protocols have been proposed for UWB-based
systems. (Legrand et al., 2003) and (Zhu & Fapojuwo, 2005) proposed a modified version of
the IEEE 802.15.3 Wireless Personal Area Network (WPAN) MAC protocol, which rely on a
centralized controller. These MAC protocols can provide guaranteed Quality of Service
(QoS) but are difficult to scale. The WHYLESS.COM project (Cuomo et al., 2002) proposed a
distributed UWB MAC, which supports QoS and is scalable but has high complexity. (Chu
& Ganz, 2004) described a hybrid MAC for WPAN, which combines the advantages of both
centralized and distributed protocols. The MAC protocol assumes that every node in a
WPAN is one hop away from every other node. Consequently, the MAC is foreseen to face

145
scalability issues when operating in multi-hop scenarios. Furthermore, a separate control
channel is used for signaling purposes, which increases the complexity and is not
lightweight for low bit-rate channels. Ultra-Wideband MAC (U-MAC) (Jurdak et al., 2005) is
a proactive and adaptive protocol. Similar to (Chu & Ganz, 2004), a separate signaling
channel is needed for exchanging a nodes state information with its direct neighbors.
(Broutis et al., 2007) and (Benedetto et al., 2005) outlined a multi-channel MAC in which
communication between two nodes takes place on orthogonal channels. The complexity and
overheads incurred by such a MAC protocol are higher than single-channel MAC protocols.
(Merz et al., 2005) proposed a combined Physical and MAC layer for very low power UWB
system. No separate control channel is needed. However, the signaling overheads incurred
by the MAC can be significant for short data packets and low bit-rate channels. In summary,
all of the above-mentioned MAC protocols were not designed for localization application in
mind. The IEEE 802.15.4a standard (Karapistoli et al., 2010; IEEE 802.15.4a, 2007) specifies a
Physical layer and a MAC layer which support localization. The IEEE 802.15.4a MAC
supports two different modes of channel access: beacon-enabled and nonbeacon-enabled.
The latter is suited for localization application. Unlike SOC-MAC, the nonbeacon-enabled
mode of the IEEE 802.15.4a MAC is based on the classical Aloha scheme or the CSMA/CA
scheme.

50 100 150 200 250 300 350 400 450 500 550
50
80
110
140
170
200
Time (s)
A
p
p
l
i
c
a
t
i
o
n

P
a
c
k
e
t

S
e
q
u
e
n
c
e

N
u
m
b
e
r

Received by BU0 (Route 1)
Received by BU1 (Route 2)
Route Change
(A)
Route Change
(B)
Delayed in-flight packets

Fig. 18. Position Reporting Packets

146
5.2 Routing protocols for ad hoc sensor networks
A large number of routing protocols, e.g. (Kulik et al., 2002; Intanagonwiwat et al., 2000;
Schurgers & Srivastava, 2001; Shah & Rabaey, 2002; Lindsey & Raghavendra, 2002;
Manjeshwar & Agarwal, 2001), have been developed for ad hoc sensor networks. Although
the considered ILS is an ad hoc sensor network, it has some profound distinctions which
mean existing ad hoc sensor routing protocols are not directly applicable. Firstly, sensor
nodes are generally assumed to have very low mobility after deployment (Al-Karaki &
Kamal, 2004) in comparison with ILS. Lastly, the relative size of ad hoc sensor networks is
huge in the order from thousands to millions of nodes (Al-Karaki & Kamal, 2004) as
compared to ILS.
6. Summary
In this chapter, we described the SOC-MAC and LAR protocols that are tailored for indoor
localization systems used to track emergency responders. The cross-layer approach is
present in the protocol design in order to optimize bandwidth and battery-energy
consumption. As a result, SOC-MAC is simple and self-organizing, which is composed of
two phases, namely RA-TDMA and reserved TDMA. The former is for initial acquisition of
time slots while the latter is for management and maintenance of time slots. In addition to
simplicity, LAR is extremely lightweight. No dedicated routing packets are needed. Instead,
routing information is carried in the network header of localization packets, which
constitutes less than 1% of the total channel capacity. We validated and studied the
performance of SOC-MAC and LAR by simulations under varying SOC-MAC and LAR
parameters.
7. Acknowledgement
The work was partially funded by the IST-004154 EUROPCOM project.
8. References
Al-Karaki, J. N. & Kamal, A. E. (2004). Routing Techniques in Wireless Sensor Networks: A
Survey, IEEE Wireless Communications Magazine, Vol. 11, No. 6
Benedetto, M.-G.; De Nardis, L.; Junk, M. & Giancola, G. (2005). (UWB)
2
: Uncoordinated,
Wireless, Baseborn Medium Access for UWB Communication Networks, Mobile
Networks and Applications (MONET), Vol. 10, No. 5
Broutis, I.; Krishnamurthy, S. V.; Faloutsos, M.; Molle, M. & Forester, J. R. (2007). Multiband
Media Access Control in Impulse-based UWB Ad Hoc Networks, IEEE Transactions
on Mobile Computing, Vol. 6, No. 4
Chu, Y. & Ganz, A. (2004). MAC Protocols for Multimedia Supporting UWB-based Wireless
Networks, Proceedings of 1st Intl Conference on Broadband Networks (BROADNETS)
Cuomo, F.; Martello, C.; Baiocchi, A. & Fabrizio, C. (2002). Radio Resource for Ad Hoc
Networking with UWB, IEEE Journal on Selected Areas in Communications, Vol. 20,
No. 9

147
Frazer, E. L. & Tee, D. (2004). A Comparison of UWB Technologies for Indoor Positioning as
an Augmentation to GNSS, Proceedings of 2nd European Space Agency (ESA)
Workshop on Satellite Navigation User Equipment Technologies (NAVITEC), Noordwijk,
The Netherlands, 2004
Harmer, D. (2008). EUROPCOM: Ultra-WideBand Radio for Rescue Services, Proceedings of
2nd Intl Workshop on Robotics for Risky Interventions and Surveillance of the
Environment (RISE), Benicassim, Spain, 2008
Harmer, D., et al. (2008). EUROPCOM: Emergency Ultra-WideBand (UWB) Radio for
Positioning and Communications, Proceedings of IEEE International Conference on
Ultra-WideBand (ICUWB), 2008
Hofmann-Wellenhof, B.; Lichtenegger, H. & Wasle, E. (2008). GNSS Global Navigation
Satellite Systems: GPS, GLONASS, and more, Springer, Vienna
IEEE 802.15.4a (2007). Wireless Medium Access Control (MAC) and Physical Layer (PHY)
Specifications for Low-Rate Wireless Personal Area Networks (WPANs)
Intanagonwiwat, C.; Govindan, R. & Estrin, D. (2000). Directed Diffusion: a Scalable and
Robust Communication Paradigm for Sensor Networks, Proceedings of ACM
MobiCom, Boston, MA, 2000
Irahhauten, Z.; Janssen, G. J. M., Nikookar, H., Yaravoy, A. & Lighart, L. P. (2006). UWB
Channel Measurements and Results for Office and Industrial Environments,
Proceedings of Intl Conference on Ultra-WideBand (ICUWB), MA, 2006
Jurdak, R.; Baldi, P. & Lopes, C. V. (2005). U-MAC: A Proactive and Adaptive UWB Medium
Access Control Protocol, Wiley Wireless Communications and Mobile Computing, Vol.
5, No. 5
Karapistoli, E.; Pavlidou, F.; Gragopoulos, I. & Tsetsinas, I. (2010). An Overview of the IEEE
802.15.4a Standard, IEEE Communications Magazine, Vol. 48, No. 1
Kulik, J.; Heinzelman, W. R. & Balakrishnan, H. (2002). Negotiation-based Protocols for
Disseminating Information in Wireless Sensor Networks, Wireless Networks, Vol. 8
Legrand, J.; Bucaille, I.; Hethuin, S.; De Nardis, L.; Giancola, G.; Di Benedetto, M.; Blazevic,
L. & Rouzet, P. (2003). U.C.A.N.s Ultra Wideband Medium Access Control
Schemes, Proceedings of Intl Workshop on Ultra Wideband Systems (IWUWBS), 2001
Lindsey, S. & Raghavendra, C. (2002). PEGASIS: Power-efficient Gathering in Sensor
Information Systems, Proceedings of Aerospace Conference, 2002
Manjeshwar, A. & Agarwal, D. P. (2001). TEEN: a Routing Protocol for Enhanced Efficiency
in Wireless Sensor Networks, 1st Intl Workshop on Parallel and Distributed Computer
Issues in Wireless Networks and Mobile Computing, 2001
Merz, R.; Widmer, J.; Le Boudec, J. Y. & Radunovic, B. (2005). A Joint PHY/MAC
Architecture for Low Radiated Power TH-UWB Wireless Ad Hoc Networks, Wiley
Wireless Communications and Mobile Computing, Vol. 5, No. 5
Mobility Framework, http://mobility-fw.sourceforge.net
OMNeT++, http://www.omnetpp.org/
Rappaport, T. (2001). Wireless Communications, 2nd edition, Prentice Hall
Schurgers, C. & Srivastava, (2001). Energy-efficient Routing in Wireless Sensor Networks,
MILCOM Proceedings on Communications for Network-Centric Operations: Creating the
Information Force, McLean, VA, 2001

148
Shah, R. C. & Rabaey, J. (2002). Energy Aware Routing for Low Energy Ad Hoc Sensor
Networks, Proceedings of WCNC, Orlando, FL, 2002
Zhu, J. & Fapojuwo, A. O. (2005). A Complementary Code-CDMA-based MAC Protocol for
UWB WPAN System, EURASIP Journal on Wireless Communications and Networking,
Vol. 2005, No. 2
8
Hybrid Access Techniques for Densely
Populated Wireless Local Area Networks
J. Alonso-Zrate
1
, C. Crespo
2
, Ch. Verikoukis
1
and L. Alonso
2

1
Centre Tecnolgic de Telecomunicacions de Catalunya (CTTC),Castelldefels, Barcelona
2
Universitat Politcnica de Catalunya (UPC), Castelldefels, Barcelona
Spain
1. Introduction
The IEEE 802.11p Task Group has recently released a new standard for wireless access in
vehicular environments (WAVE). It constitutes an amendment to the 802.11 for Wireless
Local Area Networks (WLANs) to meet the requirements of applications related to road-
safety involving inter- and intra-vehicle communications as well as communications from
vehicle to the roadside infrastructure. Indeed, the importance of the targeted applications
has forced authorities to allocate some dedicated bandwidth (nearby the 5.9GHz) to ensure
the security of the communications. However, despite the suitability of this standard for use
in high-speed vehicular communications, it is not possible to pass over the unprecedented
market penetration of the popular 802.11 networks, the so-called WiFi networks. Before we
can see a world where all the cars are equipped with 802.11p devices, current and near-
future applications might probably run on the original 802.11. Moreover, interaction
between humans and vehicles will probably be carried out by means of the 802.11, which is
the standard that is flooding most of personal tech devices, such as laptops, mobile phones,
gaming consoles, etc. Therefore, it is important to keep on working in the improvement of
the 802.11 Standard for its use in, at least, some vehicular applications.
This is the main motivation for this chapter, where we focus on the Medium Access Control
(MAC) protocol of the 802.11 Standard, and we propose a simple mechanism to improve its
performance in densely populated applications where it falls short to provide users with
good service. Envisioned applications include those were a high number of vehicles and
pedestrians coexist in a given area, such as for example, a crossing in a city where all the
cars share information to coordinate the drive along the crossing and prevent accidents.
Into more detail, the Distributed Coordination Function (DCF) is the mandatory access
method defined in the widely spread IEEE 802.11 Standard for WLANs [1]. This access
method is based on Carrier Sensing Multiple Access (CSMA), i.e., listen before transmit, in
combination with a Binary Exponential Backoff (BEB) mechanism. An optional Collision
Avoidance (CA) mechanism is also defined by which a handshake Request to Send (RTS)
Clear to Send (CTS) can be established between source and destination before the actual
transmission of data. This CA mechanism aims at reducing the impact of the collisions of
data packets and to combat the hidden terminal problem. The DCF can be executed in either
ad hoc or infrastructure-based networks and is the only access method implemented in most
commercial hardware. Despite the doubtless commercial success of the DCF, the simplicity

150
of a CSMA-based protocol comes at the cost of a trial-and-error approach where a
transmission attempt can result in a collision if several users contend for the access to a
common medium as the traffic load of the network increases. Therefore, those networks
based on the 802.11 suffer from really low performance when either the number of users or
the traffic load is high.
In this chapter, we introduce the idea of combining the DCF with the Point Coordination
Function (PCF), also defined in the 802.11 Standard, to overcome its limitations under heavy
load conditions. The PCF is defined as an optional polling-based access method for
infrastructure-based networks where there is no contention to get access to the channel and
the access point (AP) polls the stations of the network to transmit data. Therefore, collisions
of data packets can be completely avoided and the performance of the network can be
boosted.
The hybrid approach of combining distributed access with reservation or polling-based
access has been already used in the context of infrastructure-based networks [2]-[6]
combining static Time Division Multiplex Access (TDMA) with dynamic CSMA access.
Most of these works propose different alternatives to use the empty slots of TDMA in the
case that the user allocated to a given slot has no data to transmit. However, to the best
knowledge of the authors, there are very few works in the literature dealing with this
approach in a distributed manner, i.e., for ad hoc networks without infrastructure. This is
the main motivation for the work presented in this chapter, where we define the Distributed
Point Coordination Function (DPCF) as a hybrid combination of the distributed access of the
DCF and the poll-based access of the PCF to achieve high performance in highly populated
networks with heavy traffic load. Indeed, the work presented in this chapter has been
motivated by the successful results presented in [7]. In that paper, a spontaneous,
temporary, and dynamic clustering algorithm has been integrated with a high-performance
infrastructure-based MAC protocol, the Distributed Queuing Collision Avoidance (DQCA)
protocol, in order to extend its near-optimum performance to networks without infrastructure.
Upon the conclusion of that work, we realized that the same approach could be applied to the
IEEE 802.11 Standard access methods and thus be able to extend the high-performance of the
PCF under heavy load conditions to the distributed environments where the DCF runs.
We have observed that there are very few works dealing with the PCF, which can indeed
potentially achieve better performance than the DCF under heavy traffic conditions. Some
contributions related to the PCF improve the overall network performance through novel
scheduling algorithms [8]-[12] or by designing new polling mechanisms that can reduce the
overhead associated to the polling process [13]. However, there have been almost no efforts
in extending the operation of the PCF to ad hoc networks in order to provide them with
some degree of QoS. The only exception can be found in [14] where a virtual infrastructure
is created into a MAC protocol called Mobile Point Coordinator MAC (MPC-MAC) in order
to achieve QoS delivery and priority access for real time traffic in ad hoc networks
maintaining both the PCF and the DCF. In summary, a clustering based mechanism is used
to achieve the correct operation of the PCF in a distributed environment. The duration of the
PCF and DCF periods and the criterion upon which a terminal is chosen to be the MPC
(acting as AP) are fixed and they are determined by the MAC protocol configuration. This
approach works well in low dynamic environments where the topology does not vary
frequently. In this situation the overhead associated to the hello messages required for the
clustering mechanism can be kept to a minimum. However, it may not be convenient for
spontaneous and highly dynamic environments, such as those present in some vehicular
Hybrid Access Techniques for Densely Populated Wireless Local Area Networks

151
applications, where the clustering overhead could impact negatively on the efficiency of the
network. In addition, this protocol does not consider that the responsibility of becoming
cluster head should be shared among all the users of the network to ensure certain fairness
regarding the extra energy consumption associated to the role of coordinating a cluster.
Taking into account this background and motivated by the success of extending DQCA to
become DQMAN [7], we contribute to the field by presenting the DPCF as an extension of
the PCF to operate over infrastructure-less networks through smooth integration with the
DCF. By combining the DCF and the PCF using a spontaneous and dynamic clustering
mechanism at the MAC layer it is possible to extend the higher performance of the PCF to
networks without infrastructure. We present a description of the protocol as well as a
comprehensive performance evaluation based on computer simulation both for single-hop
and multi-hop networks.
The chapter is organized as follows. The DCF and the PCF of the IEEE 802.11 Standard are
overviewed in Section II. The DPCF protocol is then described in Section III. In Section IV,
we present a comprehensive performance evaluation of the protocol by means of computer
simulation. Finally, Section V concludes the chapter and outlines some future lines of
research.
2. IEEE 802.11 MAC protocol overview
An overview of the operation of the DCF and the PCF of the IEEE 802.11 Standard is
included in this section. A comprehensive description of them can be found in [1]. Following
the naming of the standard, we will refer herein to a vehicle or pedestrian equipped with a
communications terminal as a mobile station, or simply, a station.
2.1 DCF overview
The DCF is the mandatory coordination function implemented in all standard compliant
devices. Two access modes of operation are defined in the DCF:
1. Basic access (BASIC) mode; the station which seizes the channel transmits its data packets
without establishing any previous handshake with the intended destination.
2. Collision avoidance access (COLAV) mode; a handshake RTS/CTS is established between
source and destination before initiating the actual transmission of data. These RTS and
CTS get the form of special control packets. The COLAV access mode is aimed at
reducing the impact of collisions of data packets and at combating the presence of
hidden terminals.
Two examples are illustrated in Figure 1 and Figure 2 representing the operation of the
BASIC and the COLAV access modes, respectively. In summary, any station with data to
transmit listens to the channel for a DCF Inter Frame Space (DIFS). If the channel is sensed
idle for this DIFS period, the station seizes the channel and initiates the data transmission
(or the RTS transmission in the COLAV mode). Otherwise, if the channel is sensed busy, the
station backs off and executes a BEB algorithm by which the size of the contention window
is doubled up upon any transmission failure and reset to the minimum value upon success.
When a data packet is received without errors, the destination sends back an ACK packet
after a Short Inter Frame Space (SIFS). This SIFS is necessary to compensate propagation
delays and radio transceivers turn around times to switch from receiving to transmitting
mode. It is worth noting that since a SIFS is shorter than a DIFS, acknowledgments have
more priority than regular data traffic.

152
DATA
ACK
Source
Destination
Others CW
NAV
DIFS SIFS DIFS
Time+
CW

Fig. 1. Example: DCF Operation (Basic Access mode)

DATA
ACK
Source
Destination
Others
CW
DIFS
RTS
SIFS
CTS
SIFS
NAV RTS
NAV CTS
NAV DATA
DIFS
Time+
SIFS
CW

Fig. 2. Example: DCF Operation (Collision Avoidance mode)
A relevant feature of the DCF is the Virtual Carrier Sensing (VCS) mechanism. Stations not
involved in an ongoing transmission defer from attempting to transmit for the time that the
channel is expected to be used for an effective transmission between any pair of source and
destination stations regardless of the actual physical carrier sensing. To do so, stations
update the Network Allocation Vector (NAV) which accounts for the time the channel is
expected to be occupied. This information is retrieved from the duration field attached to
the overheard RTS, CTS, and data packets. This mechanism is mainly aimed at combating
the presence of hidden terminals.
2.2 PCF overview
The PCF can only run on infrastructure-based networks wherein an AP sequentially polls
stations to transmit data and thus collisions are totally avoided. This mechanism was
initially designed for the provision of QoS over WLANs.
When the PCF is executed, time is divided into Contention Free Periods (CFP), wherein the
AP sends poll messages to give transmission opportunities to the stations, and Contention
Periods (CP), where the DCF is executed. Since the PCF is an optional coordination function
and is not implemented in all standard-compliant devices, DCF periods are necessary to
ensure access to DCF-only stations. The interleaving of CFPs and CPs is illustrated in Figure
3. As also shown in this figure, a CFP is initiated and maintained by the AP, which
periodically transmits a beacon (B). The first beacon after a CP (DCF access) is transmitted
after a PCF Inter Frame Space (PIFS). The duration of a PIFS is shorter than a DIFS but
longer than a SIFS, providing thus the initialization of a CFP with less priority than the

153
transmission of control packets, but with higher priority than the transmission of data
packets. The periodically transmitted beacons contain information regarding the duration of
both the CFP and the CP and allow a new arrived station to associate to the AP during a
CFP. The CFP is finished whenever the AP transmits a CFP End (CE) control packet.

Contention Period
DCF Access
Contention Free Period
Polling Access
NAV
Access Point (AP)
802.11 Station
B CE
CFP
B
Time +
B B
PIFS
NAV

Fig. 3. IEEE 802.11 PCF Interleaves CFPs with CPs
During a CFP, the only station allowed to transmit data is the one being polled by the AP or
any destination station which receives a data packet and has to acknowledge (ACK) it, if
applicable, and can combine the ACK with data in a same packet. In PCF, some packets can
be combined together in order to reduce the number of MAC and PHY headers and thus
increase the efficiency of the communications. In any case, the access to the channel is
granted one SIFS after the reception of either the poll or the data packet, respectively. A
polled user can either transmit a data packet to the AP or to any other station in the
network, establishing a peer-to-peer link. If a polled station has no data to transmit, it
responds with a special type of control packet, referred to as NULL packet.

Access Point (AP)
Station 1
Station 2
B POLL 1+Data (1)
Contention Free Period (CFP)
SIFS
ACK+Data (AP)
ACK+POLL 2+Data (2)
ACK+Data (1)
ACK
CE
Time+

Fig. 4. Example: PCF Operation
An example of PCF operation is illustrated in Figure 4. In this example, the AP initiates a
CFP by transmitting a beacon (B). After a SIFS, it combines a poll packet with data to station
1. Upon the reception of this combined packet, station 1 acknowledges the data packet
received and responds to the poll by transmitting a data packet to the AP. Note that this is
also a combined packet. Then, the AP acknowledges the data packet received from station 1
and combines a poll packet with data to station 2. Upon the reception of the packet, station 2
acknowledges the packet to the AP and transmits data to station 1. Upon the reception of the
packet, station 1 acknowledges the received packet. The CFP is finished with the
transmission of a CE packet.
3. A new MAC protocol: DPCF
The Distributed PCF (DPCF) protocol is presented in this section as an adaptation and
extension of the PCF to operate on distributed infrastructureless wireless ad hoc networks.

154
As already mentioned before, the main idea is to use the DCF to create spontaneous and
temporary clusters wherein the PCF can be executed, having a station acting as the AP for
the life time of each cluster.
We consider a set of terminals equipped with WLAN cards forming a spontaneous ad hoc
network. Any station must be able to operate in three different modes regarding the
clustering mechanism: idle, master, and slave. Initially, all the stations operate in idle mode
but they must be able to change the mode of operation when necessary.
Idle stations with data to transmit get access to the channel using the regular DCF.
Whenever a station gets access to the channel, it transmits an RTS targeted to the intended
destination of the data packet. This packet initiates a clustering process. Upon the reception
of the RTS, the intended destination of the packet becomes master and responds to the RTS
with a beacon (B) followed by a poll targeted to the station which transmitted the RTS. A
cluster is established and a CFP is initiated inside this cluster. All the idle stations which
receive the beacon become slaves and get synchronized to the master at the packet level.
Cluster membership is spontaneous and soft-binding: there are no explicit association and
disassociation processes and a station belongs to a cluster as long as it can receive the
beacons broadcast by the master. As in the PCF, a cluster is broken when the master
transmits a CE packet. Upon the reception of this CE packet, all the slaves revert to idle
mode and execute a backoff in order to avoid a certain collision if more than one station has
data to transmit and initiates the DCF access period. Therefore, according to this operation,
the clustering algorithm of DPCF is spontaneous in the sense that the first idle station with
data to transmit initiates the clustering algorithm.

ACK +
POLL N
NULL
CE
RTS
B
Station 1
Cluster
PCF
POLL 1
Station 2
SIFS
DATA for station 2
Station N
Time +
Station 2 becomes Master, and
stations 1,3,,N become Slave All the stations revert to idle mode
All the stations are idle
CW
CW
CW

Fig. 5. Example: DPCF Operation
An example of operation is represented in Figure 5. In this example, station 1 has data to
transmit to station 2. Once the station 1 successfully seizes the channel executing the rules of
the DCF, it transmits an RTS to station 2. Upon the reception of the packet, station 2
becomes master and transmits a beacon. The first poll is then sent to station 1, which has a
data packet ready to transmit. Station 1 transmits the data packet to station 2. Then, station 2
acknowledges the reception of the packet and polls station N with a combined packet. Since
station N has no data packets to transmit, it sends a NULL packet. Finally, station 2
transmits the CE packet to indicate the end of the cluster phase. All the slave stations revert

155
to idle mode and execute a backoff to reduce the probability of collision if more than one
station has data to transmit.
Within a cluster, the master can poll the slaves following any arbitrary order. Regardless of
the specific polling policy, the master has to have some knowledge of the local
neighborhood in order to be able to carry out the polling mechanism. To do so, all the
stations overhear the ongoing packet transmissions in their vicinity in order to create a
neighbor table with an entry for each station in the local neighborhood. This table should be
updated along time. The specific scheduling of the polling mechanism is out of the scope of
the basic definition of DPCF. Only as an example, a round robin polling scheme can be
executed following the entries of the neighbor table. In any case, once a station is polled by
the master, it may transmit a data packet to any other slave (peer-to-peer communication
model) without routing all the data through the master. Therefore, the master only acts as
an indirect coordinator of the communications, but not necessarily as a concentrator of
traffic (as the AP does in a regular centralized network).
The duration of a cluster is variable and depends on the traffic load of the network. An
inactivity mechanism is considered to avoid the transmission of unnecessary polls when there
are no more data packets to be transmitted. This mechanism consists of the following: any
master maintains a counter that is incremented by one unit upon each NULL packet
received from a polled station with no data to transmit. This counter is reset to zero
whenever a station responds to a poll with the transmission of a data packet. If the counter
gets to a specified value (tunable), the cluster is broken and a CE packet is sent.
On the contrary, it may happen that under heavy traffic conditions once a station becomes
master it operates as such for the whole operation of the network due to the absence of idle
periods. This would be unfair in terms of sharing the responsibility of being master in the
network among all the stations. Therefore, it is necessary to upper-bound the maximum
time that a station can operate as master without interruption. This limit is especially
important in infrastructureless networks where fair energy consumption is a must. The
approach in DPCF is the following: any master has a Master Time Out (MTO) counter which
determines the maximum duration of a cluster. The value of the MTO corresponds to the
maximum number of beacons (MTO=N
beacons
) that a master can transmit without
interrupting the operation of its cluster. The MTO counter is decremented by one unit after
each beacon is transmitted. Whenever the MTO counter expires, a CE packet is transmitted
and the cluster is broken regardless of the traffic load or activity of the stations. Therefore,
the maximum time that a station can operate as master is denoted by T
MAX
and can be
computed as

MAX beacons polls polls
T N N MIFS MTON MIFS . = = (1)
N
polls
denotes the number of polls transmitted between beacons, which can also be tuned,
and MIFS is the Maximum Inter Frame Space whose duration corresponds to the maximum
time between two consecutive polls. The duration of a MIFS can be computed as the time
elapsed when:
1. The master station combines an ACK of a recently received data packet with a poll and
a data packet.
2. The station polled acknowledges the reception of the data packet from the master and
combines the ACK with data for a third station.
3. The third station transmits the ACK of the data packet received from the second station.

156
Master Station
Slave 1
Slave 2
ACK+POLL 1
+Data for station 1
SIFS
ACK
+Data for station 2
ACK
SIFS SIFS
MIFS
Time+
POLL 2

Fig. 6. Definition of MIFS

Poll user i
NULL
received?
Limit reached?
Transmit CE
Inactivity
counter++
MTO=0?
MTO counter --
Receive RTS from
user i
Transmit First Beacon
Idle
Revert to Idle mode
Set to Master mode
yes
Poll next user
no
no
no
yes
yes
Idle
Data to
transmit to
station i?
Transmit RTS to
station i
Beacon from
station i?
Binary Exponential
BackOff
Set to Slave mode
CE received?
PCF operation
yes
yes
yes
no
no
no
(b) Station triggering clustering
(a) Station becoming master
Reset Inactivity Counter

Fig. 7. DPCF Clustering Flowchart

157
The definition of a MIFS is illustrated in Figure 6. Note that it also corresponds to the
minimum period of time that a station has to listen to the channel before establishing a new
cluster in order to reduce the probability that another master is present.
In order to summarize the whole operation of DPCF, a general flowchart is shown in Figure
7. The left branch of the chart models the operation of any station becoming master when
requested by any other stations and the right branch of the chart represents the operation of
the station initiating the clustering algorithm when it has data to transmit.
4. Performance evaluation
In order to evaluate the performance of DPCF, we have implemented the protocol rules in a
custom-made C++ link-level simulator. The simulator works in an object-oriented basis and
the source code of each station runs in parallel. The implemented code could be directly
integrated in a wireless card to execute the protocol rules. The main motivations for
implementing the protocol in a custom-made C++ simulator rather than in any other well
known system simulation platform (such as ns-2, for example) are:
1. The faster execution of the simulations.
2. The possibility of isolating the MAC protocol performance from the rest of the network.
3. The possibility to implement the protocol in a hardware testbed.
The system parameters have been set according to the PHY layer of the IEEE 802.11g
Standard [1] and they are summarized in Table 1.

Parameter Value Parameter Value
Data Packet
Length
(MPDU)
1500 bytes
Constant Message
Length
1500 bytes
Data Tx. Rate 54 Mbps Control Tx. Rate 6 Mbps
MAC header 34 bytes PHY preamble 96 s
SIFS, PIFS, DIFS 10, 30, 50 s SlotTime () 10 s
RTS, BEACON,
CF_END and
POLL packets
20 bytes
CTS and ACK
packets
14 bytes
CW
min
16 CW
max
256
MTO 3 Polls per beacon 19
Table 1. System Parameters for Evaluation of DPCF
4.1 Single-hop networks
We first consider the case of a single-hop network composed of 20 stations, all of them
within the transmission range of each other. All the stations generate data packets of fixed-
length following a Poisson arrival distribution and they contribute equally (homogeneously)
to the total aggregate data traffic of the network. The destination of each packet is randomly
selected among all the stations of the network with equal probability. In order to focus on
the MAC layer, all the packets are assumed to be received without errors and thus the
results herein presented correspond to an upper-bound of the performance of the protocol.

158
It is also assumed that an ideal round robin scheduling is performed to poll all the stations
once a cluster is established. Three different networks have been studied (they all have been
implemented in the simulator):
1. DCF: a network wherein all the stations only execute the DCF with the collision
avoidance access method.
2. PCF: a network wherein an AP manages the access to the channel. However, stations
transmit directly to the intended destination without routing traffic through the AP. In
this network, we consider that the AP also has data to transmit as any other regular
station.
3. DPCF: a network wherein all the stations execute the proposed DPCF protocol.
According to the parameters presented in Table 1, the number of polls between beacons has
been set to 19 and it indicates that all the slaves within a cluster are polled exactly once by
the master between the transmission of two consecutive beacons. In addition, the setting
MTO=3 indicates that all the slaves are polled at most three times when a cluster is
established unless the inactivity mechanism is triggered by the master.

0
3
6
9
12
15
18
21
24
27
30
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
Total Offered Traffic Load (Mbps)
T
h
r
o
u
g
h
p
u
t

(
M
b
p
s
)
DPCF
PCF
DCF
25 %
250 %

Fig. 8. Throughput Comparison DPCF, PCF, and DCF in a Single-hop Network
The throughput of the three different networks is plotted in Figure 8 as a function of the
total aggregate offered load to the network. As expected, the three curves grow linearly until
they reach the saturation throughput. The three protocols are stable for heavy traffic
conditions without entering in congestion and thus they can operate under sporadic
situations of peak high traffic loads without collapsing the network. The saturation
throughput of DPCF is remarkably higher than that of DCF, achieving an improvement of
approximately 250%. Collisions and backoff periods are reduced in the DPCF network
compared to the DCF network, thus yielding higher performance. In addition, the
performance of DPCF is even superior to the regular PCF, attaining 25% higher saturation
throughput.

159
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
P
r
o
b
a
b
i
l
i
t
y

o
f

t
r
a
n
s
m
i
t
t
i
n
g

w
h
e
n

b
e
i
n
g

p
o
l
l
e
d
DPCF
PCF
45 %

Fig. 9. Probability of Transmitting when Being Polled in a Single-hop Network
In order to further analyze this apparently counter-intuitive result, Figure 9 shows the
probability that a station transmits a data packet when it is polled. It has been considered for
this calculation that the AP (in the PCF network) and the masters (in the DPCF network) are
virtually polled every time they poll a station as they have the possibility to combine the
polls with data and ACK packets. The probability of transmitting data when being polled is
quite similar in the two networks for low traffic loads. However, this probability is much
higher in the DPCF network than in the PCF network for high traffic loads. While the
efficiency of the polling in DPCF gets close to 98% for high traffic loads, it remains close to
55% in the PCF network. This efficiency translates directly into a higher efficiency of DPCF,
since the ratio of data packets transmitted per control overhead is higher. The reason for
these figures is that there is a severe unbalance between the channel access opportunities
between the AP and the regular stations in the PCF network. This can be seen in Figure 10,
where we plot again the probability of actually transmitting when being polled. Now, two
different curves for the PCF network are represented corresponding to the average
probability among of all the regular stations and to the probability for the AP alone,
separately. The AP has a channel access opportunity every time it polls another station, but
most of these transmission opportunities are not used for the actual transmission of data
(note that the probability of transmitting when being polled is below 10% in all cases for the
AP), decreasing the overall efficiency of the polling mechanism.
This unbalance between the AP and the stations is avoided in DPCF by sharing the
responsibility of being master among all the stations of the network. It is well known that
the DCF is fair in the long-term, and so is the clustering algorithm of DPCF. Since all the
stations of the network get the role of master periodically, the unbalanced access of the AP
in the PCF network is shared in the DPCF network. Every time a station is set to master it
can transmit all its backlogged data packets and thus take advantage of the prioritized
access to empty its data buffers while operating as master. Indeed, the fact that a station

160
operating in master mode has more channel access opportunities than a slave station can be
seen as an implicit mechanism to provide with some incentive to stations to become master
despite the extra actions they must carry out and the corresponding increase in energy
consumption.

0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
1,1
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
P
r
o
b
a
b
i
l
i
t
y

o
f

t
r
a
n
s
m
i
t
t
i
n
g

w
h
e
n

b
e
i
n
g

p
o
l
l
e
d Stations in PCF network
Stations in DPCF network
AP in PCF network

Fig. 10. Probability of Transmitting when Being Polled in a Single-hop Network

0
20
40
60
80
100
120
140
160
180
200
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
A
v
e
r
a
g
e

P
a
c
k
e
t

T
r
a
n
s
m
i
s
s
i
o
n

D
e
l
a
y

(
m
s
)
DCF
PCF
DPCF

Fig. 11. Average Packet Transmission Delay in a Single-hop Network

161
The performance in terms of average packet transmission delay is plotted in Figure 11. We
define this delay as the average time elapsed since a packet arrives at the MAC layer until it
is successfully acknowledged by the intended destination. It is worth seeing that for low
offered loads, the best performance is attained in the DCF network. This is an expected
result since every time a station has data to transmit it can successfully seize the channel
immediately without needing to wait for being polled (the probability of finding the channel
busy and the probability of collision are low due to the low offered traffic load). However,
as the offered load grows, the average packet transmission delay in the DCF network grows
sharply for traffic loads over 10 Mbps. On the other hand, the DPCF attains average delays
below 200 ms for traffic loads up to 22 Mbps, increasing the throughput of the standard DCF
network and attaining superior performance than the PCF. These results confirm the idea
that PCF-like mechanisms are worthy when the traffic load and the number of transmitting
stations are relatively high.
4.2 Multi-hop networks
We now consider a multi-hop network. Without loss of generality and as a representative
example, we consider a tandem network formed by 5 static stations set in line and equally
spaced as the one represented in Figure 12. The distance between the stations, the
transmission powers, and the channel propagation parameters have been adjusted so that:
1. Every station can transmit directly to immediate neighbors at one-hop distance.
2. Every station at two hops of a transmitting station can sense the channel busy, but
cannot decode the transmitted information.
3. Every station at three hops of a transmitting station is oblivious to the transmission.

IDLE
Transmission Range
Interference Range
SLAVE
MASTER IDLE
IDLE

Fig. 12. Tandem Multi-hop Network
A collision occurs if two simultaneous transmissions are received within either the
transmission or the interference range of the transmitters. We assume that all the stations
have perfect routing information and thus route the packets through the station in its
transmission range that is closer to the intended destination. The rest of the parameters have
been set as in the previous section for the single-hop evaluation.

162

0
1
2
3
4
5
6
7
8
9
10
1 2 3 4 5 6 7 8 9 10 11 12
T
h
r
o
u
g
h
p
u
t

t
o

d
e
s
t
i
n
a
t
i
o
n

(
M
b
p
s
)
DPCF
DCF
80%

Fig. 13. Throughput to Destination of DPCF in a Multi-hop Network

0
30
60
90
120
150
180
210
240
270
300
1 2 3 4 5 6 7 8 9 10 11 12
A
v
e
r
a
g
e

P
a
c
k
e
t

T
r
a
n
s
m
i
s
s
i
o
n

D
e
l
a
y

(
m
s
)
DPCF
DCF

Fig. 14. Average Packet Transmission Delay of DPCF in a Multi-hop Network
The total throughput delivered to destination is plotted in Figure 13 as a function of the total
offered load to the network. Note that the traffic delivered to the intermediate stations in a
multi-hop route is not accounted for this calculation. The curves show that DPCF
outperforms DCF for all traffic loads. Indeed, for low traffic loads both protocols behave

163
almost identically delivering all the data traffic offered to the network. However, while DCF
saturates around 5 Mbps, DPCF is capable of delivering up to 9 Mbps (80% higher
saturation throughput), almost doubling up the capacity of the legacy DCF. Comparing
these results to the ones obtained for the single-hop case, it is possible to see that the total
offered load that can be conveyed in the multi-hop network is considerably lower. This is
mainly due to the fact that in the multi-hop environment some packets need to travel along
several hops to get to the final destination.
The average packet transmission delay is plotted in Figure 14 for both the DPCF and the
DCF networks. In this case, this measure is defined as the average time elapsed from the
moment a packet arrives at the MAC layer of the source station until it is successfully
delivered to the final destination (end-to-end time). The curves show that the DCF attains
lower average packet transmission delay for low traffic loads. Two are the main reasons for
this lower average delay. First, the longer MIFS of DPCF (compared to the DIFS of DCF)
adds latency to all the transmissions, increasing the average packet transmission delay in the
DPCF network for low traffic loads. In addition, in the DPCF network, slaves cannot
transmit immediately whenever they have data to transmit but they have to wait to be
polled by a master, increasing thus the average access delay. However, note that the average
delay is lower than 300 ms for loads up to 8 Mbps in the DPCF network and it gets
unbounded in the DCF network for traffic loads over 6 Mbps. Therefore, the DPCF protocol
attains better performance when the traffic load of the network is higher, attaining up to
25% better performance than the DCF in this multi-hop setting.
5. Conclusions
We have presented in this chapter a simple mechanism to improve the performance of the
802.11 Standard under heavy loaded conditions. These conditions appear in some vehicular
scenarios, such as in traffic-light crossings, where vehicles and pedestrians meet together
and a number of safety applications may arise.
The key idea consists in combining both distributed and point coordinated access methods
to manage the access of the users to the wireless channel. The specific approach has been
based on an extension of the PCF of the IEEE 802.11 Standard to operate over distributed
wireless ad hoc networks without infrastructure. The main idea of DPCF is that the stations
of the network get access to the channel by executing the rules of the DCF. Any station
which seizes the channel transmits its data and also establishes a temporary dynamic cluster
to manage the pending transmissions of all the neighbors with data ready to be transmitted.
The key of this mechanism is that there is no cluster head selection, but clusters are created
in a spontaneous manner. This reduces the control overhead to establish a fixed clustering
architecture and increases the capability of the network to dynamically adapt to the
unpredictable nature of ad hoc networks. Comprehensive performance evaluation of the
protocol through link-level computer simulation shows that the new proposal improves the
performance of ad hoc networks when compared to current standards.
The results presented in this chapter are rather promising and, in fact, future work will be
aimed at theoretically evaluating and optimizing the design of DPCF and at implementing
the protocol in a testbed to evaluate its actual performance in a real environment. Ongoing
work is being carried out to evaluate the coexistence feasibility of this new approach with
legacy implemented networks based on the 802.11.

164
6. Acknowledgments
The research leading to these results has received funding from the research projects
NEWCOM++ (ICT-216715), CO2GREEN (TEC2010-20823), CENTENO (TEC2008-06817-C02-
02), and GREENET (PITNGA2010264759).
7. References
[1] IEEE, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY)
Specifications, IEEE Std. 802.11 2007.
[2] D. J. Goodman, R. A. Valenzuela, K. T. Gayliard, and B. Ramamurthi, Packet Reservation
Multiple Access for Local Wireless Networks, IEEE Trans. on Communications, vol.
37, no. 8, Aug. 1989.
[3] I. Chlamtac, A. Farag, A. D. Myers, V. R. Syrotiuk, and G. V. Zruba, ADAPT: A
dynamically self-adjusting media access control protocol for ad hoc networks, in
Proc. of the GLOBECOM, pp. 11 15, Dec. 1999.
[4] A. Rhee, M. Warrier, and J. Min, ZMAC: A hybrid MAC for wireless sensor networks, in
Proc. of Sensys 2005, San Diego, California.
[5] M. Shakir, I. Ahmed, P. Mugen, and W. Wang, Cluster Organization based Design of
Hybrid MAC Protocol in Wireless Sensor Networks, in Proc. of the Third
International Conference on Networking and Services, pp. 78 83, Jun. 2007.
[6] A. Muir and J. J. Garcia-Luna-Aceves, An efficient packet sensing MAC protocol for
wireless networks, Springer Mobile Networks and Applications, vol.3, no. 2, pp. 221
234, Aug. 1998.
[7] J. Alonso-Zrate, E. Kartsakli, L. Alonso, and Ch. Verikoukis, Performance Analysis of a
Cluster-Based MAC Protocol for Wireless Ad Hoc Networks, EURASIP Journal on
Wireless Communications and Networking, Special Issue on Theoretical and
Algorithmic Foundations of Wireless Ad Hoc and Sensor Networks, vol. 2010,
Article ID 625619, 16 pages, March 2010.
[8] A. Kanjanavapastit and B. Landfeldt, A performance investigation of the modified PCF
under hidden station problem, in Proc. of the ICCCAS 2004, vol. 1, pp.428 432, Jun.
2004.
[9] B. Anjum, S. Mushtaq, A. Hussain, Multiple Poll Scheme for Improved QoS Using IEEE
802.11 PCF, in Proc. of the IEEE INMIC05, pp.1 6, Dec. 2005.
[10] D. Ping, J. Holliday, A. Celik, Dynamic scheduling of PCF traffic in an unstable wireless
LAN, in proc. of the CCNC. 2005, pp. 445 450, Jan 2005.
[11] K. Byung-Seo, K. Sung Won, W. Yuguang Fang, Two-step multipolling MAC protocol
for wireless LANs, IEEE Journal on Selected Areas in Communications, vol. 23, no. 6,
pp. 1276 1286, Jun. 2005.
[12] K. Young-Jae and S. Young-Joo, Adaptive polling MAC schemes for IEEE 802.11
wireless LANs, in Proc. of the VTC 2003, vol. 4, pp. 2528 2532, Apr. 2003.
[13] A. Kanjanavapastit and B. Landfeldt An analysis of a modified point coordination
function in IEEE 802.11, in Proc. of the IEEE PIMRC03, vol. 2, pp. 1732 1736, 2003.
[14] Y. Tiantong, H. Hassanein, H. T. Mouftah, Infrastructure-based MAC in wireless mobile
ad-hoc networks, in Proc. of the 27th Annual IEEE Conference on Local Computer
Networks, pp. 821 830, 2002.
0
Hybrid Cooperation Techniques
Emilio Calvanese Strinati and Luc Maret
CEA, LETI, MINATEC
France
1. Introduction
A major challenge in the design of next generation wireless communication systems is to
achieve both reliable and spectral efcient communication with large coverage range. To
tackle this problem, advanced diversity techniques combined with adaptive mechanisms have
to be designed in order to combat or even exploit the variability of the radio propagation
medium across time, frequency and space. Diversity techniques create signal redundancy,
by repeating the information across multiple, independent channel realizations. This is
accomplished by allowing the receiver to experience the average channel effect rather than
an instantaneous fade. As a consequence diversity techniques improve the link reliability at
the expense of the system spectral efciency. By adjusting the transmission parameters to the
momentary link quality, adaptive mechanisms aim at improving both spectral efciency and
link reliability. Nevertheless, in order to guarantee the Quality of Service (QoS) constraints
from the upper layers, adaptive mechanisms implement a sub-optimal trade-off between link
robustness and bandwidth efciency (Calvanese Strinati E., 2006). Therefore in this chapter
we propose and analyze a novel cooperation protocol, the hybrid cooperation protocol and
we combine it with link adaptation techniques such as Adaptive Modulation and Coding
(AMC) and power allocation. Our task is to minimize the outage probability and maximize
the spectral efciency of transmission, while limiting the cooperation cost in terms of MAC
signalling overhead.
The scientic content of this chapter is based on some innovative results presented in three
conference papers (E. Calvanese Strinati and S. Yang and J-C. Belore, 2007) (E. Calvanese
Strinati and Luc Maret, 2008) (M. Baydar and E. Calvanese Strinati and J. C. Belore, 2008)
presented in 2007 and 2008.
The goals of this chapter are for the reader to have an understanding of cooperative
communication issues and challenges and, to be well informed of the state-of-the-art research
development. Eventually, the chapter will present what we have done to improve the
performance of currently proposed cooperation techniques, comparing performance of our
proposed approaches with state-of-the-art one. A critical discussion on advantages and
weakness of the proposed approaches, including future research axes, will conclude the
chapter.
The innovative contribution in this chapter is threefold.
First, in this chapter we introduce and details challenges and possible solutions for the
so-called cooperative diversity (E. Erkip A. Sendonaris and B. Aazhang: Part I, 2003; E. Erkip
A. Sendonaris and B. Aazhang: Part II, 2003) techniques where a source terminal cooperates
with several relays to exploited the spatial diversity in a distributed manner. From a physical
9
layer viewpoint, cooperation drives to improved transmission diversity and consequent
improved outage probability performance. Nevertheless, from a MAC layer viewpoint, xed
cooperation requires to probe the network and acquire channel state information (CSI) about
all active relays at least with a channel coherence time frequency. This cooperation probing
makes xed cooperation expensive in terms of signalling overhead, battery consumptions of
active relays and protocol delay. Recently, researchers showed that cooperating is not always
the best solution in terms of outage probability minimization (D. Gunduz and E. Erkip,
2005). For instance in AF cooperation, when the noise is large, cooperative relays can amplify
the noise instead of helping. Alternatively to xed cooperation, we propose to introduce a
cooperation controller that can decide when and how to cooperate. The basic idea is to cooperate
when it is advantageous (cooperative mode), and not cooperate otherwise (non-cooperative mode).
The problem in such approach is how the source-destination pair can decide if it is worth to
cooperate for a given channel instance? In xed topology networks an heuristic approach is
to analyze the geometry of the network and determine areas where cooperation can help. In a
wireless mobile communication scenario the cooperation protocol should use the momentary
channel state information to make its decision. This feedback information introduces a large
processing delay and signalling overhead and it is impractical for the destination to acquire
full CSI about all active relays. In this chapter we present the innovative approach proposed
in (E. Calvanese Strinati and S. Yang and J-C. Belore, 2007) of introducing a cooperation
controller which makes its decision on non-cooperative/cooperative mode based only on
the momentary direct link quality information. This information is directly available at the
cooperation controller each time the source sends a request to send (RTS). More precisely, for
a selected transmission rate R and direct link channel instance (
2
n
, f , etc.), the cooperation
controller can check if direct non-cooperative transmission will be certainly in outage. If an
outage is forecasted, the receiver can switch to cooperative mode trying to avoid transmission
outage improving the overall link quality with cooperative diversity. This protocol is called
hybrid cooperation.
Second, the chapter presents how hybrid cooperation protocol and AMC mechanism can be
jointly designed. Eventually, we detail the hybrid cooperative AMC mechanism (E. Calvanese
Strinati and S. Yang and J-C. Belore, 2007) in which adaptation is designed for the hybrid link
quality and cooperation is activated only when the instantaneous direct source-destination
link quality is not good enough to support the aimed spectral efciency.
Third, we face the problem of optimal power allocation between source and relay in a
cooperative network. We present the interesting results proposed in (M. Baydar and E.
Calvanese Strinati and J. C. Belore, 2008). The authors succeed in nding a close form
power allocation algorithm which can only be applied to OAF cooperative transmission.
At a rst glance, this solution can be not of interest due to the worse performance of the
OAF cooperation strategy. We verify that classical NAF outperforms classical OAF also if an
optimal power allocation is done for the OAF cooperation and sub-optimal power allocation
is done for NAF. This is due to the suboptimal performance of classical OAF. To solve this
problem we further investigated the hybrid AF protocol that we propose.
2. Preliminaries on Cooperative transmission techniques
The core topic investigated in this chapter is the improvement of outage probability
performance in a cooperation network. In the literature (see, e.g., (H. Blcskei and R. U.
Nabar and F. W. Kneubhler, 2004; H. El Gamal K. Azarian and P. Schniter, 2005; S. Yang
and J-C. Belore, 2006)), three main cooperative transmission protocols have been proposed:
the amplify-and-forward (AF), the decode-and-forward (DF) ((D. N. Tse J. N. Laneman and
G. W. Wornell, 2004) (H. Blcskei and R. U. Nabar and F. W. Kneubhler, 2004) (H. El Gamal
K. Azarian and P. Schniter, 2005)) and the compress-and-forward (CF) for which there has
been, recently, a grown interest. Nevertheless, most prior works focused on two principal
classes of protocols. The rst is the class of AF protocols, where the relay simply amplies
and re-transmits the observed signal. The second is the class of DF protocols, where the relay
decodes, re-encodes and re-transmits the message it receives. The DF protocols offer good
performance but have clearly a higher complexity compared to the AF protocols which are
used in practice, due to their low complexity and low relay power consumption. Actually,
for most ad hoc wireless networks, it is not realistic for other terminals to decode the signal
from a certain user, because the codebook is seldom available and the decoding complexity is
unacceptable in most cases.
The second topic treated in this chapter is the design of hybrid cooperation protocol combined
with an AMC mechanism. Design of cooperation protocol and AMC algorithms have
been extensively investigated separately. Nevertheless, joint design of advanced cooperation
protocols with AMC algorithms has not been intensively investigated yet. Aiming at
maximizing the physical layer throughput, in (Z. Lin and E. Erkip and M. Ghosh, 2005) the
authors study adaptive modulation performance for one relay coded cooperative protocols.
In the paper the authors nd that coded cooperation combined with adaptive modulation
offers better physical layer throughput performance than non-cooperative mode. The authors
suggest that cooperative mode MCS selection should be decided based on all direct and
relays link quality. However, if the cooperative protocol includes more than one relay, this
approach can be complex and catastrophic adaptation can occur as for frequency selective
block fading channels (M. Lampe and H. Rohling and W. Zirwas, 2002) since it is hard to
obtain a reliable predicted packet error rate (PER
pred
). In (E. Yazdian and M. R. Pakravan,
2006) the application of adaptive modulation to one relay AF cooperation is investigated.
The authors aim at evaluating the energy saving achieved through cooperation due to
the improvement in average bit/symbol transmission. Furthermore, the authors study the
performance improvement as a function of cooperating users location to identify areas
where cooperation is useful. However, in the paper the possible occurrence of detrimental
cooperation is not considered.
The third topic investigated in this chapter is the combination of cooperative diversity
techniques with power control algorithms. Optimal power allocation between source and
relay in a cooperative network has been studied in (M. Hasna and M-S. Alouini, 2004)
(Q. Zhang and C. Shao and Y. Wang and P. Zhang and J. Zhang and Z. Zhang, 2004).
A total amount of transmit power over the two slots required for relaying is shared
between the source and relay. In (I. Hammerstrom and A. Wittneben, 2006), an iterative
joint power allocation method is presented for two-hop communications schemes using
OFDM modulation. This method is based on the Karush-Kuhn-Tucker (KKT) conditions.
Power allocation optimization for NAF cooperative transmissions is classically done using
waterlling techniques. Its effectiveness depends on the a priori choice of the power allocated
to the relay (P
r
). Unfortunately, the optimal selection of P
r
can be a challenging task. An
iterative search may improve the power allocation algorithm performance at the expense of
both search latency and algorithm complexity. Alternatively, the power allocation problem
can be faced for OAF schemes for which the a priori knowledge of P
r
is not required. In such
case, the complexity of the power allocation algorithm is strongly reduced at the expense of
performance.
167 Hybrid Cooperation Techniques
s
d
r
h
f
g
Fig. 1. A relay channel with one source s, one destination d and one active relay r.
d s
r
i
r
j
f
h
i
i, j
g
j
h
k
g
k
r
k
Fig. 2. A relay channel with one source s, one destination d and N active relays r
i
.
3. System model
The considered system model consists of one source s, one destination d and N
relays (cooperative terminals) r
1
, . . . , r
N
. The physical links between terminals are slowly
faded and are modelled as independent quasi-static Rayleigh channels, i.e., the channel gains
do not change during the transmission of a cooperation frame. This assumption implies that
we assume the channel coherence time to be much larger than the maximum delay that can
be tolerated by the application. All the terminals (source, relay and destination) are equipped
with only one antenna and work in half duplex mode, i.e., they cannot receive and transmit
at the same time. Two simple illustrations of the channel model are given in Fig. 1 and
Fig. 2 respectively for a cooperative network with only one and N active relays per frame
transmission.
The gain of the channel connecting s and d is denoted by f . Similarly, g
i
and h
i
respectively
denote the channel gains between r
i
and d and the ones between s and r
i
.
ij
is used to denote
the channel gain between r
i
and r
j
. Channel quality between terminals is parameterized by
the variance of the channel gains. We assume that the receiver can gain perfect knowledge
of the channel gains for the whole network activating the relay probing procedure (S. Yang
and J-C. Belore, 2006). We consider two cases for the power allocated to source and relays.
First, we impose a total average transmit power constraint and no power control is allowed
in our scheme. In this case, in order to simplify the analysis, we consider a suboptimal
power allocation scheme where the source transmits at full power in the non-cooperation
mode and both the source and the relays transmit at half power in the cooperation mode.
Then, we rene this assumption proposing a power allocation algorithm for hybrid OAF
cooperative protocols with only one active relay per transmission frame (see section 4.3).
Also, we suppose that the terminals are subject to the half-duplex constraint, i.e., they cannot
transmit and receive simultaneously. We also assume using the capacity achieving code so that
the outage analysis holds. The PER prediction is based on the computation of a link quality
metric (LQM) that is linked to the predicted PER by means of a look up table (LUT). Ideally,
we consider perfect PER prediction. In our work we consider the Amplify-and-Forward
(AF) protocol (orthogonal and non-orthogonal) where the relay simply scales and forwards
the received signal. We study half-duplex slotted amplify-and-forward (SAF) cooperative
schemes proposed in (S. Yang and J-C. Belore, 2006). For an N-relay M-slot scheme, the
cooperation frame, composed of M slots of l symbols, is of length Ml. During any slot i,
i = 1, . . . , M, the source s transmits a sub-frame of l symbols, denoted by a vector x
i
C
l
and the relay r
j
, j = 1, . . . , N, can transmit x
r
j
,i
C
l
, a linear combination of the vectors it
received in previous slots. Under the half-duplex constraint, a relay does not receive while
transmitting. For example, the NAF scheme (H. El Gamal K. Azarian and P. Schniter, 2005) is
an N-relay (2N)-slot scheme and the non-orthogonal relay selection scheme (D. P. Reed and
A. Bletsas and A. Khisti and A. Lippman, 2005) is an N-relay two-slot scheme. Obviously, the
transmission of a cooperation frame with any SAF scheme is equivalent to l channel uses of
the following vector (MIMO) channel
y =
SNRHx +z
where x is the transmitted signal, z CN[
z
] is the equivalent additive coloured noise with
covariance matrix
z
and H is an M M lower-triangular matrix representing the equivalent
space-time channel between the source and the destination. Moreover, we have H
ii
= c
i
f
with c
i
being a constant related to the transmission power. Let H denotes the equivalent
channel matrix (S. Yang and J-C. Belore, 2006) for a Non-Orthogonal AF scheme
1
H =
_
f 0
P
r
bgh
1+P
r
bg
2
f
1+P
r
bg
2
_
.
The matrix coefcients, h
i,j
are functions of f, g, h, P
r
the relay transmission power and the
normalization factor b which veries b
2
=
1
1+P
s
h
2
. The input covariance matrix is a diagonal
matrix, denoted Q and whose diagonal elements are P
s1
and P
s2
, the source transmission
powers in the rst and the second slot, respectively.
4. Improving Cooperative transmission protocols effectiveness
This section will be divided in three parts. First we will present and explain the Hybrid Amplify
and Forward Cooperation Protocol proposed in (E. Calvanese Strinati and S. Yang and J-C.
Belore, 2007) which has been designed to overcome the suboptimal error rate performance
of AF cooperative schemes in the low SNR region. Second, we will describe the Adaptive
1
An Orthogonal AF scheme is a particular case of the NAF scheme in which the source does not transmit
simultaneously with the relay in the second slot (i.e.,h
2,2
= 0)
Modulation and Coding Combined with the Hybrid Cooperation protocol proposed in (E.
Calvanese Strinati and S. Yang and J-C. Belore, 2007; E. Calvanese Strinati and Luc Maret,
2008). In a third part of the section, we will present the Power Allocation Optimization for
Hybrid Cooperation Protocols which has been proposed in (M. Baydar and E. Calvanese
Strinati and J. C. Belore, 2008).
4.1 Improving amplify and forward Cooperation protocol: Hybrid amplify and forward
We present in this section the novel cooperative protocol proposed in (E. Calvanese Strinati
and S. Yang and J-C. Belore, 2007). The protocol is named hybrid cooperation and it has been
designed to overcome the suboptimal error rate performance of AF cooperative schemes in
the low SNR region. The protocol proposal is based on the observation that cooperating is
not always the best solution in terms of outage probability minimization (D. Gunduz and
E. Erkip, 2005). For instance in AF cooperation, in the low SNR regime, the relays amplify
the noise instead of helping. Alternatively to xed cooperation, a cooperation controller can
decide when and how to cooperate. The principle of the hybrid cooperation protocol is
simple: based on the direct source-destination link quality, a cooperation controller decides
if and how to run cooperation. Indeed, cooperation is activated only when the instantaneous
direct source-destination link quality is not good enough to support the aimed spectral
efciency. The problem in such approach is how the source-destination pair can decide
if it is worth to cooperate for a given channel instance? In xed topology networks an
heuristic approach is to analyze the geometry of the network and determine areas where
cooperation can help. In a wireless mobile communication scenario the cooperation protocol
should use the momentary channel state information to make its decision. This feedback
information introduces a large processing delay and signalling overhead and it is impractical
for the destination to acquire full CSI about all active relays. In (E. Calvanese Strinati and
S. Yang and J-C. Belore, 2007) the authors propose that the cooperation controller makes
its decision on non-cooperative/cooperative mode based only on the momentary direct link
quality information. This information is directly available at the cooperation controller each
time the source sends a request to send (RTS). More precisely, for a selected transmission
rate R and direct link channel instance (
2
n
, f , etc.), the cooperation controller can check if
direct non-cooperative transmission will be certainly in outage. If an outage is forecasted, the
receiver can switch to cooperative mode trying to avoid transmission outage improving the
overall link quality with cooperative diversity.
Classical NAF outperforms classical OAF also if an optimal power allocation is done for
the OAF cooperation and sub-optimal power allocation is done for NAF. This is due to the
suboptimal performance of classical OAF. To solve this problem we further investigated the
hybrid AF protocol. Calvanese Strinati et al. rst propose an OAF hybrid cooperation protocol
under the same power constraint adopted above: impose a total average power constraint
and no power allocation is considered. If P denotes the total power constraint, in case of NAF
cooperation, we impose P
s1
= P/2 for the power allocated to the source in the rst slot and
P
s2
= P
r
= P/2 the power allocated to the source and the relay respectively in the second slot.
In the OAF scheme, the authors propose to x P
s
= P
r
= P/2. The mutual information of the
direct channel, the cooperative channel
2
and the OAF channel are respectively:
2
Factor
1
2
comes from the fact that two time slots (i.e., two channel uses) are needed to transmit symbols
0 5 10 15 20 25 30 35 40
10
4
10
3
10
2
10
1
10
0
SNR (dB)
P
e
noncooperative
cooperative
hybrid NAF
Fig. 3. Outage probability for Cooperative/non-cooperative/hybrid NAF cooperative
transmission with N = 2, M = 5
I
d
= log
2
(1 +
P
2
| f |
2
)
I
NAF
=
1
2
log
2
det(I +H Q H
)
I
OAF
=
1
2
log
2
(1 + P
s
| f |
2
+
P
s
P
r
|bgh|
2
1 + P
r
|bg|
2
)
Based on these mutual information expressions, in (E. Calvanese Strinati and S. Yang and
J-C. Belore, 2007) the authors propose to numerically compare non-cooperative, NAF
cooperative, hybrid NAF cooperative and hybrid OAF cooperative protocols in terms of
outage probability versus average SNR. Let O
d
denotes the direct channel outage event,
O
d
= {I
d
< R}, and O
c
denotes the cooperative channel outage event, O
c
= {I
c
< R}.
The equivalent channel is in outage if both events, O
d
and O
c
, are realized.
4.1.1 Simulation results
We report here some signicant simulation results to evaluate effectiveness of the proposed
hybrid cooperation protocol when applied to NAF cooperation on Fig 3.
Next, we extend the study of the hybrid cooperation protocol to OAF schemes and
we introduce a power allocation algorithm well designed for OAF hybrid cooperative
transmission. We nd out that transmission outage is slightly smaller adopting hybrid
cooperation for OAF scheme than for NAF one. Nevertheless, there are other important
10
-4
10
-3
10
-2
10
-1
10
0
0 5 10 15 20 25 30 35 40 45
P
e
P (dB)
non-cooperative
NAF
hybrid NAF
hybrid OAF
Fig. 4. Outage probabilities for the non-cooperative, NAF, Hybrid-NAF and Hybrid OAF
scheme. Considered information rates: 2 and 4 BPCU.
advantages in adopting an OAF hybrid cooperation protocol. First, the cooperation
complexity and cost are reduced. Second, the hybrid strategy reduces signicantly the
complexity of the algorithm implemented to determine the outage probability. This is the
key reason for which we succeeded in nding an optimal power allocation algorithm for OAF
hybrid cooperation schemes. We now show some simulation results for hybrid cooperative
transmission without power allocation. Performance is compared in terms of average outage
probability versus average SNR.
Based on these mutual information expressions, we numerically compare non-cooperative,
NAF cooperative, hybrid NAF cooperative and hybrid OAF cooperative protocols in terms
of outage probability versus average SNR. Let O
d
denotes the direct channel outage event,
O
d
= {I
d
< R}, and O
c
denotes the cooperative channel outage event, O
c
= {I
c
< R}. The
equivalent channel is in outage if both events, O
d
and O
c
, are realized.
Other simulation results are shown in Figure 4 for the case of one active relay and transmission
rate of 2 and 4 bits per channel use (BPCU). We nd out that, adopting the proposed
OAF hybrid cooperation protocol, transmission outage performance is better than for both
non-cooperative and NAF hybrid cooperation transmissions. This result conrms our choice
of using an orthogonal scheme: since the channel is assumed to be quasi-static, if the direct
link is in outage in the rst slot, it will remain in outage in the second one. The outage
performance improvement is not our major achievement. Combining hybrid cooperation with
OAF scheme, we obtain a cooperation protocol with both reducedcomplexity and cooperation
cost. Furthermore, the proposed hybrid strategy permits to reduce the complexity of the
outage probability computation. This is the key reason for which we succeeded in nding
an optimal power allocation algorithm only for OAF hybrid cooperation schemes.
The Orthogonal AF strategy, sub-optimal in a full time cooperation scheme, is optimal with
the hybrid strategy. In fact, since the channels are assumed to be slow fading, if the direct link
is in outage in the rst slot of the frame, it will be the case in the second. So it is better not to
transmit in the second slot, and thus economize power, since we are sure that the reliability of
the information is not guarantied. The mutual information is in this case
I
OAF
=
1
2
log
2
(1 + P
s
| f |
2
+
P
s
P
r
|bgh|
2
1 + P
r
|bg|
2
) (1)
4.2 Proposed Adaptive Modulation and Coding Combined with the Hybrid Cooperation
Protocol
In this section we present the mechanism proposed in (E. Calvanese Strinati and S. Yang and
J-C. Belore, 2007) in which the authors propose to combine the hybrid cooperation protocol
with an AMC mechanism. The protocol is named hybrid cooperative AMC mechanism. A ow
chart of the proposed algorithm is shown in Fig. 5. I
noncoop
is the instantaneous mutual
information when transmission is done in non-cooperative mode and R is the transmission
rate.
The algorithm is summarized as follows:
Step 1: S sends a RTS each time it wants to transmit new data.
Step 2: After receiving a RTS, the AMC mechanism (in D) selects R for next data transmission.
R is selected from the set of LUT of PER versus LQM for hybrid cooperation transmission
performance, given the LQM computed at previous received packet.
Step 3: D estimates the instantaneous channel conditions of the direct source-destination link
(
2
, f , etc.) and computes I
noncoop
( f ,
2
)
Step 4: The cooperation controller in D decides if cooperate or not:
- if I
noncoop
< R, non-cooperative transmission is forecasted to be in outage: the
cooperation controller starts cooperation (go to step 5)
- otherwise, cooperation mode is not activated (go to step 9)
Step 5: D checks if the relay probing is up to date:
- YES (go to step 9)
- NOT (go to step 6)
Step 6: relay probing: D probes the relays available for cooperation and estimates the channel
coefcients of the cooperation links.
Step 7 and 8: Each relay calculates the product gain |g
i
h
i
| and reacts by sending an availability
frame after t
i
time which is anti-proportional to |g
i
h
i
|. Therefore, the relay with the strongest
product gain is identied as relay 1, and so on.
Step 9: D sends a clear to send (CTS) that includes information on transmission rate R, M, relay
identiers, etc.
Step 10: S starts data transmission at rate R
Step 11: After receiving data from S, D derives PER
pred
from the LUT of hybrid cooperation
and selects R for next transmission of S.
Summarizing, based on the direct source-destination link quality, a cooperation controller
decides if and how cooperate. We call this cooperation protocol as hybrid cooperation. The rate
R is chosen after each received packet by the AMC that aims at maximizing the throughput
performance of the hybrid transmission mode meeting the QoS constraints imposed by the
upper layers.
Note that the AMC mechanism selects R based on a set of pre-computed AMC switching
points that depends on N, M, PER
target
, transmission scenario, etc. Such switching points
are chosen based on the average PER versus average performance of the hybrid cooperation
protocol. Given N, M and R, there is a crossing point (PER
cross
) between non-cooperative
and cooperative average performance. For PER PER
cross
cooperation outperforms
non-cooperative mode. Hence the gain of hybrid cooperation is high since the direct
link results more often in outage that cooperative transmission. When PER > PER
cross
,
non-cooperative transmission outperforms cooperation. In such case the gain of hybrid
cooperation is reduced and asymptotically (for PER
cross
0) hybrid cooperation performs
as non-cooperative transmission since cooperation is never activated. In order to fully exploit
the proposed hybrid cooperative AMC to improve the average system performance, AMC
mechanism and hybrid cooperation protocol have to be designed jointly. As an example,
given our system model, we computed the minimum values of M (M
min
) for which hybrid
cooperative AMC outperforms both classical non-cooperative and cooperative AMC. A
selection of our results are shown on table 1 for maximum transmission rates R
max
at which
the system can operate and typical PER
target
values imposed to the AMC. Indeed, given
N M
min
PER
target
R
max
2 9 10
1
10
2 5 10
2
10
2 7 10
1
8
2 3 10
2
8
2 5 10
1
6
2 3 10
2
6
2 5 10
1
4
2 3 10
2
4
2 3 10
1
2
2 3 10
2
2
Table 1. Minimum values of M (M
min
) for typical PER
target
values
PER
target
and R
max
, we can dene an M
min
from which hybrid cooperation is benecial. Note
that the larger M is the more complex the cooperation protocol is. There is indeed a trade off
between cooperation performance and cooperation complexity.
In this section, we show by means of numerical simulations the effectiveness of combining
the hybrid cooperation protocol with the AMC mechanism. Results rst show how the
proposed mechanism drives to improved average system throughput performance. Then, we
outline the advantage introduced by the hybrid cooperation protocol in terms of reduction of
cooperation signalling overhead, cooperation protocol delay and average power consumed
by the active relays. Simulation results are given here for the system model presented
in section 3. In the system both AMC and ARQ are implemented. The simulated AMC
algorithm selects the MCS which maximizes the throughput while meeting the PER
target
D: I
non-coop
R
D: channel estimation
up to date?
D: compute I
non-coop
(
2
z
)
D: rate R selected based
on the LQM
S: RTS
D: relay probing
Available relays
answer in order
D: selection of the best relays
D: CTS(R,M,relay identiers,etc.)
S: starts transmission
D: PER prediction
Yes
Yes
No
No
1
2
3
4
5
6
7
8
9
10
11
Fig. 5. Flow chart of the proposed hybrid opportunistic cooperation combined with AMC
0 5 10 15 20 25 30 35 40
0
1
2
3
4
5
6
7
8
SNR dB
R

(
1
P
E
R
)
cooperative
noncooperative
hybrid
Fig. 6. Cooperative/non-cooperative/hybrid cooperative transmission with N = 2 , M = 3
and PER
target
= 10
2
QoS constraints. The set of MCS corresponds to the transmission rate set R=1, 2, 4, 6, 8. We x
the PER
target
= 10
2
. Moreover, a total average power constraint is imposed and no power
allocation is considered here. We access the average physical layer throughput of a system
that can performdata transmission with three different transmission modes: non-cooperative,
cooperative and hybrid. Performance is compared in terms of average throughput versus
average SNR. The link between source, destination and relays are assumed to be symmetric
and with independent fading coefcients.
On Fig. 6 we show the performance of the AMC algorithm combined with cooperation for
N = 2 and M = 3. From these results, we observe three regions for the SNR : the low, medium
and high SNR regions. At lowSNR, the non-cooperation mode outperforms cooperation mode
since the noise power dominates the received power at the relays. In the medium SNR region,
the cooperative scheme outperforms the non-cooperative scheme with a gain up to 6 dB. This
gain is due to the better diversity-multiplexing trade-off (DMT) of the cooperative scheme.
However, this gain decreases for increasing SNR since we x PER
target
= 10
2
while R
max
= 8
and M = 3 (hence M < M
min
, see table 1). Therefore, when M < M
min
, the cooperative
scheme is not preferable at high SNR.
On Fig. 7 the performance of the case N = 2 and M = 5 is shown. As demonstrated in (S. Yang
and J-C. Belore, 2006), the DMT is improved with the number of slots M. This improvement
translates into a better performance in both cases. We observe that the decrease of SNR gain
at medium to high SNR is slower than the previous case. Cooperation is always better than
the non-cooperation since M M
min
. Best performance is always reached when using hybrid
cooperation. We remark that the hybrid scheme alleviates the performance loss of cooperation
0 5 10 15 20 25 30 35 40
0
1
2
3
4
5
6
7
8
SNR dB
R

(
1
P
E
R
)
cooperative
noncooperative
hybrid
Fig. 7. Cooperative/non-cooperative/hybrid cooperative transmission with N = 2 , M = 5
and PER
target
= 10
2
in both the low SNR and the high SNR regions. In case of M = 3 and M = 5, we observe
respectively up to 5 and 7.5 dB of gap from xed-cooperation and 1.5 and 2 dB of gap from
non-cooperative transmission.
Hereafter we enlarge the investigation on hybrid cooperation protocols performance for a
realistic communication scenario such as, OFDMA based wireless mobile communication
transmission which employs limited modulation alphabets and real FEC codes. We access
the effectiveness of hybrid cooperation protocol in real communication scenarios in terms of
average PER versus average SNR, average system throughput enhancement and average
cooperation cost reduction. The set of parameters used in this simulations are chosen
according to the IEEE 802.16e standard . The mobile wireless channel is modelled according
to (Spatial Channel Model Ad Hoc Group, 2003).
We propose to use an OAF hybrid cooperation protocol under the following power constraint:
we impose a total average power constraint and no power allocation is considered. If P
denotes the total power constraint, we impose P
s
= P/2 for the power allocated to the
source in the rst slot and P
r
= P/2 the power allocated to the relay in the second slot.
Hereafter we adopt the following graphical notation: we represent respectively with the solid
blue line, dashed red line and solid green line, non-cooperative, persistent cooperative and
hybrid cooperative transmission mode performance.
Simulation results are given here for the system model presented in section 3. We use as
Forward Error Correcting (FEC) code the LDPC codes as specied by the standard IEEE
802.16e (IEEE Standards Department, 2005) for the different coding rates.
8 10 12 14 16 18 20 22 24 26 28
10
3
10
2
10
1
10
0
SNR dB
P
E
R
Persistent Cooperation
Non Cooperation
Hybrid Cooperation
64 QAM 2/3
64 QAM 3/4
64 QAM 1/2
Fig. 8. Cooperative/non-cooperative/hybrid cooperative transmission
On gure 8 we compare the three transmission mode performance in terms of average PER
performance versus average SNR. Results are reported here only for 64-QAM modulation
with coding rates R
c
= 1/2, 2/3, 3/4. From these results, we observe that there is a
crossing point (PER
cross
) between non-cooperative and cooperative average performance.
For PER PER
cross
cooperation outperforms non-cooperative mode. Hence the gain of
hybrid cooperation is high since the direct link results more often in outage that cooperative
transmission. Note that the PER that corresponds to this crossing point depends on the code
correcting power: stronger codes present the crossing point at higher PER. For sake of simplicity
we impose same codeword length for each MCS. Therefore, the information block length is
larger for higher coding rate which results in a stronger correcting code. This is veried on
gure 8. When PER > PER
cross
, non-cooperative transmission outperforms cooperation.
When PER
cross
0, hybrid cooperation performs as non-cooperative transmission since
cooperation is never activated. Hybrid cooperation notably outperforms both cooperative
and non-cooperative transmissions for PER values close to PER
cross
. Note that in the
present simulation we also introduce a feedback delay between MI
noncoop
estimation and
cooperation controller action. Due to this delay, hybrid cooperation performance is slightly
decreased comparing to equivalent results presented in (E. Calvanese Strinati and S. Yang and
J-C. Belore, 2007).
In order to show the effectiveness of hybrid cooperative AMC mechanism, which combines
AMC with hybrid cooperation, we compare the three transmission modes in terms of average
system throughput versus average SNR. The simulated AMC algorithm selects the MCS
which maximizes the throughput while meeting the PER
target
QoS constraints (Calvanese
Strinati E., 2006). Typical values for the target PER is a few percent. For instance, imposing
PER
target
10
1
results in a residual PER below 10
5
after 4 retransmissions.
The set of MCS corresponds to the transmission rate set dened by the IEEE 802.16e standard.
In our simulation results we show the per-user performance, having one data region of 24
sub-carriers (in frequency) and 16 data OFDM symbols (in time). Under this assumption, the
set of MCS schemes and the related nominal throughputs r
mcs
and information block lengths
N
Info
are given in table 2.
Modulation Code Rate N
Info
r
mcs
QPSK 1/2 384 (bits) 215 (Kb/s)
QPSK 3/4 576 (bits) 315 (Kb/s)
16-QAM 1/2 768 (bits) 420 (Kb/s)
16-QAM 3/4 1152 (bits) 630 (Kb/s)
64-QAM 1/2 1152 (bits) 630 (Kb/s)
64-QAM 2/3 1536 (bits) 840 (Kb/s)
64-QAM 3/4 1728 (bits) 945 (Kb/s)
Table 2. Modulation and Coding Schemes of IEEE 802.16e
When PER
target
< PER
cross
, then cooperation is always better than the non-cooperation.
Otherwise, non-cooperation transmission can outperform persistent cooperation
transmission. As an example, we report respectively on gure 10 and 9 our simulation
results for PER
target
= 10
1
, 5 10
2
.
As it is shown on gure 9, with PER
target
= 5 10
2
, persistent cooperation outperforms
non-cooperative transmission over all the considered SNR range since, PER
target
< PER
cross
for all MCS.
In this case, hybrid cooperation outperforms non-cooperative and persistent cooperative
transmission respectively with a gain up to 1.75 dB and 0.75 dB. Relaxing the constraint on the
PER
target
to PER
target
= 10
1
, there are some MCS for which PER
target
> PER
cross
. As a
consequence, non-cooperation outperforms persistent cooperation in same parts of the considered
SNR range. Again, hybrid cooperation outperforms non-cooperative and persistent cooperative
transmission respectively with a gain up to 1.25 dB and 0.9 dB (see gure 10).
We report hereafter also some simulation results aimed at understanding the average relaying
activation ratio ) - which is the ratio between the number of frames were the relay is active
over the total number of transmitted frames - versus the average SNR adopting the proposed
hybrid cooperation protocol. Results are shown on Fig 11 for PER
target
= 10
1
. Two working
zones of an AMC mechanism can be distinguished. In the rst zone, even if AMC selects the
minimum MCS at which the system can operate, we have that PER > PER
target
. Therefore,
since PERis large, is large too. For such link quality conditions the AMCmay decide to avoid
transmission since AMC cannot assure the QoS constraints imposed by the upper layers. The
second zone starts when MCS selected for transmission assures PER PER
target
. In this
zone each saw tooth corresponds to a change of MCS. Our results outline that when AMC
can assure a PER PER
target
, is very small ( PER
target
) since the hybrid cooperation
protocol activates the cooperative mode only when direct link transmission is in outage. At
the end of the second zone transmission is done at the highest MCS and the system operates
at PER PER
target
, with consequent 1. Note that, contrary to the cooperative AMC
protocol case for which = 1 over the whole SNR range, when AMC can assure a PER
PER
target
and the proposed hybrid cooperation protocol is adopted, is reduced to the same
0 5 10 15 20 25 30
0
100
200
300
400
500
600
700
800
900
SNR dB
T
h
r
o
u
g
h
p
u
t

(
K
b
/
s
)
Persistent Cooperative Transmission
Non Cooperative Transmission
Hybrid Cooperative Transmission
Fig. 9. Cooperative/non-cooperative/hybrid cooperative transmission with
PER
target
= 5 10
2
order of magnitude of PER
target
. Note that the major result in our investigation is reduction of
average relaying activation and not the improvement in average system throughput achieved
with hybrid cooperative AMC mechanism.
The reduction of average relaying activation ratio achieved with the proposed hybrid AMC
protocol presents three main advantages. First, the average power consumed by the active
relays is strongly reduced especially when cooperation does not help and consequently
cooperation activation results in a waist of relays processing power. Second, the delay
caused by the cooperation protocol and consequently the packet delivery delay can be
strongly reduced adopting our proposed hybrid cooperation protocol. For instance, when
direct non-cooperative transmission is not forecasted to be in outage, the destination can
immediately send a clear to send (CTS), without waiting for the relay probing process. This
is an important attribute for scheduling algorithm with delay QoS constraints. Third, the
average computing complexity is reduced by decreasing the number of average operation
associated to cooperation.
4.3 An efcient power allocation optimization for hybrid cooperation protocols
In this section we combine the OAF hybrid cooperation protocol presented in section 4.1 with
an optimal power allocation algorithm. The goal is to maximize the mutual information of the
equivalent cooperative channel via optimal power allocation between the source and the relay.
It is well known that the performance of a cooperative scheme is improved by relaying with
optimal power values. Hereafter we assume that a maximal overall transmit power is xed
by using, for instance, a suitable power control algorithm in order to minimize co-channel
0 5 10 15 20 25 30
0
100
200
300
400
500
600
700
800
900
SNR dB
T
h
r
o
u
g
h
p
u
t

(
K
b
/
s
)
Persistent Cooperative Transmission
Non Cooperative Transmission
Hybrid Cooperative Transmission
Fig. 10. Cooperative/non-cooperative/hybrid cooperative transmission with
PER
target
= 10
1
interference. The overall total transmitting power should then be optimally shared between
the source and the relay. The simplicity of an OAF cooperation scheme leads to an outage
probability expression easier to handle than in the NAF case. Basically, we optimize the power
allocation by minimizing the outage probability in the high SNR regime.
4.4 Outage probability approximation
First we should nd the expression of the outage probability, denoted P
_
O
c
, O
d
_
, and
approximate it in the high SNR regime. Proposition 1: Let P denotes the total power constraint
in the network, P
s
= P and P
r
= (1 )P the fractions of P allocated to the source and
the relay, respectively. Let C
=

g
h
and C
R
=
1
2
R
+1
. Then, the approximation of the outage
probability in the high SNR regime is
P
_
O
c
, O
d
_
=
2(2
R
1)
2
(2
R
+1)
2
h
_
1 +C
(1 )
__
1 C
R
_
Proof: The following Lemma will be used in our proof
Lemma 1: Let be positive, and let r
=
vw
v+w+
where v and w are independent exponential
random variables and
v
and
w
are, respectively, their parameters. Let h() be continuous
with h() 0 as 0. Then
lim
0
1
h()
P
_
r
< h()
_
=
v
+
w
0 5 10 15 20 25 30
10
3
10
2
10
1
10
0
SNR dB

:
c
o
o
p
e
r
a
t
i
o
n

a
c
t
i
v
a
t
i
o
n

r
a
t
i
o
Fig. 11. Average relaying activation ratio for hybrid cooperative transmission with
PER
target
= 10
1
P
_
O
c
, O
d
_
= P
_
| f |
2
+
|h|
2
(1 )|g|
2
|h|
2
+ (1 )|g|
2
+ P
1
<
2
2R
1
P
,
| f |
2
2
<
2
R
1
P
_
(2)
= P
_
u +
vw
v + w +
< (2
2R
1)P
1
, u < 2(2
R
1)P
1
_
(3)
= P
_
r
< g
1
() u, u < g
2
(, )
_
(4)
P
_
O
c
, O
d
_
= 2(2
R
1)
2
f
_
+

g
1
__
(2
2R
1) (2
R
1)
_
(5)
We know that
P
_
O
c
, O
d
_
= P
_
I
c
< 2R, I
d
< R
_
The outage probability can be expressed as in (2), if we dene u = | f |
2
, v = |h|
2
, w =
(1 )|g|
2
, = P
1
, g
1
() =
(2
2R
1)
P
, and g
2
(, ) = 2
(2
R
1)
P
.
Let
u
,
v
and
v
be the parameters of the exponential random variables u, v and w,
respectively. For i = f , h, we have
i
=
1
2
i
=
1
i
and
w
=
1
(1 )
2
g
= (1 )
1
g
10
-4
10
-3
10
-2
10
-1
10
0
0 5 10 15 20 25 30 35 40
P
o
u
t
P (dB)
no cooperation
Hybride NAF plus 10 dB
Hybride OAF plus 10 dB
Hybride OAF PC plus 10 dB
10
-4
10
-3
10
-2
10
-1
10
0
0 5 10 15 20 25 30 35 40
P
o
u
t
P (dB)
no cooperation
Hybride NAF moins 10 dB
Hybride OAF moins 10 dB
Hybride OAF PC moins 10 dB
Fig. 12. Outage probabilities for the non-cooperative, Hybrid-NAF, Hybrid-OAF and
Hybrid-OAF with power allocation scheme. One relay network. Considered information
rates: 1, 2, 3 and 4 BPCU. C
= 10 dB.
Using Lemma 1, we get
P
_
O
c
, O
d
_
=
_
g
2
0
P{r
< g
1
() u}p
u
(u)du
=
_
g
2
0
(
v
+
w
)(g
1
() u)p
u
(u)du.
Knowing the pdf of the exponential variable u, the expression of P
_
O
c
, O
d
_
is developed
(calculation details are omitted due to length constraints). This expression is then
approximated in the high SNR regime, using the second order Taylor development of e
a
when 0, a being positive, which leads to expression (5).
Eventually, dene C
=

g
h
and C
R
=
1
2
R
+1
which, when substituted in (5), complete the
proof.
For a given spectral efciency R and channels variances, optimizing the power allocation
consists in minimizing the outage probability and thus, nding the optimal , denoted
,
that veries
(C
C
R
1)
2
+2
1 = 0 (6)
In order to clarify the impact of the proposed power allocation algorithm we compare
non-cooperative, NAF cooperative, hybrid NAF cooperative and hybrid OAF cooperative
protocols in two different transmission scenarios. Fist we suppose that both path-loss and
shadowing effects are the same between source, relay and destination. This scenario is
specied by C
= 0 dB, so that we have

2
h
=
2
g
. In this case
is
=
1
1 +
1 C
R
We observe that minimizing the outage probability leads to almost an equal power allocation
between the source and the relay since
takes values around 0.5 independently from the

transmission spectral efciency. We evince that, when C
= 0 dB, the algorithm of power

10
-4
10
-3
10
-2
10
-1
10
0
0 5 10 15 20 25 30 35 40
P
o
u
t
P (dB)
no cooperation
Hybride NAF plus 20 dB
Hybride OAF plus 20 dB
Hybride OAF PC plus 20 dB
10
-4
10
-3
10
-2
10
-1
10
0
0 5 10 15 20 25 30 35 40
P
o
u
t
P (dB)
no cooperation
Hybride NAF moins 20 dB
Hybride OAF moins 20 dB
Hybride OAF PC moins 20 dB
Fig. 13. Outage probabilities for the non-cooperative, Hybrid-NAF, Hybrid-OAF and
Hybrid-OAF with power allocation scheme. One relay network. Considered information
rates: 1, 2, 3 and 4 BPCU. C
= 20dB.
allocation optimization performs as an equal power allocation P
s
= P
r
= P/2. This is obvious
since source-relay and relay-destination links have the same link quality.
As a second scenario, we consider the more realistic case where C
0 dB. Actually, having

C
0 dB, we assume that one of the links, source-relay or relay-destination, has a better
quality, i.e., (
2
h

2
g
). Optimizing the power allocation becomes more worthy in this situation
since allocating more power to the worst channel helps. In this case,
can be derived from

(6) as follow:
=
1
1 +
_
C
(1 C
R
)
On Figures 12 and 13 we consider the case of C
> 0 dB, having respectively, C
= 10 dB
and C
= 40 dB. In this scenario, e.g., the attenuation between source and relay is much
smaller than between relay and destination. In this case, if the cooperation is activated by
the hybrid cooperation controller, our power optimization allocates a higher fraction of the
overall transmit power P to the relay.
A more challenging scenario is when C
< 0 dB or equivalently
2
h
<
2
g
. In this case, an
optimal power allocation algorithm can drive to notable performance improvement. Mainly,
making reliable the transmission between the source and the relay is imperative since the
relay amplies and then forwards the received signal. That is why our optimization technique
allocates, in this case, a higher fraction of P to the source. Simulation results for C
= 10 dB
and C
= 40 dB are given on Figures 12 and 13.

5. Conclusion
In this chapter we present an effective scheme to improve the system performance of a
cooperative system, reduce cooperation complexity, signalling overhead and cooperation
protocol delay, while meeting the QoS constraints from the upper layer. For this reason, we
looked for a novel AF cooperative protocol, and its combination with adaptive mechanisms
such as AMC and power allocation.
First, we propose a novel cooperation protocol for half-duplex AF cooperative networks. We
call this protocol hybrid cooperation. We prove by simulation that, NAF hybrid cooperation
outperforms both non-cooperative and classical full-cooperative transmission. To evaluate
the improvement due to this new strategy, we also propose an hybrid cooperative AMC
mechanism, which is the combination of AMC mechanism and hybrid cooperation protocol.
We show that the advantages of hybrid cooperative AMC are twofold. First, its average
throughput performance is higher than both AMC combined with non-cooperative and
with xed-cooperation transmission for all values of SNR. This results is benchmarked
by our simulation results. Second, the proposed algorithm drives to a reduction of both
average power consumed by the active relays and cooperation probing cost. This results in
a reduced average packet delivery delay since both throughput performance is improved
and cooperation probing delay is strongly reduced. Moreover, we showed how the proposed
hybrid cooperative AMC mechanism drives to a reduction of cooperation signalling overhead
that from a MAC layer point of view, may result in an additional throughput enhancement at
the top of the MAC layer.
We further investigate the proposed hybrid AF cooperation protocol. We compared hybrid
OAF and hybrid NAF protocols. Imposing a total average power constraint and no power
allocation, we showed that the orthogonal strategy (OAF), suboptimal in the case of a classical
amplify-and-forward scheme, outperforms both classical NAF cooperative and hybrid NAF
schemes. Moreover, we pointed out that from an implementation point of view, the hybrid
OAF protocol reduces signicantly the cooperation complexity.
Furthermore, we prot of the simplicity of the outage probability expression for the OAF
cooperation scheme to derive an optimal power allocation algorithm. The proposed algorithm
optimizes the system performance by minimizing the outage probability of the channel at
high SNR. We underlined that the need of such an optimization increases with the increasing
quality difference within the links (source-relay and relay-destination). Indeed, we succeeded
in nding a low complexity algorithm that optimizes the power allocation in the case of a
hybrid-OAF schemes.
6. References
E. Calvanese Strinati. Radio link control for improving the qos of wireless packet transmission. PhD
thesis, Ecole Nationale Suprieure des Tlcomunications de Paris, December 2005.
E. Calvanese Strinati, S. Yang, and J-C. Belore. Adaptive Modulation and Coding for Hybrid
Cooperative Networks. June 2007.
Emilio Calvanese Strinati and Luc Maret, Performance Evaluation of Hybrid Cooperation
Protocol in IEEE 802.16e, IEEE Vehicular Technology Conference (VTC Spring),
Singapore, Mai 2008.
Maya Badar and Emilio Calvanese Strinati and Jean-Claude Belore, Optimal Power
Allocation for Hybrid Amplify-and-Forward Cooperative Networks, IEEE Vehicular
Technology Conference (VTC Spring), Singapore, Mai 2008.
E. Erkip A. Sendonaris and B. Aazhang. User cooperation diversity-part 1: System
description. 51:19271938, November 2003.
E. Erkip A. Sendonaris and B. Aazhang. User cooperation diversity-part 2: Implementation
aspects and performance analysis. 51:19391948, November 2003.
D. Gunduz and E. Erkip. Outage Minimization by Opportunistic Cooperation. 2:14361442,
June 2005.
D. N. Tse J. N. Laneman and G. W. Wornell. Cooperative diversity in wireless networks:
Efcient protocols and outage behavior. 50:30623080, December 2004.
H. Blcskei R. U. Nabar and F. W. Kneubhler. Fading relay channels: Performance limits and
space-time signal design" algorithm. ieeeJSAC, pages 10991109, August 2004.
H. El Gamal K. Azarian and P. Schniter. On the achievable diversity-multiplexing tradeoff in
half-duplex cooperative channels. 51:41524172, December 2005.
S. Yang and J-C. Belore. Optimal space-time codes for the mimo amplify-and-forward
cooperative channel. IEEE Trans. Inform. Theory, May. 2006.
Z. Lin and E. Erkip and M. Ghosh. Adaptive Modulation for Coded Cooperative Systems.
pages 615
U619, June 2005.

M. Lampe and H. Rohling and W. Zirwas, Misunderstandings about link adaptation for
frequency selective fading channels, IEEE International Symposium on Personal,
Indoor, and Mobile Radio Communications , September 2002.
E. Yazdian and M. R. Pakravan. Adaptive Modulation Techniques for Cooperative Diversity
in Wireless Fading Channels, in proceeding of IEEE PIMRC, September 2006.
M. Hasna and M-S. Alouini. Optimal power allocation for relayed transmission over
rayleigh-fading channels, IEEE Trans. Inform. Theory, 3(6), November. 2004.
Q. Zhang and C. Shao and Y. Wang and P. Zhang and J. Zhang and Z. Zhang. Zhang,
Adaptive optimal transmit power allocation for two-hop non-regenerative wireless
relaying systems, Vol 41 :124
U133, September 2004.

D. P. Reed and A. Bletsas and A. Khisti and A, Lippman. A simple cooperative diversity
method based on network path selection, IEEE Journal on Selected Areas of
Communication, 2005.
I. Hammerstrom and A. Wittneben, On the optimal power allocation for nonregenerative
OFDM relay links, in Proc. IEEE Int. Conf. Communications (ICC), June 2006.
Spatial Channel Model Ad Hoc Group (Combined ad-hoc from 3GPP and 3GPP2), Spatial
Channel Model Text Description, SCM-134, April 22, 2003.
IEEE Standards Department, Part 16: Air Interface for Fixed Broadband Wireless Access
Systems - Amendment 2: Physical and MediumAccess Control Layers for Combined
Fixed and Mobile operation in Licensed Bands and Corrigendum1, NewYork, IEEE
Std 802.16e-2005, Feb. 2006. (Available online at: http://www.ieee.org).
0
Adaptative Rate Issues in the WLAN Environment
Jerome Galtier
Orange Labs
France
1. Introduction
In this chapter, we investigate the problem of mobility in the WLAN environment. While
radio conditions are changing, the Congestion Resolution Protocol (CRP) plays a key role in
controling the quality of service delivered by the distributed network. We investigate different
types of CRP to show the impact of one user to all the other ones. We place our work in an
urban context where the users (bus passengers, walkers, vehicule network applications) are
using an accessible WLAN (for instance WiFi) network via an access point and interact with
one another through the network.
Accessing the network via an access point has become in the last years a more and more
popular technique to do some networking at low cost. The reason for it is that the
WLAN technologies such as WiFi do not require complex user registration, handovers,
downlink/uplink protocol synchronization, or even planication for existing base stations
(such as GSM BTS or UMTS Node-B). Of course such transmissions achieve much lower
performance prole, but they are often delivered for free or almost for free, for instance simply
to attract new clients in cafs or restaurants.
As a result, we come up wih new habits of communications which are not exactly the use for
which engineers have designed WiFi for L (and other WLAN networks).
2. Overview of 802.11 modulation techniques
2.1 Techniques employed
In the course of its developpment, the 802.11x family has developped a surprising number of
modulation techniques that deeply impact the nal performance of the system. We summarize
these techniques for a 20 MHz band in the 2.4 GHz frequency area in Tab. 1 (we skip here
all the historical modulations that have since been abandonned). All these cards implement
backward compatibility, which means that the most recent and sophisticated one also handles
previous rates in order to be able to communicate with simpler/older cards. As a result, a
new 802.11n card with 4 streams will be able to produce modulations in 44 different modes!
We give in the following some explanations on the different modulation techniques employed
for WiFi.
BPSK Binary Phase Shift Keying is a modulation technique that uses the phase of two
complementary phases to code the bis 0 or 1. We plot its constellation diagram in Fig. 1.
QPSK Quadrature Phase Shift Keying uses four phases instead of two to code the signal, so
that each symbol carries 2 bits instead of 1 for the BPSK.
10
2 Will-be-set-by-IN-TECH
Protocol Data rate per stream (Mbits/s) Modulation & Coding
(# streams)
- 1,2 DSSS/BPSK QPSK/Barker seq.
b 5.5,11 DSSS/QPSK/CCK
g 6,9,12,18,24,36,48,54 OFDM/BPSK QPSK QAM/Conv. coding
n (1 st.) 7.2,14.4,21.7,28.9,43.3,57.8,65,72.2 OFDM/MIMO/Conv. coding
n (3 st.) 21.7,43.3,65,86.7,130,173.3,195,216.7 OFDM/MIMO/Conv. coding
Table 1. Different rate parameters for 802.11x at 2.4GHz within a 20 MHz band.
(a) BPSK. (b) QPSK. (c) QAM.
Fig. 1. Constellation diagram for main 802.11 modulation techniques.
QAM Quadrature Amplitude Modulation combines amplitude modulation with phase
modulation to carry more information. 16-QAM carries 4 bits, while 64-QAM carries 6
bits.
DSSS The Direct Spread Sequence Spectrum is a modulation technique that uses the whole
band (here, 20 MHz) to encode the information via some coding techniques, more precisely
Barker codes or CCK in the 802.11 context.
Barker sequences For the 1 Mbit/s coding, the pseudo-random sequence (10110111000) is
used to code the 1 symbol, and its complement (01001000111) to code the 0 symbol, in
a PSK modulating scheme. The 2 Mbit/s version is obtained by using QPSK modulation
instead of PSK.
CCK The CCK (Complementary Code Keying) technique consists in using 16 or 256 different
sequences coded in eight chips (QPSK symbols). The 16 or 256 different sequences allow
to identify 4 or 8 bits of information.
OFDM The OFDM (Orthogonal Frequency-Division Multiplexing) is a technique that
consists in dividing the channel into close sub-carriers to transmit data through these
parallel sub-channels. Using orthogonality of signals, this technique allows to reduce
signicantly the spacing between sub-carriers and therefore improves spectral efciency.
MIMO The Multiple-Input and Multiple-Output technique consists in using several input
antennas and several output antennas in the devices, in order to use spatial diversity and
therefore increase the throughput capacity.
2.2 Range of communication
In a paper on adaptativiy and mobility, of course, the range of communication is a crucial
parameter. Unfortunately, this very range is very variable depending on radio conditions. We
try in this subsection to give a more accurate opinion on that question without falling into two
main defaults of the literature on that topic, that would be (1) rely exclusively on simulations
or (2) explain the theoretical context without answering the question of range.
Data rate Modulation Coding rate
(Mbits/s) (R)
6 BPSK 1/2
9 BPSK 3/4
12 QPSK 1/2
18 QPSK 3/4
24 16-QAM 1/2
36 16-QAM 3/4
48 64-QAM 2/3
54 64-QAM 3/4
Table 2. Rate parameters in 802.11a and 802.11g.
We aim to obtain a realistic model for WLAN communications the following way. We
take in this section as an example the model of IEEE 802.11g which is a precise, popular
industrial context, in which the problem is really accurate. As mentioned in (Part 11: wireless
LAN medium access control (MAC) and physical layer (PHY) specications, 1999, chapter 17), the
different rates in which such networks operate are the ones of Tab. 2. We can see a continuum
of rates that varies from 6 Mbits/s to 54 Mbits/s. However, the radio conditions impact a lot
the effective performance of terminals. We do not want to enter too deeply into some specic
situations in this paper, but instead we try to extract general enough properties that can be
extrapolated to numerous contexts. We need to say that, surprisingly, the theory says that
BPSK has exactly the same performance as QPSK, so we will only plot curves for QPSK. We
use rst the berfading function of mathlab, and evaluate the coding gain as 1/R. This simple
implementation gives the plots of Fig. 2, for respectively, (a) a Rayleigh channel of diversity
order 2, and (b) a Rice channel of diversity order 3 and K-factor 5.
Although these curves are quite different, they have some characteristics that are kept not
only in these two cases, but with a large set of different diversity values ranging from 2 to 10
and, for the Rice channel, K-factors ranging from 1 to 100 or more. We plot in Fig. 3 the same
curves, that we normalized by taking the logarithm, substracting the mean of the logarithms
in the six cases, and scaling by the standard deviation.
This simple approach matches very well the data that some authors are giving on the range
of operations for different congurations, as we plot in Fig. 4, with indoor ranges of Romano
(2004), and ranges of WLAN - 802.11 a,b,g and n (2008). In (Segkos, 2004, page 93, Fig. 58),
again, similar ranges are shown. The conclusion are always the same:
1. within each group of modulation (QPSK, 4-QAM, 64-QAM) the curves are closer one to
another, than when one jumps from one group of modulation to another.
2. the closest curves are that of 48 and 54 Mbits/s,
3. the farthest curves are that of 36 and 48 Mbits/s.
189 Adaptative Rate Issues in the WLAN Environment
(a) Rayleigh channel of diversity order 2. (b) Rice channel of diversity order 3 and K=5.
Fig. 2. Various analytical SNR to BER behaviors.
(a) Rayleigh channel of diversity order 2. (b) Rice channel of diversity order 3 and K=5.
Fig. 3. SNR to normalized BER.
Some differences, however, appear. For instance, it sounds like in the simulations and analytic
tools (Fig 2 and (Segkos, 2004, page 93, Fig. 58)), the curves of rates 48 and 54 Mbits/s are much
closer than they are in the tutorial curves of Fig. 4. We have observed this phenomenon with
lots of different parameters for the Rayleigh and Rice channels.
As a result, we can conclude that the behavior of the channel for WLAN networks deeply
depends on the type of radio conditions that are experienced. However, all the experiments
we have done suggest that we can use a pathloss model where the gain follows a law in
K
r
(d/)
, where K
r
depends on the rate r of the connection, d is the distance to the access
point, and the pathloss parameter, depending on radio conditions. Fig. 3 gives a sufciently
precise behavior of all the mechanisms to evaluate the channel.
3. Existing rate adaptation algorithms
There exist several algorithms that are intended to nd the optimal rate of communication
for WiFi terminals while exploring the channel. In order to adapt the rate of the packet, such
Fig. 4. Comparative given ranges for 802.11g in the litterature.
an algorithm will certainly take into account the fact (or not) that the receiver emitted an
acknowledgement at the end of the transmission, since this is mandatory in the legacy mode
of IEEE 802.11x.
A very popular approach to that question is that of ARF (Auto-Rate Fallback) Kamermann &
Monteban (1997). The sender begins to send packets at the minimal rate. After N successes,
the sender increases the rate to the next available rate in terms of speed. In case a failure occurs
when the new rate is rst tried, the rate is xed to the previous one, and will not be improved
again until N successes occur. If, at a given rate, two successive failures occur, then the rate is
also decreased, and will also wait N successes before a new try to a faster rate is done.
An improved version, AARF (Adaptive Auto-Rate Fallback) Lacage et al. (2004), plays with the
value N of successes before a try to faster rate is done. If a rst test at a new rate fails, then the
systemwill wait 2Nsuccesses to try again. This results in an improvement of the performance
of the system.
The TARA scheme (Throughput-Aware Rate Adaptation) Ancillotti et al. (2009) combines the
information of the Congestion Window (CW) of the MAC of 802.11 with specic parameters
to improve the rate mechanism.
Another variant of ARF, ERA (Effective Rate Adaptation) Wu & Biaz (2007), tests, in case of
collision, a retry at the lowest rate. This retry is used to infer whether the failure is due to a
collision or to a radio (SNR) problem. The rate is changed only if the problem is supposed not
to be a collision problem, that is, when the retransmission at lower rate is successful.
Indeed, it is not true that all defaults of acknowledgement are due to radio conditions (and
accordingly rate of transmission). In fact, in saturation mode, up to 30% of packets are lost
because of collisions. This has led to additional research work mainly in four directions:
Obtain a measurement on the radio conditions. Two main methods are then possible. First,
one can assume link symmetry, so that the transmitter will evaluate the signal-to-noise
ratio of a packet from the receiver, Pavon & Choi (2003). Second, one can suggest to
modify the RTS/CTS mecchanismso that the CTS would send back the received SNRto the
receiver (Holland et al. (2001); Saghedi et al. (2002)). These ideas have several drawbacks as
mentioned in Ancillotti et al. (2008). The former can only give an assumed SNR, while the
latter supposes that both receiver and emitter implement this RTS/CTS modication. But
the main inconvenient is that both mechanisms assume the knowledge of a SNR-to-Rate
table, which is not easy to obtain and/or update.
Distinguish physically loss reason. Some algorithms lie on the fact that the physical layer
may distinguish packets that are lost due to channel collisions, from packets that are lost
because of collisions. This could be for instance implemented by a feature that allows the
receiver to say that he could decode the header of the packet (transmitted at minimal rate)
but not the payload in itself, see Pang et al. (2005). Of course this solution requires a very
specic hardware and nevertheless one cannot be sure that the collision detection feature
is fully reliable.
Test the channel in case of collision by an RTS/CTS. Several mechanisms J.Kim et al.
(2006); Wong et al. (2006) decide, in case of collision, to send an RTS/CTS to test the
channel. Of course, in case the channel comes close to saturation, this has a terrible impact
on performance. This has also the terrible side-effect of employing the hidden station
procedure (RTS/CTS) for a different purpose, which has several side-effects. Note that
this defaults are partially corrected using probabilities in Chen et al. (2007).
Use Beacon information. In the case where one terminal is connected to an Access Point,
some SNR information can be used to have a rst idea of the channel quality, see Biaz &
Wu (2008).
4. Analytical model
In this section, we make the use of Markov analysis to infer important properties of the rate
adaptation algorithm. We model in the following the ARF mechanism, knowing that this
model is very popular, and can possibly be extended to alternative approaches. This can be
viewed as an extension of the model of Bianchi (2000). It gives an interesting model where,
as expected, the rate is connected to the size of the congestion window of the backoff process.
Indeed, in IEEE 802.11x, when a station needs to send a packet, it goes through a phase
called contention resolution protocol (CRP) that aims at deciding which stations - among the
contending ones - will send a packet. In order to do that, it uses two main variables: the
contention window (CW) and the backoff parameter (b). The contention window is set at
the begining to a minimal value (CWMin) and doubles each time a failure is experienced,
to a maximum CWMax. When the maximum is reached, the CW parameter will stay at
this value for a xed number of retries as long as the transmission fails, and then aborts
sending and falls to CWMin. In case of success, in all these cases, the next CW is set to
CWMin. This typical behavior of CW has been discussed in many ways in the literature
(Galtier (2004); Heuse et al. (2005); Ibrahim & Alouf (2006); Ni et al. (2003)), therefore in the
following, we will simply use a series of constants CW
0
, . . . , CW
m
, with CW
0
=CWMin and
CW
1
being the value of CW after the rst augmentation (2CWMin in the legacy case), and
CW
mretries+1
= = CW
m
=CWMax. It gives, in the legacy case, the following formula:
CW
i
= min(2
i
CWMin, CWMax).
Meanwhile, when a transmission is to be done, the value of b is taken randomly between 0
and CW-1. If b equals zero, the station emits its packet as soon as the channel is available
(that is, when the previous transmission is completed). Otherwise, b is decremented and the
station waits for a small period (called a minislot) and listen to the channel to see if some other
station did not start to transmit. If it is the case, it postpones its own emission to the end of
the current transmission, and therefore freezes CW and b. Then, if b is equal to zero, it starts
emitting. Otherwise b is decremented and so on.
We describe our model in Fig. 5. In this gure the state s
i
j
corresponds to the situation where
the rate is j and the congestion window CW
i1
. The probability p
j
represents the probability
that a transmission at rate j fails. One can see in the gure also small states between s
i
0
and
s
i+1
0
. Those states represent the fact that N successful transmissions at rate r
i
are necessary to
try to send at rate r
i+1
. The colors of the vertices (or the levels of gray) correspond to the rate
of transmission of the state in question. The state s
r
+
represents the case where the current rate
is r and at least the last transmission at that rate was successful.
rate 1 rate 0 rate r rate r1
1
p
1
CW
4
CW
3
CW
2
CW
1
CW
0
CW
m1
CW
m
s
0
0
s
1
1
s
2
1
s
3
1
CW
r1
CW
r
CW
r+1
CW
r+2
s
m
0
s
m1
0
s
r
1
s
r1
0
s
4
0
s
1
0
s
2
0
s
3
0
s
r+2
0
p
2
p
2
p
2
p
2
p
2
p
2
p
1
p
1
p
1
p
1
p
1
p
1
p
1
p
1
p
0
p
0
p
0
p
0
p
0
p
0
p
0
p
0
p
0
1
s
r+1
0
s
r
0
1
p
0
1
p
0
1
p
0
1
p
0
1
p
0
1
p
0
s
0
1
1
p
1
1
p
1
1
p
1
1
p
1
1
p
1
1
p
1
1
p
0
s
r1
2
s
r3
2
s
r2
2
s
1
r1
s
2
r1
p
r
p
r
p
r1
1 p
r
1 p
r
p
r
1
p
r
1
1
p
r
1
1
p
r
1
1
p
r
1
1
p
r
1
1
p
r
1
1
p
r
s
1
r
s
0
r s
+
r
1
p
1
1
p
2
s
0
r1
1
p
r
1
s
r1
1
s
r2
1
s
2
2
s
1
2
1
p
2
s
0
2
Fig. 5. Model including ARF in the backoff process.
Note that, if we extend the model similarly to Bianchi (2000) to represent the backoff variable,
this is a discrete Markov process. Note also that a simpler model also appears in the literature
(Singh & Starobinski (2007)) that expresses two models, one for the rates, and one for the ARF
internal behavior, and mixes them based on semi-markov properties. Note that our model
captures ne properties of backoff behavior, and complex mechanisms of rate improvement.
4.1 Analysis of the left-hand part
Let now design by
j
i
the probability that the transmitter is in state s
i
j
. We also note
r
+
the
probability that the transmitter is in state s
+
r
. The analysis of the left part of Fig 5 gives:
_
r
+
= (1 p
r
)(
r
+
+
r
0
+
r
1
),
r
1
= p
r
r
+
.
(1)
And therefore
r
1
=
1 p
r
p
r
r
0
. (2)
4.2 Analysis of the intermediate part
Let us now analyze an intermediate level of the model. We can see on Fig. 5 that, for k
{2, . . . , r},
k1
i+1
= p
k
k
i
, for i {1, . . . , r +1 k},
k
0
= (1 p
k1
)
N
i=r+2k
i=0
k1
i
,
k1
1
= p
k
k
0
+ p
k1
((1 p
k1
) + (1 p
k1
)
2
+ + (1 p
k1
)
N1
)
i=r+2k
i=0
k1
i
.
(3)
Rearranging the two last lines of (3) gives
k1
1
=
_
p
k
1 +
1
(1 p
k1
)
N1
_
k
0
for k {2, . . . , r}. (4)
We show by induction that
i=r+1k
i=0
k
i
=

k
0
p
k
. (5)
Obviously, using (2), equation (5) is true for k = r. Now we suppose that (5) is true for the
values {k, . . . , r}. Using the rst line of (3) and (4) we have
i=r+2k
i=0
k1
i
= p
k
i=r+1k
i=0
k
i
+
_
1
(1 p
k1
)
N1
1
_
k
0
+
k1
0
.
Using the induction we have
i=r+2k
i=0
k1
i
=
1
(1 p
k1
)
N1
k
0
+
k1
0
.
We now replace by the second line of (3) to get:
i=r+2k
i=0
k1
i
= (1 p
k1
)
i=r+2k
i=0
k1
i
+
k1
0
.
Hence the result for k {1, . . . , r} in (5).
If we combine equation (5) and the second line of (3) we have
k
0
=
(1 p
k1
)
N
p
k1
k1
0
. (6)
Now, before evaluating the last stage of the Markov model, we get an analytical expression of
any other state
j
i
with j 1 in terms of
1
0
. The simplest expression comes from (6):
k
0
=
(1 p
1
)
N
. . . (1 p
k1
)
N
p
1
. . . p
k1
1
0
for k {2, . . . , r}. (7)
Then, using (4), one can deduce
k
1
=
_
p
k+1
1 +
1
(1 p
k
)
N1
_
(1 p
1
)
N
. . . (1 p
k
)
N
p
1
. . . p
k
1
0
for k {1, . . . , r 1}. (8)
Now, let us see a more general case, with i 2 and j 1.
j
i
=
i+j1
1
p
2
. . . p
i
using the rst line of (3)
=
_
p
i+j
1 +
1
(1p
i+j1
)
N1
_
p
2
. . . p
i
(1p
1
)
N
...(1p
i+j1
)
N
p
1
... p
i+j1

1
0
using (8), with i + j r
We then get the following formulas:
1
i
=
(1 p
1
)
N
. . . (1 p
i1
)
N
p
1
_
(1 p
i
) (1 p
i+1
)(1 p
i
)
N
_
1
0
,
for i {2, . . . , r 1},
(9)
and
j
i
=
1
p
1
(1 p
1
)
N
. . . (1 p
i+j2
)
N
p
i+1
. . . p
i+j1
_
(1 p
i+j1
) (1 p
i+j
)(1 p
i+j1
)
N
_
1
0
,
for i 2, j 2, i + j r.
(10)
Of course, the case i + j = r +1 remains. It gives:
r+1i
i
=
r
1
p
2
. . . p
i
using the rst line of (3)
=
1p
r
p
r
p
2
. . . p
i

r
0
using (2)
=
1p
r
p
r
p
2
... p
i
p
1
... p
r1
((1 p
1
)
N
. . . (1 p
r1
)
N
)
1
0
using (7).
We distinguish the cases i = r, i = r 1, and others, and we obtain
1
r
=
1 p
r
p
1
_
(1 p
1
)
N
. . . (1 p
r1
)
N
_

1
0
, (11)
2
r1
=
1 p
r
p
r
_
(1 p
1
)
N
. . . (1 p
r1
)
N
_

1
0
, (12)
r+1i
i
=
1 p
r
p
1
p
r
(1 p
1
)
N
. . . (1 p
r1
)
N
p
i+1
. . . p
r1
1
0
for i {1, . . . , r 2}. (13)
4.3 Analysis of the right-hand part
The analysis of the chain of Fig. 5 gives the following formulas:
0
i+1
= p
o
0
i
for i {k +1, . . . , m1}
0
i+1
= p
o
0
i
+ p
1
1
i
for i {1, . . . , k}
1
0
= (1 p
0
)
N
i=m
i=0
0
i
+ p
0
(1 p
0
)
N
0
m
0
1
= (1 p
0
)
_
1 (1 p
0
)
N1
_ i=m
i=0
0
i
+ p
1
1
0
+ p
0
0
0
+ p
0
_
1 (1 p
0
)
N1
_
0
m
(14)
Since there is no entering state for s
0
0
, the corresponding probability is a stationary state will
verify
0
0
= 0. Then, if we denote
0
j
=
i=m
i=j
0
i
,
we can deduce from the last line of equation (14) the following:
0
1
=
(1 p
0
)
_
1 (1 p
0
)
N1
_
0
2
+ p
1
1
0
+ p
0
_
1 (1 p
0
)
N1
_
0
m
1 (1 p
0
)
_
1 (1 p
0
)
N1
_ (15)
We can now make use of the following family of polynomials:
_
Q
1
(X) = 1 X + X
N
,
Q
p+1
(X) = Q
p
(X) (1 X)
p
(X X
N
), for p 1.
(16)
We can rewrite this formula as follows for p 1
Q
p+1
(X) = Q
1
(X) ((1 X) + + (1 X)
p
) (X X
N
),
= 1 X + X
N
(1 X)
1(1X)
p
X
(X X
N
),
= 1 X + X
N
(1 X) (1 (1 X)
p
) (1 X
N1
).
Finally it gives:
Q
p+1
(X) = X
N1
+ (1 X)
p+1
(1 X
N1
), for p 1. (17)
We then aim at proving the following formula for i {2, . . . , r}:
Q
i
(1 p
0
)
0
i
= p
i1
0
_
(1 p
0
) (1 p
0
)
N
_
0
i+1
+ p
i
0
_
1 (1 p
0
)
N1
_
0
m
+
j=i2
j=1
p
ij2
0
(1 p
0
)
N
. . . (1 p
j
)
N
(1 p
j+1
)
1
0
+p
i1
0
p
1
1
0
+ p
i2
0
(1 p
1
)Q
1
(1 p
0
)
1
0
Q
i1
(1 p
0
)(1 p
1
)
N
. . . (1 p
i1
)
N
(1 p
i
)
1
0
(18)
The case i = 2 comes from the second line of equation (14) and equations (15) and (8). The
following cases are obtained by induction using again the second line of (14) in conjunction
with (9).
Now, using the rst line of equation (14) we have
0
r+1
=
0
r+1
+
0
r+2
+ +
0
m
=
0
r+1
+ p
0
0
r+1
+ + p
mr1
0

0
r+1
=
1p
mr
0
1p
0

0
r+1
.
We then start with the second line, again, of equation (14), that iis
0
r+1
= p
0
0
r
+ p
1
1
r
, and
combining with equations (18) for i = r and (11) we get:
0
r+1
=
1 +
j=r1
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
r
0
1
0
(1 p
0
)
N1
. (19)
Naturally, it gives, using now the rst line of equation (14),
0
k
=
1 +
j=r1
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
k1
0

1
0
(1 p
0
)
N1
for k {r +1, . . . , m}.
(20)
Back-tracking further with the help of the second line of equation (14) along with equations
(11) and (8), we can write:
0
k
1
0
=
1 +
j=k2
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
k1
0
(1 p
0
)
N1
(1 p
1
)
N
. . . (1 p
k1
)
N
(1 p
k
) for k {1, . . . , r}.
(21)
We summerize all the different values of the states with respect to
1
0
in Tab. 3. We come to the
conclusion that, even if this model is still analytical, the complexity is much more important
than that of Bianchi (2000). All the formulas show the role of the barrier of value (1 p
k
)
N
to
come from rate k to rate k +1.
5. Conclusion
In this article, we have opened an approach to study the impact of mobility over WLANs.
First, after reviewing different modulation aspects, we have shown some constants that
appear in terms of the shape of the area where some rate of communication is likely to operate
Eq. State Ratio to
1
0
(7)
k
0
(1 p
1
)
N
. . . (1 p
k1
)
N
p
1
. . . p
k1
k {2, . . . , r}
(8)
k
1
_
p
k+1
1 +
1
(1 p
k
)
N1
_
(1 p
1
)
N
. . . (1 p
k
)
N
p
1
. . . p
k
k {1, . . . , r 1}
(9)
1
i
(1 p
1
)
N
. . . (1 p
i1
)
N
p
1
_
(1 p
i
) (1 p
i+1
)(1 p
i
)
N
_
i {2, . . . , r 1}
(10)
j
i
1
p
1
(1p
1
)
N
...(1p
i+j2
)
N
p
i+1
... p
i+j1
((1 p
i+j1
) (1 p
i+j
)(1 p
i+j1
)
N
)
i 2, j 2, i + j r
(11)
1
r
1 p
r
p
1
_
(1 p
1
)
N
. . . (1 p
r1
)
N
_
(12)
2
r1
1 p
r
p
r
_
(1 p
1
)
N
. . . (1 p
r1
)
N
_
(13)
r+1i
i
1 p
r
p
1
p
r
(1 p
1
)
N
. . . (1 p
r1
)
N
p
i+1
. . . p
r1
i {1, . . . , r 2}
(19)
0
r+1
1 +
j=r1
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
r
0
(1 p
0
)
N1
(20)
0
k
1 +
j=r1
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
k1
0
(1 p
0
)
N1
k {r +1, . . . , m}
(21)
0
k
1 +
j=k2
j=0
(1 p
0
)
N
. . . (1 p
j
)
N
p
j+1
0
(1 p
j+1
)
p
k1
0
(1 p
0
)
N1
(1 p
1
)
N
. . . (1 p
k1
)
N
(1 p
k
)
k {1, . . . , r}
Table 3. Stationary probabilities for the model of Fig. 5.
well. These gures have been obtained in a very general context, taking into consideration
practical and theoretical channel conditions, including Rice and Rayleigh channels.
On top of that, we have described different existing systems for adapting the rate of
communication in an unknown medium. We have shown many characteristics and also
differences of the approaches.
Finally, we could open a new approach to the difcult, and yet unanswered question of
designing a reliable analytical model to explain the behavior of such systems, taking the ARF
scheme as a model one. In that case, we were able to highlight an high correlation between the
Congestion Window (CW) of the system, and the rate at which packets are emitted. Not only
that, but the analysis showed that all the stationary probabilities of the states of this Markov
chain can be described with a closed formula. This opens new ways to research in that area
and shows that the different mechanisms that have been implemented in the MAC systems of
WLAN cards have strong correlations with one another and therefore have to be redesigned
with at least a global understanding of channel access problems (backoff and collisions) and
rate adaptation questions.
6. References
Ancillotti, E., Bruno, R. & Conti, M. (2008). Experimentation and performance evaluation
of rate adaptation algorithms in wireless mesh networks, Proc. of 5th ACM
PE-WASUN08, Vancouvert, BC, Canada, pp. 714.
Ancillotti, E., Bruno, R. & Conti, M. (2009). Design and performance evaluation of
throughput-aware rate adaptation protocols for ieee 802.11 wireless networks,
Performance Evaluation 66: 811825.
Bianchi, G. (2000). Performance analysis of the IEEE 802.11 distributed coordination function,
IEEE Journal on Selected Areas in Communications 18(8): 535547.
Biaz, S. & Wu, S. (2008). Loss differentiated rate adaptation in wireless networks, IEEE WCNC
2008, Las Vegas, NV, USA, pp. 16391644.
Chen, X., Qiao, D., Yu, J. & Choi, S. (2007). Probabilistic-based rate adaptation for IEEE 802.11
WLANs, Globecom07, Washington DC, USA, pp. 49044908.
Galtier, J. (2004). Optimizing the IEEE 802.11b performance using slow congestion window
decrease, Proccedings of the 16th ITC Specialist Seminar on performance evaluation of
wireless and mobile systems, Antwerpen, pp. 165176.
Heuse, M., Rousseau, F., Guillier, R. & Duda, A. (2005). Idle sense: An optimal access
method for high throughput and fairness in rate diverse wireless LANs, SIGCOMM,
Philadelphia, USA.
Holland, G., Vaidya, N. & Bahl, P. (2001). A rate-adaptative MAC protocol for multi-hop
wireless networks, MobiCom2001, Rome, Italy, pp. 236251.
Ibrahim, M. & Alouf, S. (2006). Design and analysis of an adaptative backoff algorithm,
Networking, pp. 184196.
J.Kim, Kim, S., Choi, S. & Qiao, D. (2006). CARA: collision-aware rate adaptation mechanism
of IEEE 802.11 WLANs, Infocom2006, Barcelon, Spain, pp. 111.
Kamermann, A. & Monteban, L. (1997). Wave LAN II: a high-performance wireless LAN for
the unlicensed band, Bell Labs Technical Journal pp. 118133.
Lacage, M., Manshaei, M. & Turletti, T. (2004). IEEE 802.11 rate adaptation: a practical
approach, MSWiM04, pp. 126134.
Ni, Q., Aad, I., Barakat, C. & Turlletti, T. (2003). Modeling and analysis of slow CW decrease
for IEEE 802.11 WLAN, PIMRC, Beijing, China.
Pang, Q., Leung, V. & Liew, S. (2005). A rate-adaptation algorithm for IEEE 802.11
WLANs based on MAC layer loss differentiation, IEEE Broadnets 2005, Boston, USA,
pp. 709717.
Part 11: wireless LAN medium access control (MAC) and physical layer (PHY) specications (1999).
IEEE Std 802.11a-1999.
Pavon, J. &Choi, S. (2003). Link adaptation strategy for IEEE wlan via received signal strength
measurement, Proc. ICC03, Vol. 2, Seattle, WA, USA, pp. 11081113.
Romano, P. (2004). The range vs. rate dilemna of (WLANs), EETimes Design.
Saghedi, B., Kanodia, V., Sabharwal, A. & Knightly, E. (2002). Opportunistic media access for
multirate Ad Hoc networks, MobiCom2002, Atlanta, Georgia, USA, pp. 2435.
Segkos, M. (2004). Advanced techniques to improve the performance of OFDM wireless LAN,
Masters thesis, Naval postgraduate school, Monterey, California.
Singh, A. & Starobinski, D. (2007). A semi markov-based analysis of rate adaptation
algorithms in wireless LANs, IEEE SECON, pp. 371380.
WLAN - 802.11 a,b,g and n (2008). NI Developper Zone.
Wong, S., Yang, H., lu, S. & Bharghavan, V. (2006). Robust rate adaptation for 802.11 wireless
networks, MobiCom2006, Los Angeles, California, USA, pp. 146157.
Wu, S. &Biaz, S. (2007). ERA: efcient rate adaptation algorithmwith fragmentation, Technical
Report CSSE07-04, Auburn University.
11
An Overview of DSA via Multi-Channel MAC
Protocols
Rodrigo Soul de Castro, Philippe Godlewski and Philippe Martins
Tlcom ParisTech, NMS research group
France
1. Introduction
The development of radio access technologies requires an increasing number of spectrum
resources. Unfortunately, spectrum bands are scarce and the development of new wireless
communication networks and services are thus more and more challenging ([jia07]). Recent
reports indicate that fixed channel allocations result in low efficiency in spectrum utilization
because a large portion of the spectrum remains underutilized ([mchenry05]).
One approach capable of dealing with the above problem is Dynamic Spectrum Access
(DSA) which allows spectrum sharing. In such an approach, unlicensed users, known as
secondary users (SUs), dynamically look for unused spectrum in licensed bands and
communicate using spectrum holes. These idle bands represent spectrum portions
assigned to licensed users (known as primary users, PUs) that are not being used at a
considered time and location ([timmers07]).
Many researchers have proposed different multi-channel MAC protocols to increase
network throughput and reduce interference caused by secondary use of the spectrum.
Many of these studies consider Wi-Fi like protocols (or IEEE 802.11 based mechanism).
Cognitive Radios (CR) are a type of radio capable of switching channels and adapting its
transmission parameters in real-time ([mitola99]). Common MAC protocols do not
provide, in general, mechanisms for channel switching. When having multiple
independent channels to be used simultaneously, the need for enhanced Multi-channel
MAC protocols becomes paramount. The IEEE 802.11 standard uses a distributed
coordination function (DCF), as the fundamental Medium Access Control (MAC)
technique. However, the distributed coordinate function, which employs carrier sense
multiple access with collision avoidance (CSMA/CA), was not designed to work in a
multi-channel environment ([ahmed07]).
Secondary users equipped with a cognitive radio, in a multi-channel environment, may
improve the efficiency of spectrum utilization and increase the network throughput.
2. Background
2.1 Secondary use of spectrum
A cognitive radio is an intelligent communication device, which has the ability to adapt its
transmission parameters such as channel frequency, modulation and power; based on the
interaction with the environment in which it operates ([jia07]).

202
There are two different approaches of secondary use of spectrum in cognitive radio context.
One is in the form of overlay, opportunistic usage of idle bands in the primary users (PU)
spectrum by cognitive radios and another in the form of underlay, using Ultra Wide Band
(UWB) technology ([cabric06]).
The rules in secondary use of frequency spectrum specify that licensed users, known as
Primary Users (PUs), have the rights for interference-free communication in certain bands.
When these bands are not used by the primary users, they can be used by Secondary Users
(SUs). As soon as a primary user starts activity in its channel, the SU has to vacate the
channel to avoid interference ([timmers07]). However, a cognitive radio (using a half duplex
transceiver) cannot scan the spectrum and transmit simultaneously in the same frequency
band. Then, for the protection of primary users, a maximum detection or sensing time must
be established. This detection time represents the maximum time of interference, from
secondary users, that a primary user can tolerate ([jia07]).
2.2 Rendezvous in multi-channel protocols
In multi-channel MAC protocols, Mobile Stations (MSs) exchange control information to
concur on the channel for data transmission in the user plane. Proposed protocols vary in
how MSs negotiate the channel to be used for data transmission and the way to solve
medium contention; these protocols can be divided according to their principle of
operation.
In single rendezvous protocols, the rendezvous between a sender and its receiver can take
place on at most one channel at any time, while in Multiple Rendezvous protocols, several
rendezvous can take place in different channels simultaneously, thereby mitigating the
control channel congestion ([mo07]).
In single rendezvous, three different classes of protocols can be distinguished based on the
mechanism of channel negotiation ([sheung07]). The Dedicated Control Channel approach,
which uses two transceivers (TRx), operates with a single channel only for control packets
exchange. In this approach, the MSs always tune one TRx to the control channel to make
agreements and be aware of neighbours negotiations. The other TRx is able to switch
channels and is used for data transmission. The Split Phase protocol uses only one TRx for
control and data packets. In this protocol, time is split into fixed periods of control and data
phases. The control phase is used as common control channel to make rendezvous, when
control phase ends, MSs switch to their selected channels and begin data transmission. The
third class of protocol is named Common Hopping, which also has only one TRx for both
control and data packets, in this protocol there is no CCCH. MSs hop synchronously
through all available channels and pauses hopping when sender and receiver agree on data
transmission using their current channel.
2.3 Hidden terminal problem in a single channel environment
Hidden terminal problem occurs when mobile stations cannot detect signal from other MSs
by carrier sensing because they do not have a physical connection to each other. Figure 2
illustrates this problem: MS A sends a message to MS B; C cannot detect the signal
from A since C is out of range of A. For station C, the channel is idle. When MS
C sends a message to B, this message will collide at B with the message sent from
A. In this scenario C is the hidden node to A.
An Overview of DSA via Multi-Channel MAC Protocols

203

Fig. 2. Hidden terminal problem in a single channel environment
2.4 Virtual carrier sensing using RTS/CTS exchange
To deal with the above problem, the IEEE 802.11 MAC layer uses the Distributed
Coordination Function (DCF) mechanism, which employs virtual carrier sensing to solve the
hidden terminal problem by using the RTS/CTS mechanism.
In this mechanism, when a mobile station wants to initiate communication, it first sends a
RTS (Request-To-Send) message and the receiver replies by sending a CTS (Clear-To-Send).
The RTS and the CTS contains the NAV (Network Allocation Vector), which is the expected
duration of time that other mobile stations, around the communication pair, must refrain
from sending data to avoid collisions.
This procedure can solve the hidden terminal problem in a single channel environment,
under the assumption that all mobile stations have the same transmission range. However,
the DCF mechanism cannot work well in a multi-channel environment, the reason is
because MSs may be transmitting or receiving data packets in different channels, missing
the RTS/CTS procedure of the DCF mechanism.
2.5 Multi-channel hidden terminal problem
This problem occurs when mobile stations in the network listen to different channels
missing the RTS/CTS procedure.
The Multi-Channel Hidden Terminal Problem is illustrated in figure 3. Initially, mobile
station A wants to communicate with B, then A sends an A-RTS to B on the
Common Control Channel (Channel 1). After receiving the A-RTS, MS B selects the Channel
2 to communicate with A and sends back an A-CTS, notifying their neighbours that the
data channel number 2 has been selected. In a single channel environment the RTS/CTS
exchange avoids collisions in the transmission ranges of A and B. However, in multi-
channel environments other mobile stations could be involved in communication in
different channels when the RTS/CTS procedure took place. That is the case of mobile
stations C and D, as they were communicating in channel 3 they did not hear the A-CTS
sent by B. When they finish their communication on Channel 3, mobile stations C and
D switch to Channel 1 and now they select Channel 2 to reinitiate communication. When
MS C sends the first message to D, this message will cause collision to mobile station
A and B on Channel 2.

204

Fig. 3. Hidden terminal problem in Multi-channel protocols (figure inspired from Jungmin
So et al. [so04])
One possible solution would be a unique channel or moment in which every MS in the
network listens to, thereby, ensuring that the RTS/CTS procedure can be heard by all the
MSs, thus avoiding the Multi-Channel Hidden Terminal Problem ([so04]).
3. Multi-channel MAC protocols
3.1 Comparison of multi channel MAC protocols [mo07]
[mo07] presents a performance comparison between different multi-channel MAC
protocols, single rendezvous protocols (dedicated control channel, common hopping and
split phase) and multi rendezvous (parallel rendezvous).
Dedicated Control Channel Approach: This protocol uses 2 TRx per Mobile Station (MS),
one is used for control information exchange and the other is able to switch between
channels for data transmission. There is no need for synchronization to make rendezvous
because the control channel is always tuned by all the MSs in the network. However, this
protocol presents two principal problems, the need for 2 TRx and the possibility of control
channel bottleneck.

205

Fig. 4. Dedicated Control Channel Approach (figure inspired from [mo07])
Common Hopping Approach: This protocol uses 1 TRx per Mobile Station (MS); this TRx
is able to switch between channels for control information exchange and data
transmission. To make rendezvous, MSs hop synchronously over all the channels and
pauses its hopping sequence when the agreement between sender and receiver is made.
This protocol uses all the channels for data transmission. However, the synchronization
among MSs is crucial.

Fig. 5. Common Hopping Approach (figure inspired from [mo07])
Split Phase Approach: This protocol uses 1 TRx per Mobile Station (MS), time is divided
into control Phase and Data phase, this division has the objective to ensure that all MSs
listen to the control phase, thus avoiding the Multi-Channel Hidden Terminal problem
(MCHTP). Two important disadvantages of this protocol are the need for global
synchronization and the wasted data channels during the control phase. However, with
only one TRx, this protocol solves the MCHTP.

Fig. 6. Split Phase Approach (figure inspired from [mo07])
3.2 McMAC: A parallel rendezvous multi-channel MAC protocol [sheung07]
McMAC protocol uses 1 TRx per Mobile Station (MS). At the beginning, a sender chooses a
hopping pattern in a pseudo-random way using a seed to generate it, neighbours learn its
hopping sequence because is included in all the senders packets. To make rendezvous, a
MS can deviate from its default hopping sequence and hops to the receivers channel. In this
protocol multiples rendezvous can be made in different channels at the same time, thus

206
improving the network throughput and avoiding control channel bottleneck. However, the
synchronization and coordination between MSs are essential.

Fig. 7. McMAC protocol (figure inspired from [mo07])
3.3 SSCH: Slotted Seeded Channel Hopping for capacity improvement in IEEE 802.11
ad-hoc wireless networks [bahl04]
SSCH protocol uses 1 TRx per Mobile Station (MS). In this protocol, each sender chooses one
of the possible hopping patterns generated in a pseudo-random way (one hopping pattern
for each available channel). To make rendezvous, a sender must wait until its current
hopping pattern intersects with that of the receiver before it can send data. The principal
disadvantage of this protocol is the time wasted waiting to coincide with the receiver.
However, multiples rendezvous can be made at the same time in different channels and the
control channel bottleneck is avoided.
3.4 Multi-channel MAC for ad hoc networks: handling multi-channel hidden terminals
using a single transceiver [so04]
In MMAC protocol, each MS is equipped with 1 TRx. Time is divided into an alternating
periods of control and data phases (split phase). An Ad Hoc Traffic Indication Message
(AR), at the start of each control interval, is used to indicate traffic and negotiate
channels for utilization during the data interval. A similar approach is used in IEEE
802.11's power saving mechanism (PSM). This scheme uses two new packets which are
not used in IEEE 802.11 PSM: the ATIM ACK (AC) and the ATIM-RES (A-RE). These
packets inform the neighbourhood nodes of the Sender (S) and Destination (D), of which
channels are going to be used during the data exchange. During the control period,
named ATIM window, all MSs have to attend the default channel and contend for the
available channels. Once reservation is successful, the MSs switch to the reserved
channel. With only one TRx this protocol solves the Multi-Channel Hidden Terminal
Problem. A Preferred Channel List (PCL) is used to select the best channel based on
traffic conditions. In this list all the channels are classified by the status: HIGH, MID, and
LOW.
The major drawback of the scheme could be the need for synchronizing beacons, which
might be difficult to implement in Ad Hoc networks and the waste of the bandwidth in
other channels during the ATIM window (control period). However, with only one TRx this
protocol solves the MCHTP.

207

Fig. 8. MMAC protocol (figure inspired from [so04])
3.5 A distributed multichannel MAC protocol for cognitive radio networks with
primary user recognition [timmers07]
In MMAC-CR protocol, time is split into alternating periods of control and data phase and
each user is equipped with 1 TRx. A similar approach is used in IEEE 802.11's power saving
mechanism (PSM). This protocol has two data structures: the Spectral Image of Primary
users (SIP), which contains the channels used by Primary Users (PUs), and the Secondary
users Channel Load (SCL), which is used to select the communication channel in terms of
traffic.

Fig. 9. MMAC-CR protocol (figure inspired from [timmers07])

208
The proposed protocol is divided into four phases: during phase I, the nodes contend to
transmit a beacon and perform a fast scan; this scanning process is used to update the SIP
value of the scanned channel. Phase II is used to determine the spectral opportunities by
listening to C minislots (each minislot correspond a data channel). Each MS informs the
others of the presence of PUs by transmitting a busy signal in the corresponding minislot. In
Phase III, using ATIM packets (AR and AC), the channels are negotiated. Phase IV is used
for data transmission or fine sensing for idle nodes.
MMAC-CR with only one TRx solves the Multi-Channel Hidden Terminal Problem.
Alternating periods of control and data phases, this protocol avoids the possibility of control
channel bottleneck. However, the synchronization and coordination between MSs are
essential to make rendezvous which might be difficult to implement in Ad hoc networks.
3.6 TMMAC: an energy efficient multi-channel MAC protocol for ad hoc networks
[zhang07]
In TMMAC, each user is equipped with 1 TRx; time is divided into control phase (ATIM
window) and data phase. The ATIM window size is not fixed and can be adapted based on
traffic conditions. The data phase is slotted, only a single data packet can be transmitted or
received during each time-slot. The purpose of the control window is twofold, the channel
negotiation and the slot negotiation. In the data phase, each node switches to the negotiated
channel and uses its respective time slot for packet transmission or reception.
This protocol has the same advantages and disadvantages presented in split phase
protocols: the need for global synchronization and the wasted data channels during the
control phase. However, with only one TRx, this protocol solves the MCHTP.
3.7 Hardware-constrained multi-channel cognitive MAC [jia07]
In HC-MAC, each MS is equipped with 1 TRx. In this protocol, there is no need for global
synchronization. To make rendezvous, HC-MAC transfers control packets using a Common
Control Channel (CCCH). Time is divided into Contention phase, Sensing phase and
Transmission phase and each phase has a RTS/CTS exchange:
1. C-RTS/C-CTS: using the RTS/CST mechanism (cf. IEEE 802.11 DCF mode), a pair of
MSs reserves all the channels (CCCH and data channels) for the following two phases
(sensing and transmission).
2. After sensing the different data channels, the pair exchanges a S-RTS/S-CTS on the
CCCH to mutually inform about channel availability. A set of channels (only one in
single Tx case) is then selected.
3. After data transmission on the different selected channels, the communication pair
informs the end of transmission by a T-RTS/T-CTS exchange. This allows neighbouring
MSs to begin the contention phase with a random back off.
Authors outline two constraints for cognitive radios, sensing and transmission, the former
used to optimize the stopping of spectrum sensing and the later used to optimize the
spectrum utilized in transmission by secondary users.
The major drawback of this scheme could be that after one communication pair wins the
CCCH, using the C-RTS/C-CTS exchange, other mobile stations must defer their sensing
and transmission. Then, for a certain time, only one pair uses all available channels and
other users must wait for the T-RTS/T-CTS notification to contend again in the control
channel.

209

Fig. 10. HC-MAC protocol (figure inspired from [jia07])
3.8 Distributed coordinated spectrum sharing MAC protocol for cognitive radio
[nan07]
This protocol uses 2 TRx per Mobile Station (MS), one is used for control information
exchange and the other is able to switch between channels for data transmission. There is no
need for synchronization to make rendezvous because the control channel is always tuned
by the MSs. In this protocol, secondary users employ a time slot mechanism for cooperative
detection of primary users around the communication pair by using the CHRPT (channel
report slots). Each node informs the others about the presence of PUs, in the sender and in
the receiver side, by transmitting a busy signal in the corresponding minislot (there is one
minislot for each data channel).

Fig. 11. Procedure of the proposed protocol (figure inspired from [nan07])
The source sends to destination the RTS which includes its available channel list. Neighbour
nodes, which hear the RTS, compare the sender list with their own; if they detect a PU

210
occupation in a channel, they reply with a pulse in the specified time slot during CHRPT
(signalling occupied channels seen by the neighbours). If necessary, the source update its
RTS sending a RTSu. The same mechanism occurs in the destination side. After the RTS
reception the destination waits to get the possible RTSu for certain time named UIFS, if the
RTSu does not arrives, the destination will handle the first RTS. After the RTS reception, the
destination sends to its neighbours the Channel Status Request (CHREQ), which includes
the destination available channel list among the listed channels of the source. At the end of
channel verification by the destination neighbours, the receiver sends the CTS with the
chosen channel.
The major drawbacks of the scheme are the time wasted in channel verification by the
neighbours and the need for two TRx. However, this procedure ensures the absence of
primary users in the vicinity of the communication pair.
3.9 Performance of multi channel MAC incorporating opportunistic cooperative
diversity [ahmed07]
In CD-MMAC, time is divided into fixed periods (split phase), each user is equipped with 1
TRx. This protocol uses the same mechanism proposed by So et al. in MMAC ([so04]). The
authors of this protocol add the notion of relays between source and destination. Time is
divided into fixed-time intervals (control phase and data phase) using beacons, a small
window, named ATIM, at the start of each interval is used to indicate traffic and negotiate
channels to be used during the data phase. This protocol uses intermediate nodes as relays
to increase the probability of transmission success.
This protocol solves the MCHTP with only one TRx. However, two drawbacks of CD-
MMAC are the need for global synchronization and the wasted data channels during the
control phase.
3.10 A full duplex multi channel MAC protocol for multi-hop cognitive radio
networks [choi06]
In this protocol, each secondary user is equipped with 3 TRx named: Receiver, Transmitter
and Controller. To communicate, the RECEIVER of the receiving node and the
TRANSMITTER of the sending node must be tuned to the same channel.
There is no need for synchronization because the CCCH is always tuned by the MSs using
the CONTROLLER. A MS selects an unused frequency band as its home channel (HCh), it
tunes its receiver to its HCh and informs the others about its selected channel by
broadcast in the control channel. This protocol uses CSMA/CA scheme of IEEE 802.11
DCF mode. With the use of three TRx, MSs can reduce communication delay by
transmitting packets while they are receiving. However, the need for 3TRx will increase
the overall cost.
3.11 A multi channel MAC for opportunistic spectrum sharing in cognitive networks
[mishra06]
In AS-MAC (Ad hoc SEC Medium Access Control) protocol, the primary user is a
TDMA/FDMA (GSM) cellular network and the secondary user is an Ad hoc network that
can decode the control information of GSM system. Sensing the vacant slots, the SU uses the
resources left utilized by the primary user, which could be a Base Station (BS) or a Mobile
Station (MS). To obtain all the parameters like synchronization, frequency correction and

211
cell information, secondary users decode the beacon channel from the BS. To make
rendezvous, this protocol employs RTS/CTS and Reservation (RES) mechanism.
3.12 Performance evaluation of a medium access control protocol for IEEE 802.11s
mesh networks [benveniste06]
CCC protocol uses 2 TRx per Mobile Station (MS), one is used for control information
exchange and the other is able to switch between channels for data transmission. There is no
need for global synchronization to make rendezvous because the control channel is always
tuned by the MSs. The CCC protocol defines a Common Control Channel (CCCH), over
which, mesh nodes will exchange control and management frames, the rest of the channels,
called Mesh Traffic (MT) channels, are used to carry the data traffic. Reservations of the
various MT channels are made by exchanging control frames on the CCCH.
This protocol has the same advantages and disadvantages presented by the dedicated
control channel approach: there is no need for synchronization to make rendezvous.
However, this protocol needs two TRx and the possibility of control channel bottleneck
exists.

Fig. 12. CCC MAC protocol (figure inspired from [benveniste06])
4. Os-MAC: an efficient MAC protocol for spectrum-agile wireless networks
[hamdaoui08]
In Os-MAC protocol, each secondary user is equipped with 1 TRx; this protocol uses the
IEEE 802.11 DCF mode. This approach seeks to exploit the available spectrum opportunities
using MSs coordination. One entity per channel is a "delegate", the delegates are chosen
among MSs and makes reports about channel quality. A single ACK notion is used in a
"multicast group" named Secondary User Group (SUG).
OS-MAC divides time into periods; each period is named Opportunistic Spectrum Period
(OSP). In each OSP, there exist three consecutive phases: Select, Delegate, and Update Phase.
In the first phase, each SUG selects the best Data Channel (DC) based on traffic conditions

212
and uses it for communication during the totality of the OSP period. During the second
phase, a Delegate Secondary User (DSU) is chosen to represent the data channel during
the Update Phase, in which, all DSUs switch to the CCCH to update each other about
their channel conditions, mean while, all non-DSUs continue communicating on their
DCs.
An important aspect of this protocol is the notion of groups and the Delegate for each DC.
This mechanism can improve the channel classification necessary to define the best channel,
based in traffic conditions, which could be used for data transmission.
4.1 Primary Channel Assignment Based MAC (PCAM) a multi channel MAC protocol
for multi-hop wireless networks [pathma04]
In PCAM protocol, each user is equipped with 3 TRx. This scheme eliminates the need for a
dedicated control channel that arise the possibility of control channel bottleneck when the
traffic increases. In this protocol, a MS selects a frequency band as its primary channel, this
will be used as a receiver channel and a secondary channel is used as transmitter while the
third TRx is used for transmitting and receiving broadcast messages. PCAM protocol
removes the constraints of time synchronization and control channel saturation because the
channels are pre-assigned. However, the need for 3 TRx will increase the overall cost and
the channel assignment procedure, in this protocol, is not specified.
4.2 Adaptive MAC protocol for throughput enhancement in cognitive radio networks
[lee08]
In this protocol, each user is equipped with 2 TRx, this protocol proposes two channels, the
first one is a WLAN channel which is always available for data transmission; the second
one, named Cognitive channel, is available sporadically. When traffic conditions restrain
the use of the cognitive channel, this channel is used for frame errors recovery by
transmitting the same information in both channels, known as frequency diversity in MIMO
systems; otherwise, the cognitive channel can be used to increase the overall throughput by
sending sequential frames using both channels.
The drawback of this scheme could be the need for two TRx. However, this procedure can
enhance the overall throughput if the Cognitive channel is available.
4.3 CREAM-MAC: An efficient Cognitive Radio-EnAbled Multi-channel MAC protocol
for wireless networks [su08]
In the Cognitive Radio-EnAbled Multi-channel MAC (CREAM-MAC) protocol, each
secondary user is equipped with 1 TRx that can dynamically utilize one or multiple
channels to communicate and also has multiple sensors that can detect multiple channels
activity simultaneously. This protocol needs neither centralized controllers nor
synchronization.
The CREAM-MAC protocol employs a Common Control Channel (CCCH) as the
rendezvous channel. With one TRx, this protocol solves the Multi-Channel Hidden
Terminal Problem employing a four-way handshake. These control packets are RTS/CTS
and CST/CSR, the RTS/CTS exchange prevents the collisions among the secondary users by
reserving the CCCH for channel negotiation. The CST/CSR exchange avoids collisions
between secondary and the primary users by allowing secondary users to share sensing
information about PUs channel occupation.

213

Fig. 13. CREAM-MAC protocol (figure inspired from [su08])
The merit of the CREAM-MAC protocol is the fact that there is no need for global
synchronization and with the use of only one TRx and multiple sensors, this protocol solves
the MCHTP.
4.4 Distributed coordination in dynamic spectrum allocation networks [zhao05]
In this paper, the notion of groups with similar views of spectrum availability is addressed.
Each secondary user is equipped with 1 TRx, this protocol employs a voting scheme for
selection of a Coordination Channel (CCH) for a group and this user group is
assembled based in similar spectrum channel availabilities.
The CCH is used as the only means to connect secondary users, thus, only members of the
same group can directly communicate with each other. To maintain network connectivity
bridge nodes, located on the edge of each group, must manage at least two different CCH
to transfer data packets between groups and connect users with different spectrum
perspective.
The advantage of this approach is its possible application in the case of secondary use of the
spectrum by WLAN devices in TV white spaces, principally, because the interference
condition with primary users is determined by distance.
4.5 Single-Radio Adaptive Channel Algorithm for spectrum agile wireless ad hoc
networks [ma07]
In the Single-Radio Adaptive Channel (SRAC) algorithm, each secondary user is equipped
with 1 TRx. This algorithm proposes an adaptive channelization, where a radio combines
multiple fixed channels with minimum bandwidth, named atomic channels, based on its
needs to form a new channel with more bandwidth, thus forming a Composite channel. In
this algorithm there is no need for global synchronization. SRAC also proposes Cross-
channel communication, utilized to enable communications when there are multiple
jamming sources and there is no common idle spectrum between the transmitter and the
receiver. A node always has a pre-assigned channel for reception, which is well known by
its neighbours and will be used to reach that node; this channel can be modified but the
selection must follows strict rules to enable future communications.
The merits of this algorithm are the adaptive channelization and the fact that it does need
neither CCCH nor synchronization because the MSs have a pre-assigned channel for
reception.

214
4.6 Cognitive radio system using IEEE 802.11a over UHF TVWS [ahuja08]
This paper presents a practical implementation of IEEE 802.22 WLAN/TV with Primary and
Secondary users. The architecture consists of Cognitive Mobile Stations (CMS) and a
Cognitive 802.11 Access Point (CAP), which performs band sensing and available channel
determination. The Cognitive Access Point has 1 TRx and 1 Rx for sensing, the Cognitive
Mobile Stations are equipped with only 1 TRx. There is no CCCH, the CAP sends a
broadcast message to inform all stations about the available channels list and time
synchronization. A Geo-location module is used to guarantees that the cognitive radio units
will never transmit on a channel that is determined to be within a licensed stations
protected contour.
4.7 Spectrum sharing radios [cabric06]
This paper proposes the utilization of overlay, opportunistic usage of idle bands, for data
transmission and underlay, using UWB technology, for control messages exchange. In this
approach, authors propose two different types of control channels. The first one is a low
throughput and wide coverage channel, named Universal Control Channel (UCC), which
is used as a CCCH allowing the co-existence of several Radio Access Technologies (RATs).
The second type of channel, named Group Control Channel (GCC), works as Group
Coordination Channel. This channel with high throughput and short coverage allows
sensing information exchange, link maintenance and performs channel allocation.
The advantage of the use of UWB Control Channels is that we could have a realistic and
reliable Cognitive Control Channel, always free of Primary users, which is one of the
principal assumptions in several propositions of Multi-Channel MAC protocols.
5. Conclusions
This chapter presents the main existing multi channel MAC protocols. The merits of several
protocols are discussed with regard to different factors: the number of transceivers, the need
for synchronization, the need for a common control channel (CCCH) and the different ways
to make rendezvous for data transmission. As we showed, each multi-channel MAC
protocol faces and resolves differently the various complications that arise in dynamic
spectrum access.
In short, Cognitive Radio (CR) technology offers the possibility for additional use of radio
spectrum by secondary users. Multiple channel protocols allow dynamic spectrum access
(DSA) due to the fact that different rendezvous and data transmissions can be performed on
different channels. This type of protocols, compared to others that use a single frequency
channel (IEEE 802.11mechanism), may improve spectrum utilization and increase total
network throughput.
6. Acronyms
ATIM: Ad hoc Traffic Indication Message
AC: ATIM ACK
AR: ATIM
A-CTS: ATIM CTS (which includes the data channel selection)
A-RE: ATIM Reservation
A-RTS: ATIM RTS (which includes the data channel selection)

215
CCCH: Common Control Channel
CR: Cognitive Radio
CREAM-MAC: Cognitive Radio-EnAbled Multi-Channel MAC protocol proposed in [su08]
CSR: Channel-State-Receiver
CST: Channel-State-Transmitter
DC: Data Channel
DCF: Distributed Coordination Function {IEEE 802.11}
DSU: Delegate Secondary User
MCHTP: Multi-Channel Hidden Terminal Problem
MC-MAC: Multi-Channel (wireless) MAC
MMAC: Multi-Channel MAC protocol proposed in [so04]
MMAC-CR: Multi-Channel MAC protocol proposed in [timmers07]
MS: Mobile Station
OSMAC: Opportunistic Spectrum Media Access Control proposed in [hamdaoui08]
OSP: Opportunistic Spectrum Period
PN: Primary Network
PSM: Power Saving Mechanism
PCL: Preferred Channel List
PU: Primary User
RAT: Radio Access Technology
SCL: Secondary users Channel Load
SIP: Spectral Image of Primary users
SU: Secondary User
SUG: SU Group
TRx: Transceiver
7. References
[ahmed07] Sabbir Ahmed, Christian Ibars, Aitor del Coso and Abbas Mohammed,
Performance of Multi Channel MAC incorporating Opportunistic Cooperative
Diversity. in IEEE Vehicular Technology Conference, April 2007.
[ahuja08] Ramandeep Ahuja, Robert Corke and Alan Bok, Cognitive Radio System using
IEEE 802.11a over UHF TVWS. New Frontiers in Dynamic Spectrum Access Networks,
2008.
[bahl04] Paramvir Bahl, Ranveer Chandra and John Dunagan, SSCH: Slotted Seeded
Channel Hopping for Capacity improvement in IEEE 802.11 Ad-Hoc Wireless
Networks. in MobiCom 2004.
[benveniste06] Mathilde Benveniste and Zhifeng Tao, Performance Evaluation of a Medium
Access Control Protocol for IEEE 802.11s Mesh Networks, in Sarnoff Symposium,
2006.
[cabric06] Danijela Cabric, Ian D. ODonnell, Mike Shuo-Wei Chen, and Robert W.
Brodersen, Spectrum Sharing Radios, in IEEE Circuits and Systems Magazine, 2006.
[choi06] Noun Choi, Maulin Patel and S.Venkatesan, A Full Duplex Multi channel MAC
Protocol for Multi Hop Cognitive Radio Networks, in Cognitive Radio Oriented
Wireless Networks and Communications Conference, June 2006.
[hamdaoui08] Bechir Hamdaoui and Kang G. Shin, Os-MAC: An efficient MAC Protocol for
Spectrum-Agile Wireless Network, in IEEE Transactions on Mobile Computing 2008.

216
[jia07] Juncheng Jia and Qian Zhang, Hardware-constrained Multi-Channel Cognitive MAC.
in IEEE Global Telecommunications Conference, November 2007.
[lee08] Byungjoo Lee and Seung Hyong Rhee, Adaptive MAC Protocol for Throughput
Enhancement in Cognitive Radio Networks, in: Information Networking, 2008.
[mishra06] Amitabh Mishra, A Multi channel MAC for Opportunistic Spectrum Sharing in
Cognitive Networks, in Military Communications Conference, 2006.
[mitola99] J. Mitola III, Cognitive radio for flexible mobile multimedia communication, in:
Proc. IEEE International Workshop on Mobile Multimedia Communications (MoMuC)
1999, November 1999, pp. 310.
[ma07] Liangping Ma, Chien-Chung Shen, and Bo Ryu, Single-Radio Adaptive Channel
Algorithm for Spectrum Agile Wireless Ad Hoc Networks, in: New Frontiers in
Dynamic Spectrum Access Networks, 2007.
[mchenry05] Mark A; McHenry, NSF, Spectrum Occupancy Measurements, Project
Summary, http://www.sharedspectrum.com/ , August 2005
[mo07] Jeonghoon Mo, Hoi-Sheung Wilson So and Jean Walrand, Comparison of Multi
channel MAC protocols, in IEEE Transactions on Mobile Computing 2007.
[nan07] Hao Nan, Tae-In Hyon and Sang-Jo Yoo, Distributed Coordinated Spectrum Sharing
MAC protocol for cognitive radio, 2007.
[pathma04] Jaya Shankar Pathmasuntharam, Amitabha Das and Anil Kumar Gupta,
Primary Channel Assignment Based MAC (PCAM) A Multi Channel MAC Protocol
for Multi-Hop Wireless Networks, in Wireless Communications and Networking
Conference, 2004.
[sahin07] Mustafa E. Sahin, Sadia Ahmed, and Hseyin Arslan. The Roles of Ultra
Wideband in Cognitive Networks, in IEEE International Conference on Ultra-
Wideband, 2007.
[sheung07] Hoi-Sheung Wilson So, Jean Walrand and Jeonghoon Mo, McMAC: A Parallel
Rendezvous Multi-Channel MAC protocol. in IEEE Wireless Communications and
Networking Conference, March 2007.
[so04] Juming So and Nitin Vaidya, Multi-Channel MAC for Ad Hoc Networks: Handling
Multi-Channel Hidden Terminals Using A Single Transceiver, in Proceedings of
AMC MobiHoc, May 2004.
[su08] Hang Su and Xi Zhang, CREAM-MAC: An efficient Cognitive Radio- EnAbled Multi-
Channel MAC Protocol for Wireless Networks. in: International Symposium on a
World of Wireless, Mobile and Multimedia Networks, 2008.
[timmers07] Michael Timmers, Antoine Dejonghe, Liesbet Van der Perre and Francky
Catthoor, A Distributed Multichannel MAC Protocol for Cognitive Radio Networks
with Primary User Recognition, in Cognitive Radio Oriented Wireless Networks and
Communications, Aug. 2007.
[zhang07] Jingbin Zhang, Gang Zhou, Chengdu Huang, Sang H. Son and John A. Stankovic,
TMMAC: An Energy Efficient Multi-Channel MAC Protocol for Ad Hoc Networks,
in the ICC 2007 proceedings
[zhao05] Jun Zhao, Haito Zheng and Guang-Hua Yang, Distributed Coordination in
Dynamic Spectrum Allocation Networks, in New Frontiers in Dynamic Spectrum
Access Networks, 2005.
Alfonso Bahillo, Patricia Fernndez, Javier Prieto, Santiago Mazuelas,
Rubn M. Lorenzo and Evaristo J. Abril
University of Valladolid
Spain
1. Introduction
Intense research work is being carried out to design and build localization schemes that
can operate in indoor environments where satellite signals typically fail. The objective is
to achieve a degree of accuracy, reliability and cost in indoor environments comparable
to the well-known Global Navigation Satellite Systems (GNSS) in open areas. These
challenging problems are being faced today to fulll commercial, public safety and
military applications (Gustafsson & Gunnarson, 2005; Pahlavan & Krishnamurthy, 2002). In
commercial applications for residential and nursing homes there is an increasing need to track
people with special needs, such as children and elderly people who are out of regular visual
supervision, navigate the blind, and nd specic items in warehouses. For public safety and
military applications, indoor localization schemes are needed to track inmates in prisons or
navigate police ofcers, re ghters and soldiers to complete their missions inside buildings.
Among the many indoor technological possibilities that have been considered for indoor
localization such as infrared, ultrasonic and articial vision, radiofrequency based schemes
predominate today due to their availability, low-cost and coverage range. Currently, few
radiofrequency infrastructures that operate inside buildings are as extensively deployed and
used as 802.11. Nowadays, many buildings such as shopping malls, museums, hospitals,
airports, etc. are equipped with 802.11 access points (APs). Therefore, it may be practical to
use these APs to determine user location in these indoor environments.
Whichever indoor wireless technology is involved, the purpose of localization schemes is
to nd the unknown position of a mobile station (MS) given a set of measurements called
localization metrics. These metrics could be the measured time-of-arrival (TOA) (Golden &
Bateman, 2007), angle-of-arrival (AOA) (Seow & Tan, 2008) or received-signal-strength (RSS)
(Mazuelas et al., 2009) of the MSs signal at the reference devices or APs. Techniques based on
RSS require channel modeling and they are not exible because they present high variability
to environmental changes; even though building and updating a RSS database is much easier
in indoor environments than in wide urban areas. The major drawback of pattern recognition
techniques still lies in substantial efforts needed in generation and maintenance of the RSS
database in view of the fact that the working environment changes constantly. Techniques
based on TOA need time synchronization between wireless nodes; and techniques based
on AOA require specialized antennas. Furthermore, it is important to point out that as the
Distance Estimation based on 802.11 RTS/CTS
Mechanism for Indoor Localization
12
measurements of metrics become less reliable, the complexity of the positioning algorithm
increases.
In this chapter, the performance of the 802.11 wireless networks for indoor localization is
based on the time delay localization metric through round-trip time (RTT) measurements. The
challenge is to develop an infrastructure that is inexpensive to design and deploy, complies
with frequency regulations, and provides a comprehensive coverage for accurate ranging. RTT
is used instead of TOA to avoid the need for time synchronization between wireless nodes.
Furthermore, due to the use of an 802.11 infrastructure, the location capabilities will be an
added value to the existing connectivity ones. The main characteristic that makes the RTT
measurements possible in any 802.11 wireless network is the common protection mechanisms
to fully reserve a shared medium, Request To Send/Clear To Send (RTS/CTS) handshake
(Bahillo et al., 2009). Therefore, the RTT measurements are obtained by measuring the latency
of a series of layer two CTS frames sent by and in response to a corresponding series of RTS
frames initiated by the MS that is going to be located. The measuring systemis integrated in a
Printed Circuit Board (PCB), which is used as additional hardware to the 802.11 adapter from
which appropriate signals, such as transmission and receiver pulses of exchange frames, are
extracted to quantify the RTT.
The results of RTT measurements in different scenarios are qualitatively consistent because, as
it was expected, the delay prole observed shifts as the actual distance between wireless nodes
in line-of-sight (LOS) increases following a linear shape. The coefcient of determination is
used to measure how much of the original uncertainty in the RTT measurements is explained
by the linear model.
Unfortunately, the assumption that a direct sight exists between two wireless nodes in an
indoor environment is an oversimplication of reality, where the obstacles usually block the
direct path. Known as non-line-of-sight (NLOS), several techniques have emerge to overcome
this problem. They can be broadly classied in two groups, techniques which attempt to
minimize the contribution of NLOS multipaths (Chen, 1999) or techniques which focus on
the identication of NLOS reference devices and discard them for localization (Cong &
Zhuang, 2005). However, their reliability remains questionable in an indoor environment
with abundant scatterers where almost all reference devices will be in NLOS. In this chapter
the PNMC (Prior NLOS Measurements Correction) technique is used to correct the NLOS
effect from distance estimates (Mazuelas et al., 2008). This technique manages to introduce
the information that actually resides in the NLOS measurements in the localization process.
The chapter is organized as follows. Section 2 presents a method to quantify the time delay
between two wireless nodes and proposes a PCB as a measuring system. Section 3 analyzes
the best statistical estimator of the time delay assuming a linear regressionmodel to relate that
estimator with the actual distance between two wireless nodes in LOS. Section 4 describes the
mitigation of the severe NLOS effect on those distance estimates using the PNMC method.
Section 5 evaluates the performance of the distance estimation technique in a rich multipath
indoor environment, and Section 6 summarizes the main achievements.
2. Time delay quantication
The TOA-based systems measure distance based on an estimate of signal propagation delay
between a transmitter and a receiver since, in free space or air, radio signals travel at the
constant speed of light. The TOA can be measured by either measuring the phase of received
narrowband carrier signal or directly measuring the arrival time of a wideband narrow
pulse. However, the challenge for this chapter is to develop a distance estimation system
that is inexpensive to design and deploy, complies with 802.11 regulations, and provides a
comprehensive coverage for accurate ranging.
That is why in this chapter the performance of the 802.11 wireless networks for indoor
localization is based on the time delay localization metric through RTT measurements. By
using RTT the need for time synchronization between wireless nodes is avoided which would
entail a major increase in the complexity of the location scheme development. Furthermore,
due to the use of an 802.11 infrastructure, the location capabilities will be an added value to
the existing connectivity ones.
The main characteristic that makes the RTT measurements possible in a 802.11 wireless
network is the common protection mechanism to reserve a shared medium, RTS/CTS
handshake (Gast, 2002).
Mobile
Station
RTS
CTS
Frame
ACK
Access
Point
RTT
Fig. 1. RTS/CTS handshake.
In wireless communication networks, medium access control (MAC) schemes are used to
manage all nodes access to the shared wireless medium. Due to the randomness of packet
arrivals and local competition, it is difcult to completely eliminate packet collisions. Since
data packet collisions are costly, researchers proposed to use the RTS/CTS dialogue to reserve
the right to channel usage. Assuming a ready node has a frame to send, if it has the RTS/CTS
technique activated (see Fig. 1), it initiates the process by sending an RTS frame. The RTS frame
serves several purposes; in addition to reserving the radio link for transmission, it silences any
station that hear it. If the target station receives an RTS, it responds with a CTS. Like the RTS
frame, the CTS frame silences stations in the immediate vicinity. Once RTS/CTS exchange is
complete, the mobile station can transmit its frames without worry of interference from any
hidden nodes. Hidden nodes beyond the range of the sending station are silenced by the CTS
from the receiver. With the use of the RTS/CTS dialogue, it is less likely that data packets will
suffer collisions.
Therefore, the RTS/CTS handshake is used to quantify the RTT by measuring the latency of a
series of layer two CTS frames sent by and in response to a corresponding series of RTS frames
initiated by the MS that is going to be located. The same as acknowledgement (ACK), CTS are
considered in the AP the highest priority frames, therefore, the minimum elapsed time in the
AP is guarantee when processing these sort of frames.
219 Distance Estimation based on 802.11 RTS/CTS Mechanism for Indoor Localization
2.1 Printed circuit board design
In order to quantify the RTT of the RTS/CTS two-frame exchange 802.11 mechanism,
appropriate signals from within the WLAN adapter chip set must be selected and accessed
by the measuring system. The aim is to extract both transmission pulses and receiver signals
in such a way that the RTS frame can be used as the trigger to start the measuring system that
would be stopped by the corresponding CTS frame.
(a) Tx port timing
(b) Rx port timing
Fig. 2. Timing behavior of TX_RDY and MD_RDY signals.
If we access the physical layer of the WLAN adapter, we lose the control of what instant
corresponds with which frame sent or received. That is way we access the interface between
the physical and MAC layers of the WLAN adapter. This way, it will be easy to associate
transmission times and reception times with sent and received frames, respectively. Among
the commercial chip sets that work as interface between the physical and MAC layers, the
Intersil HFA3861B baseband processor is fully free documented (Intersil, 2002).
MCLK
TX_RDY
MD_RDY
RTT
COUNT
RTS frame
departure
CTS frame
arrival
(a) RTT
MCLK
TX_RDY
MD_RDY
AP processing
time
COUNT
RTS frame
arrival
CTS frame
departure
(b) AP processing time
Fig. 3. Timing diagram to mesure the RTT and the processing time of the AP.
From an inspection of the Intersil HFA3861B component pinout diagram, three appropriate
leads (and common ground) were identied, TX_RDY, MD_RDY and MCLK (Intersil, 2002).
TX_RDY is an output of the external network processor indicating that preamble and header
information has been generated and that the HFA3861B is ready to receive the data packet
from the network processor over the TXD serial bus (see Fig. 2(a)). MD_RDY is an output
signal of the network processor, indicating that header data and a data packet are ready
to be transferred to the processor (see Fig. 2(b)). MCLK is the 44 MHz master clock that
governs MAC layer processing. In Fig. 2(a) and 2(b) TXCLK and RXCLK clocks govern
the signals TX_RDY and MD_RDY, respectively. These signals, TXCLK and RXCLK, are
generated through the master clock MCLK. Therefore, as TX_RDY and MD_RDY signals are
synchronized with TXCLK and RXCLK, respectively, they will be also synchronized with the
master clock, MCLK.
Thus, in case the RTT is measured, the falling edge of the TX_RDY signal is used to start the
counter (RTS frame departure) and the rising edge of the MD_RDY signal is used to stop it
(CTS frame arrival). If the processing time of the AP is wanted to be measured, the rising edge
of the MD_RDY signal is used to start the counter (RTS frame arrival) and the falling edge of
the TX_RDY signal is used to stop it (CTS frame departure).
To quantify the RTT a 16-bit counter is used as measuring system and its input is the
aforementioned MCLK lead, which allows to measure times up to 1.489 ms (2
16
MCLK cycles).
The triggers of the counter will be the TX_RDY and MD_RDY leads. As these triggers are
MCLK synchronized, the count accuracy will not improve although a higher clock frequency
was used.
The counter is integrated in a PCB (see Fig. 4). The PCB is made up of four serial 4-bit counter
resulting in a 16-bit counter; one Flip-Flopand one XORgate to formthe RTT signal, managing
the TX_RDY and MD_RDY triggers so that the RTS and CTS frames can be used as triggers
to start and stop the measuring system; three 4-bit multiplexers to read the state of the four
4-bit counters; and the corresponding bypass capacitors between power supply and common
ground to speed up the PCB commutation times, and to form a low pass lter which will
prevent from high frequency disruptions. With the aim of reducing the loop area both for the
supply and the signal tracks, top and bottom planes of the PCB were poured of copper, one
attached to the power supply lead and the other attached to the common ground. This will
minimize the impedance of the return path. In order to control the measuring system, the PCB
is governed by the MS through the universal serial bus (USB) port.
The PCB measures the RTT as follows:
1. The MS enables the counters prior to send the RTS frame.
2. The last bit of the sending RTS frame starts the counters.
3. The rst bit of the receiving CTS frame stops the counters.
4. Once the RTS/CTS two-frame exchange is completed the MS disables the counters.
5. The MS saves the state of the four 4-bit counters through the multiplexers.
The measuring system proposed has some limitations. First of all, as the MCLK that governs
the PCB is 44 MHz frequency, the 16-bit counter implemented on the PCB cannot measure
RTTs over 1.489 ms, but this time is enough for wireless networks range. Secondly, as a frame
coming from other wireless nodes could activate or deactivate the count within the short
lapse of time in which the measuring system is enabled, a lter that rejects these undesirable
measurements is implemented. Filter limits have been chosen based on previous trials where
there were no other wireless nodes interfering. Finally, according to (Bahillo et al., 2009) the
elapsed time in the AP, between receiving a RTS frame and sending the corresponding CTS
frame, can be assumed to be constant when there are no other processes competing for the AP
resources. Obviously, although the CTS frame has the highest priority (Gast, 2002), it could be
concurrent RTS frames coming fromother MS at the same AP increasing the load of the AP. In
that case, if there are not enough APs in range to apply the localization algorithm, the wireless
localization system delay increases, but the accuracy is not degraded thanks to the previous
lter that rejects the RTT measurements that are out of the expected range.
(a) Top layer (b) Bottom layer
(c) Layer connections
U
S
B

I
N
T
E
R
F
A
C
E
F
I
L
P
-
F
L
O
P

a
n
d

X
O
R
MULTIPLEXERS
TX_RDY, MD_RDY, MCLK COUNTERS
C
A
P
A
C
I
T
O
R
S
(d) Printed circuit board
Fig. 4. Printed Circuit Board, 6974 [mm
2
] size, used to measure the RTS/CTS two-frame
exchange.
Regardless of the PCB size, the core of the measuring system is the counter, because the
other components have to control the measuring system. If the measuring system would be
integrated in the WLAN adapter, only the counter component and the driver to control it
would be needed.
2.2 Experimental validation
After using the circuit simulation software Pspice to check that the PCB design works as
expected, it is necessary to check that the PCB connected to the WLAN adapter and managed
by the MS is able to measure the time delay when interchanging RTS/CTS frames.
The wireless devices involved in the experimental validation can be found in most wireless
networks. They are:
1. A MS, an 802.11b wireless cardbus adapter, specically a Cisco Aironet AIR-PCM340
with the HFA3861B baseband processor. The wireless adapter has been connected to the
computer through a cardbus extender to be able to access to the HFA3861B pinout. This
wireless adapter includes two on-board patch antennas with a diversity switch which
toggles to and from, and stops when a signicant amount of radio frequency power is
detected.
2. An AP, a Linksys WRT54GL 802.11b/g. This AP includes two rubber duck omnidirectional
antennas in diversity mode that never work at the same time, since diversity circuitry
switches to the one with better reception. Rubber duck antennas provide vertical
polarization with 360 degrees of coverage in the horizontal plane and 75 degrees in the
vertical one. The AP was congured to send a beacon frame each 100 ms at constant power
on 802.11 frequency channel 1 (2.412 GHz).
Using these wireless devices, several measurement campaigns were carried out in two
different scenarios: an esplanade with a few streetlamps and trees, in the following exterior;
and the corridor of a building 50 4.3 3.5 m
3
(length, width and hight) size with wooden
and metal doors and a few people walking around, in the following corridor. In all scenarios,
the two WLAN devices which were involved in the scheme of the experimental setup were
always in line-of-sight on a cardboard box 1.5 m high each, in order to guarantee the rst
Fresnel zone clearance.
Flying time
processing
Mobile
Station
Access
Point
RTS
CTS
RTT
Flying time
(a) RTT
Flying time
AP
processing
time
Mobile
Station
RTS
CTS
Flying time
Last bit
departure
First bit
arrival
Access
Point
PCB
10 m
(b) AP processing time
Fig. 5. Experimental setup.
As shown in Fig. 5(a), in case the RTT is measured, the last bit of the RTS frame departure
is used to start the counter and the rst bit of the CTS frame arrival is used to stop it. If the
processing time of the AP is wanted to be measured (see Fig. 5(b)), the rst bit of the RTS
frame arrival is used to start the counter and the last bit of the CTS frame departure is used to
stop it.
2.2.1 RTT measurements
RTT measurements were performed using one Cisco WLAN adapter acting as a MS and one
Linksys WRT54GL acting as an AP. The PCB was connected to the WLAN adapter of the
MS. Three campaigns of 5000 RTT measurements were conducted at each position for several
distances from 0 to 40 m.
Fig. 6 shows the RTT measurements obtained in terms of the number of MCLK cycles elapsed
at each distance and environment, exterior and corridor. Each mark in Fig. 6 represents the
mean of each group of 150 RTT measurements. The result is qualitatively consistent because,
as it was expected, the delay prole observed shifts to the right as the actual distance between
the two WLAN nodes increases, and it follows a linear form.
As Fig. 6 shows, although a direct path exists between the MS and the AP in all scenarios and
distances, the delay prole observed is spread around 4 MCLK cycles. Besides the random
behavior of the electronics, there are two main reasons: rst of all, the frequency clock that
governs the MS and the AP is 44 MHz and 20 MHz, respectively. Secondly, because of the
multipath and the scatters of the environment, the direct path is not always the one selected by
WLAN adapters. Therefore, to estimate the distance between the MS and the AP, a statistical
estimator of the delay prole observed has to be selected. It will be discussed in the following
section.
6999 6992 6994 6996 6998 7999 7992 7994 7996 7998
9
9
19
19
29
29
39
39
49
Nmero de ciclos de MCLK
D
i
s
t
a
n
c
i
a

[
m
]
Exterior
Number of MCLK cycles
D
i
s
t
a
n
c
e
(a) Exterior
6999 6999 7999 7999 7919 7919
9
9
19
19
29
29
39
39
49
D
i
s
t
a
n
c
i
a

[
m
]
Pasillo
D
i
s
t
a
n
c
e
Corridor
(b) Corridor
Fig. 6. RTT measurements between two WLAN nodes in LOS at different distances in two
different scenarios, exterior and corridor.
2.2.2 AP processing time measurements
The AP processing time is measured in the two scenarios described above (exterior and
corridor) in order to check that the AP processing time is constant when the RTS/CTS
two frame exchange is performed. As the WLAN MAC chip set of the Linksys WRT54GL
802.11b/g is not for public access, in this section the Cisco AIR-PCM340 WLAN adapter is
used in AP mode. Therefore, a testing for this AP processing time approach was performed
using two Cisco WLAN adapters, one acting as a MS and the other acting as an AP. The
PCB was in the AP. By using a driver designed by ourselves, based on the LORCON library,
the MS sends a RTS frame to the AP identied through the MAC address and it waits for the
corresponding CTS frame response. When the AP processing time is measured, the measuring
system was not disabled between two RTS/CTS frames exchange because the AP does not
know, a priori, the time the RTS frames arrival because they are not synchronized. As other
frames coming from other WLANs could start or nish the count, the AP processing time
measurements have to be ltered a posteriori, because the triggers that starts and nishes the
count could not match with our RTS/CTS frames exchange.
As the AP processing time is independent of distance, the measurements were conducted for
a distance of 10 m between the two WLAN adapters in the two scenarios, exterior and corridor.
N
u
m
b
e
r

o
f

m
e
a
s
u
r
e
m
e
n
t
s
Corridor
(a) Zoom out
8489 8499 8999 8919 8929 8939 8949 8999 8969
9
99
199
199
299
299
399
399
N
m
e
r
o

d
e

M
e
d
i
d
a
s
Pasillo
Exterior
N
u
m
b
e
r

o
f

m
e
a
s
u
r
e
m
e
n
t
s
Corridor
(b) Zoom in
Fig. 7. AP processing time measurements in two different scenarios, exterior and corridor.
As Fig. 7 shows, there are mainly four different behaviors about AP processing time
measurements, which are around 0, 8500, 50000 and 65000 MCLK cycles. The extreme values,
0 and 65535 MCLK cycles, mean that the measuring system has not started the count before
reading the counters state and the count has overow, respectively. Measurements around
50000 MCLK cycles are not due to the RTS/CTS frame exchange because this behavior does
not appear in the exterior environment, where there were not signals coming fromother 802.11
devices. Measurements around 8500 MCLK cycles do not appear when the MS does not send
the RTS frames. Therefore, these measurements, around 8500 MCLK cycles, are due to the
AP processing time. When a RTS/CTS frame exchange is performed, it is assumed that the
AP processing time is roughly constant because more than 50% of measurements, which are
around 8500 MCLK cycles, were exactly 8494 MCLK cycles, although this actual value of the
AP processing time is not needed to apply the ranging method we propose in next section.
3. Distance estimation in LOS
According to (Chen & Ling, 2002), the range resolution is determined by the bandwidth of the
transmitted signal when RTT measurements are used. High-precision location would require
large transmission bandwidths or the use of multiple frequency channels. Furthermore, when
using a 44 MHz clock as input of the measuring systemto quantify the RTT measurements, the
maximum resolution achievable, if only one sample is taken, is hampered by that frequency
clock. Moreover, even in a LOS environment the RTT measurements have a random behavior
due to the error introduced by the standard noise from electronic devices, that is always
present. Therefore, to overcome these limitations several RTT measurements have to be
performed at each distance and a representative value, called the location estimator, from
this group of RTT measurements has to be selected as the distance estimation. The selection of
the location estimator is based on the model that relates the location estimator to the distance
that separates the MS and the AP.
The location of a random variable distribution can usually be presented by a single number,
the location estimator. In (Olive, 2008), several location estimators of a random variable are
analyzed. The mean, median, mode and the scale parameter of the Weibull distribution
(scale-W) are examples of the location of a random variable. In this chapter, they have
been analyzed and compared as location estimators of the RTT measurements in terms
of the coefcient of determination, r
2
. This coefcient measures how much of the original
uncertainty in the RTT measurements is explained by the model (Weisberg, 2005). In this
chapter, a simple linear regression function is assumed to be the model that relates the actual
distance between the two wireless nodes involved in RTT measurements to the location
estimators in LOS.
Analytically,
d
LOS
RTT
=
0
+
RTT
1
= d +
LOS
, (1)
where,

d
LOS
RTT
and d are the estimated and the actual distance between the MS and the AP
in LOS, respectively,
RTT is the location estimator of RTT measurements,

0
and
1
are the
intercept and slope of the simple linear regression function, respectively, and
LOS
is the error
introduced by
RTT. The error term

LOS
has been modeled as a zero-mean Gaussian random
variable, because the estimators used are asymptotically Gaussian and a large amount of
measurements have been used, so
LOS
N(0,
LOS
). (2)
In this case, as the expression (1) is a simple linear regression function, r
2
is simply the square
of the correlation coefcient, r
2
RTT,d
.
The parameters
0
and
1
that characterize the simple linear regression function do not
depend on the environment where the wireless localization system is going to be deployed,
but on the communication system used, i.e. the MS and the AP. These parameters are
computed so as to give a best t of the location estimators to the actual distance. Most
commonly, the best t is evaluated by using the least squares method, but this method
is actually not robust in the sense of outlier-resistance. Hence, robust regression has been
performed as it is a form of regression analysis designed to circumvent some limitations of
least squares estimates for regression models (Olive, 2008).
Assuming LOS between the MS and the AP without any scatter nearby and guaranteeing
the rst Fresnel zone clearance of the link between both nodes, three campaigns of RTT
measurements were conducted for several distances from 0 to 40 m. Fig.8 shows the robust
linear regressionfunction which best ts each location estimator to be analyzed. Each location
estimator has been computed from each group of 150 RTT measurements at each distance.
The different location estimators are analyzed and compared in terms of the coefcient of
determination value, r
2
.
The mode (
RTT
md
) is the value that is most likely to be sampled, thereby it could be a good
candidate for the location estimator, but the value that occurs the most frequently in a data
6985 6990 6995 7000 7005 7010 7015
5
0
5
10
15
20
25
30
35
40
45
CLK cycles [l]
D
is
t
a
n
c
e

[
m
]
E(d|l = l
1
)
E(d|l = l
0
)
l
0
l
1
N(
0
md
+
1
md
l
0
,
md
LOS
)
N(
0
md
+
1
md
l
1
,
md
LOS
)
d
LOS
RTT
=
0
md
+

RTT
md
1
md
(a) Mode (md)
6985 6990 6995 7000 7005 7010 7015
5
0
5
10
15
20
25
30
35
40
45
CLK cycles [l]
D
i
s
t
a
n
c
e

[
m
]
l
1
E(d|l = l
0
)
l
0
E(d|l = l
1
)
N(
0
mdn
+
1
mdn
l
0
,
mdn
LOS
)
N(
0
mdn
+
1
mdn
l
1
,
mdn
LOS
)
d
LOS
RTT
=
0
mdn
+

RTT
mdn
1
mdn
(b) Median (mdn)
6985 6990 6995 7000 7005 7010 7015
5
0
5
10
15
20
25
30
35
40
45
CLK cycles [l]
D
is
t
a
n
c
e

[
m
]
l
1
E(d|l = l
1
)
E(d|l = l
0
)
N(
0mn
+
1mn
l
0
,
mn
LOS
)
N(
0mn
+
1mn
l
1
,
mn
LOS
)
d
LOS
RTT
=
0
mn
+

RTT
mn
1
mn
l
0
(c) Mean (mn)
6985 6990 6995 7000 7005 7010 7015
5
0
5
10
15
20
25
30
35
40
45
CLK cycles [l]
D
is
t
a
n
c
e

[
m
]
l
1
l
0
E(d|l = l
1
)
E(d|l = l
0
)
N(
0sc
+
1sc
l
0
,
sc
LOS
)
N(
0sc
+
1sc
l
1
,
sc
LOS
)
d
LOS
RTT
=
0
sc
+

RTT
sc
1
sc
(d) Scale-W (sc)
Fig. 8. Robust linear regression function that best ts each location estimator to be analyzed:
the mean, median, mode and scale-W parameter. Where each location estimator is computed
from groups of 150 RTT measurements at each distance.
set is a discrete value. Therefore, the resolution achieved, T
MCLK
, is not enough for indoor
localization systems. The same resolution is achieved with the median (
RTT
mdn
) as it is a
discrete value separating the higher half of a data set. Fig.8 (a) (b) show that the Gaussian
distributions that characterize the errors
LOS
of the mode and the median are the widest,
md
LOS
= 7 m being r
2
md
= 0.64 and
mdn
LOS
= 3.6 m being r
2
mdn
= 0.9.
The mean (
RTT
mn
) is equivalent to the center of gravity of the distribution and it does
not take discrete values, thereby the resolution is improved. Although the mean is rather
sensitive in the presence of outliers, the use of a robust regression function circumvents this
limitation. Fig.8 (c) shows the errors committed when using the mean as location estimator
are characterized by a Gaussian,
mn
LOS
= 2.7 m, being r
2
mn
= 0.94, lower than the error
commit when the median. However, Fig.8 (d) shows that the best location estimator is the
scale-W parameter (
RTT
sc
) once Weibull distribution is tted to the RTT measurements. In
this case
LOS
is characterized by
sc
LOS
= 2.3 m and r
2
sc
= 0.96. Therefore, the assumption of
a linear function as the model that relates RTT measurements to distance is corroborated by
a correlation coefcient value close to the unit. This value indicates that the regression line
nearly ts the
RTT
sc
perfectly.
There is no phenomenological explanation for choosing the scale-W parameter as location
estimator of the RTT measurements set, however this parameter is another kind of a
location estimator since the maximum likelihood estimator (MLE) of the scale-W parameter
is the Hlder mean (Borwein & Borwein, 1986), a generalized form of the Pythagorean
means, taking as parameter the shape parameter of Weibull distribution (for more detail see
appendix).
0 1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Error [m]
P
r
o
b
a
b
ilit
y
Mean
Median
Mode
Scale~W
RSS~based
Fig. 9. CDFs of distance errors performed with four different location estimators and a
RSS-based method.
Once scale-Wparameter is found as the statistical estimator of the RTT measurements that best
ts the actual distance when using a simple robust linear regressionfunction as the model that
relates the estimator to the distance, its performance is compared to a RSS-based solution to
evaluate the goodness of the proposed one. The same two wireless nodes have been used in
the same LOS environment. As it is well known the distance between two wireless devices
causes an attenuation in the RSS values. This attenuation is known as path-loss and it is
modeled to be inversely proportional to the distance between both devices raised to a certain
exponent. According to (Mazuelas et al., 2009), the distance between two wireless nodes can
be estimated from RSS measurements by
d
RSS
= 10
P
re f
P
10
(3)
where

d
RSS
is the estimated distance between the MS and the AP, P
re f
is the RSS measured in
logarithmic units at the reference distance of 1 m, P is the average RSS in logarithmic units
at the actual distance, and is the path-loss exponent. According to (IEEE Standard, 2007),
for any distance under 20 m in LOS, is recommended to be 2 while = 3.5 for longer
distances. Therefore, having taken this value for the path-loss exponent and from the RSS
values measured between both devices, the distance between the two wireless nodes can be
estimated by using the expression (3).
Fig.9 shows the cumulative distribution function (CDF) of distance errors. As the mode and
the median take discrete values, the CDF has a step-shape with large errors. The mean has a
good behavior with an error lower than 2 mon average. However, the scale-Wparameter with
an error lower than 3 m for a cumulative probability of 80% achieves the best behavior. Also
in Fig.9 it can be appreciated that the scale-W parameter outperforms the RSS range based
method, specially for cumulative probabilities larger than 50%.
4. Distance estimation in NLOS
The assumption that LOS propagation conditions are present in an indoor environment is an
oversimplication of reality. In such environments the transmitted signal could only reach
the receiver through reected, transmitted, diffracted, or scattered paths. Hence, these paths
could positively bias the actual distance caused mainly by the blocking of the direct path or
due to experiencing a lower propagation speed through obstacles (Allen et al., 2007).
Known as the NLOS problem, this positive bias has been deeply considered through the
literature with the aim of mitigating its effect on distance estimates (Mazuelas et al., 2008;
Tang et al., 2008; Wylie & Holtzman, 1996), however, in all of them the NLOS is mainly
discussed within the cellular networks. Note that such techniques usually assume that the
bias for the NLOS range measurements changes over time and has larger variances than LOS
range measurements (Gven et al., 2008), assumptions that could not be assured in an indoor
environment (Yarkoni & Blaunstein, 2006).
In an indoor environment, the easiest method for dealing with NLOS conditions is simply to
place APs at additional locations and select those from LOS, however one of the objectives
of this chapter is to deploy a wireless localization system in a common and unmodied
wireless network. Therefore, in this chapter, the feasibility of the PNMC method presented
in (Mazuelas et al., 2008) is analyzed in an indoor environment, taking the PCB proposed in
previous section as measuring system, the scale-W as statistical estimator of the RTT and the
simple linear regression as the model to relate the scale-W to the actual distance.
The PNMC method relies on the statistical distribution of NLOS errors and on the major
variance that NLOS errors present with respect to LOS. The distribution type of NLOS
errors depends on the particular environment. Hence, it can follow different statistical
distributions such as Gaussian, Exponential, Gamma, etc. (Mazuelas et al., 2008). Regarding
the distribution, its parameters can be assumed to be constant in that particular environment.
Moreover, those parameters can be obtained before the process of getting distance estimates
(Cong & Zhuang, 2005) or directly from the estimated delay spread at that moment (Urrela et
al., 2006). In this chapter, those parameters have been obtained beforehand by a campaign of
RTT measurements in NLOS.
Let d be the actual distance between the MS and the AP, thus
d
NLOS
RTT
= d + , (4)
where,

d
NLOS
RTT
and d are the estimated and the actual distance between the MS and the AP,
respectively. The term denotes the error in the estimation of the distance. This error is the
sum of two independent errors, =
LOS
+
NLOS
, where
LOS
describes the noise form
electronic errors, while
NLOS
is the error due to the lack of direct sight between the MS and
the AP. On the one hand, the term
LOS
has been evaluated in the previous section and it
was found as a zero-mean Gaussian with
LOS
= 2.3 m. On the other hand, the term
NLOS
can follow different statistical distributions. However, regarding the distribution of
NLOS
, it
can be characterized by its mean and standard deviation. These parameters, as well as the
distribution type of
NLOS
, depends on the particular environment, but it can be assumed
that the NLOS propagation conditions do not change signicantly in the time window that
contains the record of range measurements, so the mean and standard deviation of
NLOS
can
be assumed to be constant. Moreover, the parameters that characterize
NLOS
can be obtained
previously to the localization process by the estimates performed in the environment where
the localization system is going to be deployed. For simplicity, the Exponential distribution
has been chosen for the term
NLOS
. Therefore,
NLOS
Exponential() (5)
where the parameter is xed previously to the localization process.
10 15 20 25 30
0
0.05
0.1
0.15
0.2
[m]
N
u
m
b
e
r

o
f

E
s
t
i
m
a
t
o
r
s

[
%
]
(a) Histogram
0 5 10 15 20 25 30
10
12
14
16
18
20
22
24
26
28
[
m
]
Samples
Original Samples
Corrected Samples
Actual Distance
(b) Estimators
Fig. 10. NLOS error correction from a record of distance estimates. (a) Histogram of distance
estimates and the probability density function that best ts the data. (b) Record of original
and corrected distance estimates after having applied the PNMC method.
In order to show the feasibility of the PNMC technique in an indoor environment, a campaign
of measurements in the second oor of the Higher Technical School of Telecommunications
Engineering (ETSIT) at the University of Valladolid has been carried out. Specically, the
PNMC technique is applied to the range measurements computed between an AP and a
MS 14 m away who is moving 5 m straight perpendicularly to the path that joins the AP
and the MS. As =
LOS
+
NLOS
, the probability density function (PDF) of the term is
the convolution of the Gaussian PDF caused by the
LOS
errors and the Exponential PDF
caused by the
NLOS
errors. Fig.10 (a) shows the histogram of the distance estimates record
and the PDF of the term that best ts these estimates, where the value of the parameter
that best ts the data is = 0.3 m
1
. Once the term is statistically characterized, the
PNMC technique can be applied. Fig.10 (b) shows the result of applying the PNMC technique
to the original range measurements computed in a time window equivalent to 5 m walking.
In this scenario, the ratio of
NLOS
errors from the record of range measurements has been
52%. Subsequently, these NLOS range measurements have been corrected by subtracting the
expected NLOS errors for each segment according to the Exponential distribution.
5. Experimental validation
The second oor of the ETSIT as a real indoor environment with several ofces, rooms and
many people walking around has been the selected scenario to test the wireless localization
systems accuracy. The 802.11 wireless network deployedin that building has been used as the
one over which the MS communicates with their APs whose positions are previously known.
Fig. 11 shows the layout of the south-wets of the second oor of the ETSIT building where
positioning tests have been carried out. The route followed by the MS describes a 40 11 m
2
rectangle walking through the middle of the corridors where each pair of continuous positions
is separated 0.75 m approximately. The corridors involved in the route are 2 m wide except
the widest one that has a width of 4.3 m. As a consequence of the heterogeneous distribution
of rooms and ofces, and the people walking around, multiple reection, diffraction or scatter
points could appear and alter the signal path. Presumably, although NLOS is always present,
multipath will be more noticeable when moving along the narrowest corridors.
Fig. 11. Indoor environment where RTT measurements have been carried out.
RTT measurements have been performedbetween an AP xed in a laboratory and a MS which
was moving on the route shown. Any of the positions on the route has a direct sight to the
AP, situation that could possibly happen in any indoor environment with high probability.
Therefore, the route followed shows different degrees of NLOS instead of LOS and NLOS
combinations.
The error introduced by the term NLOS is corrected by using the PNMC technique, where
NLOS is observed to be exponentially distributed with =0.3 m
1
and a time window
equivalent to 5 m walking is used. As location estimator, the scale-W parameter has been
implemented to reduce the error produced by the term LOS.
Fig. 12(a) shows the actual distance from the MS to the AP at each position through the route
shown in Fig. 11. The distance estimate, in red, at each position is shown by using the scale-W
as statistical estimator of the RTT, having applied the linear regressionmodel. As it is observed
in Fig. 12(a), due to the fact that the positions where the MS is going to be located do not have
a direct sight to the AP, the distance estimates are almost always higher than the actual one.
Therefore, PNMC method is going to correct distance estimates with severe NLOS. In blue,
the distance estimates having applied the PNMC method on the computed distance estimates
are shown. They are more similar to those that would be obtained in the absence of severe
NLOS propagation.
It can be concluded that although the difference between
LOS
and
NLOS
is not so great,
the presence of severe NLOS in the record is detected and corrected. As it is shown in
0 10 20 30 40 50 60 70 80
0
5
10
15
20
25
30
35
Positions separated each other 75 cm
D
i
s
t
a
n
c
e

[
m
]

Actual distance
scaleW
scaleW with PNMC
(a) Correction of severe NLOS measurements
with PNMC
0 2 4 6 8 10 12 14 16 18
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Error [m]
P
r
o
b
a
b
i
l
i
t
y

scaleW
scaleW with PNMC
(b) CDF
Fig. 12. NLOS error mitigation from a record of distance estimates using a window size of
5 m walking. (a)scale-W distance estimates before and after applying PNMC method. (b)
Comparison of CDFs errors in distance estimate before and after applying PNMC method.
Fig. 12(b) the improvement of applying the PNMC method can be observed through the CDF
of errors in distance estimates. Generally speaking, the distance estimate can be improved on
approximately 2 m for cumulative probabilities higher than 30% when applying the PNMC
method.
6. Conclusions
The achievable positioning accuracy of traditional wireless localization systems is limited
when harsh radio propagation conditions like rich multipath indoor environments are
present. In this chapter a novel RTT-based ranging method is proposed over a PCB that
performs RTT measurements. The effect of hardware errors has been minimized by choosing
the scale-W parameter as RTT estimator. A coefcient of determination value of 0.96 achieved
with this estimator in LOS justied the simple linear regression function as the model that
relates distance estimates to RTT measurements in LOS. As LOS is not guaranteed in an
indoor environment, the accuracy of the proposed localization algorithm has been tested in
a rich multipath environment without any NLOS error mitigation technique achieving an
error lower than 4 m on average. However, this error is improved after having implemented
the PNMC technique to correct NLOS errors. Once reliable RTT-based ranging estimates are
obtained, simple geometrical triangulation methods can be used to nd the location of the MS
(Pahlavan & Krishnamurthy, 2002).
Indoor localization schemes have experienced a urry of research in recent years. However,
there still remain multiple areas of open research that will help systems to meet the
requirements of applications that have to operate in indoor propagation environments
where GNSS typically fails. These are: i) Interference mitigation: To date, the majority of
research effort ignores the effects of interference on time estimation accuracy, and few papers
propose robust interference mitigation techniques. ii) Inertial Measurements Units (IMU):
the integration of traditional localization metrics, such as TOA, RSS or AOA with IMU
information, such as the one reported by accelerometers, gyroscopes and magnetometers,
could provide location estimations more precisely and continuously, since IMU-based
localization is a beacon-free methodology. iii) Secure ranging: In certain scenarios the
localization process may be subject to hostile attacks. While some works have presentedsecure
localization algorithms (see, e.g., (Li et al., 2005; Zhang et al., 2006)), less attention has been
paid to secure ranging.
7. Acknowledgment
This research is partially supported by the General Board of Telecommunications of the
Council of Public Works from Castilla-Len (Spain) and by the spanish national project
LEMUR (TIN2009-14114-C04-03).
8. Appendix
8.1 Maximum likelihood estimator of the scale parameter of the Weibull distribution
The scale-W parameter is estimated by using the MLE method and assuming that the shape
parameter is known.
The probability density function of a Weibull (two-parameter) random variable x is
f (x; k, ) =
k
k1
e
(
x
)
k
x 0
=
k
k
x
k1
e
(
x
)
k
x 0
where k > 0 is the shape parameter and > 0 is the scale-W parameter.
Let X
1
, X
2
, ...X
n
be a random sample of random variables with two-parameter Weibull
distribution, k and . The likelihood function is
L(x
1
, ...x
n
; k, ) =
n
i=1
f (x
i
; k, )
Therefore,
lnL(x
1
, ..., x
n
; k, ) =
n
i=1
lnf (x
1
, ...x
n
; k, )
=
n
i=1
ln
+ (k 1) ln
x
i
x
i
= n ln
+ (k 1)
n
i=1
ln
x
i
i=1
x
i
k
= n (ln(k) ln()) + (k 1)
n ln() +
n
i=1
ln(x
i
)
i=1
x
i
k
= n ln(k) + (k 1)
n
i=1
ln(x
i
) n k ln()
k
i=1
x
k
i
thus,
lnL
= n k
1
+ k
1
k+1

n
i=1
x
k
i
in order to nd the maximum,
lnL
= 0 then,
0 = n k
1
+ k
1
k+1

n
i=1
x
k
i
=

n
i=1
x
k
i
n
k
k+1
=
n
i=1
x
k
i
n
k
hence, the MLE of the scale-W parameter
1
n
n
i=1
x
k
i
1
k
this expression is known as the generalized mean or Hlder mean.
The Hlder mean is a generalized mean of the form,
M
p
(x
1
, x
2
, ..., x
n
) =
1
n
n
i=1
x
p
i
1/p
(6)
where the parameter p is an afnely extended real number, n is the number of samples and
x
i
are the samples with x
i
0. The Hlder mean is an abstraction of the Pythagorean means
which for example includes minimum (M
), harmonic mean (M
1
), geometric mean (M
0
),
arithmetic mean (M
1
), quadratic mean (M
2
), maximum (M
), and the MLE of the scale-W

parameter (M
k
) where k is the shape parameter of Weibull distribution.
9. References
Bahillo, A., Mazuelas, S., Lorenzo, R.M., Fernndez, P., Prieto, J. & Abril, E.J. (2009a).
Indoor location based on IEEE 802.11 round-trip time measurements with two-step
NLOS mitigation. Progress In Electromagnetics Research B, PIERB, Vol.2009, No.15,
(September 2009) 285-306, ISSN: 1937-6472.
Bahillo, A., Prieto, J., Mazuelas, S., Lorenzo, R.M., Blas, J. & Fernndez, P. (2009b). IEEE
802.11 distance estimation based on RTS/CTS two-frame exchange mechanism,
Proceedings of 69th international conference of Vehicular Technologies, pp. 1-5, ISBN:
978-1-4244-2517-4, Barcelona, June 2009, IEEE VTC, Spain.
Golden, S.A. & Bateman, S.S. (2007). Sensor measurements for wi location with emphasis on
time-of-arrival ranging. IEEE Transactions on Mobile Computing, Vol.6, No.10, (October
2007) 1185-1198, ISSN: 1536-1233.
Gustafsson, F. & Gunnarson, F. (2005). Mobile positioning using wireless networks:
possibilities and fundamental limitations based on available wireless network
measurements. IEEE Signal Processing Magazine, Vol.22, No.4, (July 2005) 41-53, ISSN:
1053-5888.
Mazuelas, S., Bahillo, A., Lorenzo, R. M., Fernndez, P., Lago, F.A., Garca, E., Blas, J. &
Abril, E. J. (2009). Robust indoor positioning provided by real-time RSSI values in
unmodied WLANnetworks. IEEE Journal of Selected Topics in Signal Processing, Vol.3,
No.5, (October 2009) 821-831, ISSN: 1932-4553.
Mazuelas, S., Lago, F.A., Blas, J., Bahillo, A., Fernndez, P., Lorenzo, R. M. & Abril, E. J.
(2008). Prior NLOS measurements correction for positioning in cellular networks.
IEEE Transactions on Vehicular Technologies, Vol.58, No.5, (November 2008) 2585-2591,
ISSN: 0018-9545.
Morrison, J.D. (2002). IEEE 802.11 wireless local area network security through location
authentication. M.S. Thesis, Naval Postgraduate School Monterey, California.
Pahlavan, K. & Krishnamurthy, P. (2002). Principles of wireless networks - A unied approach,
Prentice-Hall Inc., 2nd edition, ISBN: 0-13-093003-2, Upper Saddle River, New Jersey.
Prieto, J., Bahillo, A., Mazuelas, S., Blas, J., Fernndez, P. & Lorenzo, R. M. (2008). RTS/CTS
mechanism with IEEE 802.11 for indoor location, Proceedings of the Navigation
Conference & Exhibition: Navigation and Location, pp. 1-5, London, UK, October 2008,
NAV & ILA.
Seow, C.K. & Tan, S.Y. (2008). Localization of omni-directional mobile device in multipath
environments. Progress In Electromagnetics Research, PIER, Vol.85, No.2008, (2008),
323-348, ISSN: 1070-4698.
Soliman, M.S., Morimoto, T. & Kawasaki, Z.I. (2006). Three-dimensional localization system
for impulsive noise sources using ultra-wideband digital interferometer technique.
Journal of Electromagnetic Waves and Applications, Vol.20, No.4, (2006), 515-530, ISSN:
0920-5071.
Gast, M.S. (2002), 802.11 Wireless networks: the denitive guide, OReilly &Associates, Inc., ISBN:
0-596-00183-5, 1005 Gravenstein Highway North, Sebastopol, CA 95472.
Chen, V.C. & Ling, H. (2002), Time-frequency transforms for radar imaging and signal analysis,
Artech House, Inc., ISBN: 1-58053-288-8, 685 Canton Street, Norwood, MA, 02062.
Olive, D.J. (2008), Applied robust statistics, Southern Illinois University, Department of
Mathematics, 4408 Carbondale, IL 62901-4408.
Weisberg, S. (2005), Applied linear regression, 3rd ed., John Wiley & Sons, Inc., ISBN:
0-471-66379-4, Hoboken, New Jersey.
Intersil, (2002), HFA3861B wireless LAN medium access controller, Intersil Data Sheet.
Borwein, J.M. & Borwein, P.B. (1986), Pi and the AGM: a study in analytic number theory and
computational complexity, John Wiley & Sons, Inc., ISBN: 0-471-83138-7, USA.
IEEE Standard for Information Technology (2007) Telecommunications and Information Exchange
Between Systems - Local and Metropolitan Area Networks - Specic Requirements - Part 11:
Wireless MediumAccess Control (MAC) and Physical Layer (PHY) Specications, IEEE Std
802.11-2007 (Revision of IEEE Std 802.11-1999).
Allen, B., Dohler, M., Okon, E., Malik, W., Brown, A. & Edwards, D., (2007), Ultra Wideband
Antennas and Propagation for Communications, Radar, and Imaging, John Wiley & Sons,
Inc., ISBN: 0-470-03255-3, West Sussex, UK.
Tang, H., Park, Y. & Qui, T., (2008). NLOS mitigation for TOA location based on a modied
deterministic model. Research Letters in Signal Processing, Vol.8, No.1, (April 2008) 1-4,
ISSN: 1687-6911.
Wylie, M. P. & Holtzman, J., (1996). The non-line of sight problem in mobile location
estimation. Proceedings of the 5th IEEE International Conference on Universal Personal
Communications, Vol.2, pp. 827-831, ISBN: 0-7803-3300-4, Cambridge, October 1996,
Mass, USA.
Gven I., Chong, C.-C., Watanabe, F. & Inamura, H., (2008). NLOS identication and
weighted least-squares localization for UWB systems using multipath channel
statistics, EURASIP Journal on Advances in Signal Processing, Vol. 2008, (April 2008)
1-14, ISSN: 1110-8657.
Yarkoni, N. & Blaunstein, N., (2006). Prediction of propagation characteristics in indoor radio
communication environment. Progress In Electromagnetics Research, PIER 59, (2006)
151-174, ISSN: 1070-4698.
Cong, L. & Zhuang, W., (2005). Non-line-of-sight error mitigation in mobile location.
INFOCOM2004, 23th Annual Joint Conference of the IEEEComputer and Communications
Societies, Vol. 4, pp. 560-572, ISBN: 0-7803-8355-9, Hong-Kong, March 2004.
Urrela, A., Sala, J. & Riba, J., (2006). Average performance analysis of circular and hyperbolic
geolocation. IEEE Transactions on Vehicular Technology, Vol. 55, (January 2006) 52-66,
ISSN: 0018-9545.
Chen, P.-C., (1999). A non-line-of-sight error mitigation algorithm in location estimation.
Proceedings of Wireless Communications and Networking Conference, Vol. 1, pp. 316-320,
ISBN: 0-7803-5668-3, New Orleans, La, USA, September 1999.
Li, Z., Trappe, W., Zhang, Y. & Nath, B., (2005). Robust statistical methods for securing
wireless localization in sensor networks, Proceedings of IEEE Information Processing
Sensor Networks, pp. 91-98, ISBN: 0-7803-9202-7, Los Angeles, California, USA, April
2005.
Zhang, Y., Liu, W., Lou, W. & Fang, Y., (2006). Location-based compromise-tolerant
security mechanisms for wireless sensor networks, IEEE Journal on Selected Areas in
Communications, Vol. 24, (February 2006) 247-260, ISSN: 0733-8716.
13
Data Forwarding in Wireless Relay Networks
Tzu-Ming Lin, Wen-Tsuen Chen and Shiao-Li Tsao
ITRI, NTHU, &NCTU,
Hsinchu,
Taiwan, R.O.C.
1. Introduction
Broadband wireless communication has brought users a number of multimedia services for
years. As wireless broadband markets mature, system operators face new problems. First,
system capacity is bounded by finite radio resources. Second, service providers increasing
spend on network deployments and services provisions. With limited radio resource and
increased costs, operators have no choice but to force users to pay higher rates to preserve
the same service quality than they had several years ago.
The expectation is that these problems are solved by introducing relay station (RS) into
traditional wireless systems. The relaying technology has proven to be a feasible technology
to expand system capacity and reduce deployment costs simultaneously. One of the most
popular technologies is cooperative communications that boost network throughput
significantly by improving the radio quality for transmission and reception (Nosratinia et
al., 2004); (Mohr, 2005); (Doppler et al., 2007). Numerical results (Soldani & Dixit, 2008)
showed the cost saving benefits when conventional communication networks adopt relay
functionality. From this, a commercial cellular system can save more than 56 % on capital
expenditures when RSs are deployed.
RS is a wireless communication station that provides relay services for receiving and
forwarding radio signals between two stations. In a cellular system, RS is in charge of
receiving, decoding, and forwarding data between the Base Station (BS) and the Subscriber
Stations (SSs). During the relaying, RS provides additional features to assist data
transmissions. For example, (Tao et al., 2007) proposed a novel data forwarding scheme in
this environment. The novel data forwarding scheme improved the transmission efficiency
by approximately 66 %.
Although data relay shows superior performances in throughput enhancements and cost
savings. Multi-hop relaying introduces new issues that impact overall performance. To
indicate a routing path in multi-hop circumstances, additional control information should be
introduced. The routing indication design for multiple radio links introduces new overhead
and may impact transmission efficiency during relay. Imprecise indications may waste RS
computation and storage since it does not know what it shall relay and where to forward.
Moreover, there may be a case that data destined to others are received by the RS that is not
in charge. Imprecise relay indication will not only increase overheads but also lead RS to use
its processor and storage inefficiently. In other words, data forwarding with ambiguous
control information would increase RS complexity and buffer storage unnecessarily. During

238
the relaying, RS shall forward data as simple as possible to prevent wasting processing
power and storage. This study proposes a burst-switch concept aiming to tackle the issues
and provides a simple and efficient data forwarding for wireless relay networks.
The rest of the chapter is organized as follows. First, wireless relaying and a conventional
relay system are overviewed in section 2. The proposed new forwarding mechanism is then
elaborated in section 3; section 4 presents the evaluation and simulation results for the
mentioned issues. At last, conclusions are given in section 5.
2. Background and related works
Relay technology has been investigated for years, and a realistic relay system will be
deployed in few years to enhance legacy wireless networks. This section overviews IEEE
802.16 communication system and introduces the relay enhancements. The data forwarding
mechanisms adopted in the system are also discussed to address the issues.
2.1 Overview of wireless relay networks
A wireless relay network consists of a BS, one or more RSs, and numbers of SSs. In the
network, directly or through the assistances of RSs. BS forwards the downstream data
coming from outside network to SS while RSs relay upstream data generated by SSs to BS.
Since all the data transmissions within the network are arranged by BS and there are no
communications between RSs, the relay network is usually constructed as a tree topology,
which is illustrated in figure 1.

Fig. 1. Wireless Relay Network
There are two types of radio links in this network: relay link and access link. The radio
link between a BS and a RS and between two RSs are called relay links, and BS constructs
a relay path by multiple relay links. The access link is the radio communication between a
SS and its access station, which can be a BS or an access RS. The access RS is a RS attached
by a SS and can helps BS for relaying data to the SS. For the example in figure 1, RS
2
is an
access RS of SS
u
and assists BS
0
to provide relay services for SS
u
. BS
0
allocates resources

239
along the relay path between SS
u
and itself so that the help on data relay in corresponding
relay links.
2.2 IEEE 802.16 and multi-hop relay network
IEEE Std 802.16e
TM
-2005 is one of the most popular wireless broadband networks nowadays,
and figure 2 shows the reference model for that system. The system consists of two layers,
Medium Access Control (MAC) and Physical (PHY) layers, to handle wireless
communications. Packets from TCP/IP layer are translated into MAC Protocol Data Units
(MPDUs) and then encoded into a PHY burst. The burst is associated with a MAP
Information Element (MAP-IE) that indicates a station for receiving and decoding the burst.
After the data process, BS transforms both the burst and the associated MAP-IE into a radio
frame and pumps it into wireless medium.

Fig. 2. Data Processing in IEEE 802.16
The overview of 802.16e frame structure is depicted in figure 3. The frame composes two
subframes: downlink and uplink subframes, and starts with a synchronization part of
preamble and Frame Control Header (FCH). The first part is used for each receiving station
synchronize with BS and abstracting the frame. Following the synchronization part, the
frame header further includes a downlink MAP (DL-MAP) and an uplink MAP (UL-MAP),
which consists of MAP-IEs to indicate the stations where and how to access data bursts. As
stated before, each data burst is associated with an MAP-IE, and one or more MPDUs
destining to a destination can be concatenated or packed into the burst. With a connection
identity (CID) in the MAP-IE, every receiving station locates and receives MPDUs in the
desired PHY burst, and has no need to check all the bursts in the frame.

240

Fig. 3. IEEE 802.16e Frame Structure
The system further specifies an option for disabling MAP to save overheads so that more
data can be allocated. In this case, receiving stations should put more efforts to process
entire frame since there are no indications in frame header any more. Without MAP
indication, the receiving station cannot but store the whole frame to check if there are any
desired MPDUs. However, it is inefficient for buffering and checking all the MPDUs in a
frame. Although, the operations brought overheads but the problem should not be as
serious as that in multi-hop communications. Because of redundant processing and
transmissions during relay, multi-hop data forwarding makes the overhead become a severe
problem.
2.3 Data forwarding and issues in 802.16j MR network
IEEE working group specifies Multi-hop Relay (MR) support for 802.16e system in IEEE Std
802.16j
TM
-2009. The specification aims to solve the capacity problem and reduce
development cost with advanced relay technologies. Efficiency during data relay is also a
major concern for implementing RSs. Two data forwarding schemes are specified to
facilitate relay functionalities and reduce overheads.
The first forwarding scheme is CID-based transmission, in which RS forwards MPDUs
based on the CIDs contained in the MAP-IE or MPDU headers. For saving signalling
overheads, relayed MPDUs do not carrier any extra routing information, and are
transmitted as in 802.16e conventional system. When MPDUs are relayed, each receiving RS
gets CIDs from MAP-IEs and checks associating bursts if there are any data required for
further forwarding. The RS discards the burst that is not indicated by the recorded CIDs in
its forwarding list. When the CID of the burst is in the forwarding list, the receiving RS
forwards the burst to the station in next hop. Besides, there is another implementation that
RSs forward MPDUs by identifying CIDs containing in MPDU headers. As mentioned, each
RS has to process all the MPDU in receiving frame, and determines the MPDUs for relay.
Figure 4 depicts the example for this forwarding scheme based on the relay network in

241
figure 1. RS
1
receives all the packets containing in first frame and stores the MPDUs
destining to RS
1
, RS
2
, RS
3
, RS
5
, and RS
6
. In second frame, RS
2
caches the MPDUs from RS
1

and checks which MPDUs it shall relay. The data for RS
2
and RS
3
are received and only RS
3

data are relayed. After that, RS
3
receives the data by checking MPDU headers and performs
relaying for SS
w
, SS
y
, and SS
y
individually. Moreover, the first option for adopting MAP-IE
indications can filter unnecessary bursts before looking into every MPDU and RS needs not
store all the data in the frame. Otherwise, checking CIDs in MPDU headers would force RSs
to receive all the data in a frame. As a result, the using of MAP is a trade-off issue between
storage and signalling because enabling MAP prevents RSs to store MPDUs unnecessarily
but brings extra control overheads.

Fig. 4. CID-Based Transmission
The second scheme is tunnel-based transmission, in which BS and access RS encapsulates
MPDUs into a Tunnel PDU (T-PDU) and transmits these data through a tunnel in between.
Figure 5 shows an example of this scheme. BS, the ingress station of a tunnel, aggregates
MPDUs into one or more T-PDUs, and transmits the data to an access RS. The access RS at
the egress of tunnel, e.g. RS
3
, is responsible for removing tunnel headers and forwarding the
decapsulated MPDUs to SSs. Besides, the intermediate RSs along a relay path relay the T-
PDU to the tunnel end through the indication of tunnel headers. As CID-based operation,
enabling MAP can help RSs to filter unwanted data, and disabling MAP would force RSs to
buffer and process all the T-PDUs in the frame. Tunnel header benefits in preventing
redundant processes for the group of MPDUs destining to same access RS. However, the

242
extra overhead brought from the tunnel header should be considered. Furthermore, the
impact caused by adopting MAP and tunnel headers at the same time shall be also
investigated.

Fig. 5. Tunnel-based Transmission
Comparing these two forwarding schemes, RSs applying tunnels forward data efficiently
since RS identifies a MPDU group by a tunnel header in replace of multiple MPDU headers.
CID-based scheme provides the forwarding as simple as that in legacy one-hop system, and
introduce no extra headers. Although both the two mechanisms can assist data forwarding
in multi-hop environments, the overheads caused by excess header processing and
unnecessary data buffering shall be discussed in advanced. First, the processing overhead
for RSs would rise with the system traffic load. Take the example in figure 1, RS
5
shall
process 2n tunnel headers in tunnel-based scheme when each RS maintains n tunnels. If m
MPDUs are scheduled for each SS in CID-based scheme, 4m MPDU headers will be handled
by RS
2
. In the example, both m and n increase with the traffic load and all the intermediate
RSs suffer. The impacts to processing complexity of the two schemes are:

_ _ Tunnel based tunnel header tunnel RS
C L N N = (1)

_ _ Station based MPDU header MPDU
C L N = (2)
where C
Tunnel_based
and C
CID_based
define the computation in bits, L
tunnel_header
and L
MPDU_ header
denote the lengths of tunnel and normal MPDU headers, N
tunnel
is the tunnel number that a
RS shall handle, N
RS
is the summation of all the RSs behind a receiving station for a relay

243
path, and N
MPDU
is the amount of MPDUs that needs to be forwarded. In a stationary relay
network, both C
Tunnel_based
and C
CID_based
grow with N
tunnel
and N
MPDU
, and both the two
numbers correspond to the network load. As system load rises, tunnel numbers would
increase with provisioned connections. If CID-based forwarding is applied in this case, the
amount of processed MPDUs would also grow with the increased connections.
Although tunnel header prevent RSs to process MPDU header redundantly, the issue of
inefficient data processing and buffering remains unsolved; every RS needs to store all the
T-PDUs or MPDUs in a frame before getting relay information. Some undesired data would
still be processed and buffered before being processed. If MAP is applied, RSs could handle
fewer data since the bursts for other destinations can be filtered by associating MAP-IEs.
However, another issue for excess processing is unsolved because there are two-levels of
control information, one for MAP and the other for MPDU or T-PDU headers. No matter
which mechanism is applied, the addressed problems cannot be taken over totally. Aiming
this, a simple data forwarding mechanism is proposed in this study for relaying data more
efficiently.
3. End-to-end burst switch and proposed network model
This study brings out a new concept of switching burst to forward data in the wireless relay
network. The burst switch mechanism uses RSs more efficiently by reducing processed
control information and buffered data. Besides, a new network model for adopting the
concept is also proposed to realize the forwarding process.
3.1 End-to-end burst switch
This literature suggests using an end-to-end burst CID in the associated MAP-IE to forward
relayed data, and proposes to relay data with a unique identifier for each relay path in burst
level. To reduce carried routing information, data goal to same destination RS are assigned
with one burst CID. With the help of the burst CID, intermediate RSs relay bursts without
checking the MPDU so as to eliminate the processing in MPDU level. Moreover, checking
CIDs in MAP also saves unnecessary processing for the data destining to other destinations.
Before relaying data, BS identifies the access RS and sets up the forwarding path for a SS. As
the legacy schemes, a data burst binding a relay path aggregates multiple MPDUs and is
transmitted in frames. Since each relay path is identified by a burst CID, each RS along a
relay path checks the MAP to locate the relayed burst. For the RS not in the path, it just
ignores the data burst after checking the MAP. During the relay process, RSs need not to
look into MPDUs to identify the data or routing. The usage of CIDs in this scheme is similar
to that used in tunnel-based scheme, but the CID is used in burst level, not in T-PDU level.
Such an enhancement saves the overheads for both control and data parts. By referring burst
CID in frame header, the RS can decide to receive the associated burst or not. If the burst for
the relay path is located, the RS transmits the entire burst toward next RS without
processing the data in further. The intermediate RSs do not decapsulate inside data but
switch the burst along the relay path until the destination RS receives it. After the burst is
switched to the end of relay path, the access RS forwards the data to SSs using individual
CIDs in access links as BS does in the conventional one-hop network. Since bursts are
identified and switched by the burst CID, the proposed forwarding scheme is so called burst
switch.

244

Fig. 6. Burst Switch Transmission
Figure 6 shows the proposed scheme in detail. In the figure, BS establishes a path for
switching the burst for SS
w
, SS
x
, and SS
y
, and assigns an end-to-end burst CID for the relay
destination, RS
3
. When receiving the MAP in first hop, RS
1
and RS
2
check the burst CID and
buffer the burst for switching in second hop. After receiving the switched burst, RS
3

decapsulates the burst and processes the MPDUs to see where the destination for the data is.
After that, access RS
3
forwards the decoded MPDUs to SS
w
, SS
x
, and SS
y
separately. The
computation for this relay process is:

_ Burst switch MAP IE RS
C L N
= (3)
Where C
Burst_Switch
is the computation cost, L
MAP-IE
denotes MAP-IE length and N
RS

represents the numbers of RSs behind the station. Since N
RS
is a fixed value as network has
been setup, neither the network load nor other traffic factors impact C
Burst_Switch
. Compared
with the tunnel-based and CID-based solutions, the proposed scheme adopts the 32 bits
MAP-IE and saves 33 % overhead while the two schemes apply 48 bits T-PDU or MPDU
headers. Therefore, it can be expected that C
Burst_Switch
< C
Tunnel_based
< C
CID_based
. Moreover,
RS with burst switch identifies relay destination before storing data and takes the advantage
of storage saving.
3.2 Proposed network model
Before providing relay services, the relay system shall construct a relay network model to
determine relay paths for all the SSs. This paragraph provides an overview of the proposed
model, in which the burst switch mechanism can be applied. In the network, BS maintains a
forwarding table that keeps the CID for every relay path and access link while RSs only
record the CIDs for the stations in behind. The CIDs includes the burst CID of access RSs

245
and also the access CIDs of SSs. Figure 6 shows the model for the example of the wireless
relay network in figure 1.

Fig. 7. Proposed Relay Network
In the example, BS can obtain the CIDs for RS
1
to RS
6
and SS
t
to SS
z
from the table, and RS
2

only holds the CIDs for RS
3
RS
6
, and SS
u
. When initializing a relay transmission, BS first
identifies a receiving SS and the associating access RS from the table. Then, radio resources
are allocated for relay links in the relay path. By the MAP indications, intermediate RSs
switch burst by checking the CIDs within MAP when receiving a frame. Only when the
received burst CID matches the recorded CID, the binding burst is switched to next hop.
Intermediate RSs switch the burst until the burst arrivals the relay destination, and the
access RS forwards MPDUs to SSs in access links after receiving.
In figure 6, data bursts for all the SSs are ready for relay, and BS assigns burst CID for each
burst based on the forwarding table. Bursts for SS
t
associates with the CID of RS
1
, and bursts
for SS
u
and SS
z
are associated with the CIDs of RS
2
and RS
6
respectively. Because SS
w
, SS
x
,
and SS
y
attach the same access RS, data for these SSs are encapsulated together and assigned
with the CID of RS
3
. When the relaying begins, RS
1
checks CIDs in the MAP and see if there
are any bursts associating with the CIDs of SS
u
, RS
2
, RS
3
, and RS
6
. Likewise, RS
2
operates
similarly, but the difference is switching bursts to RS
3
and RS
6
individually due to separated
burst CIDs. After receiving the bursts, RS
3
and RS
6
forward the data to their SSs using the
CIDs of SS
w
, SS
x
, SS
z
, and SS
u
to complete the relay process.
The major outcomes of the proposed scheme are determined control overhead, simple
forwarding, and efficient buffer usage. Unlike tunnel-based and CID-based approaches,
processing overheads of proposed method increase with static RS number, not dynamic
network traffic load. Moreover, it is simple because each RS identifies desired data by
matching CIDs in frame header. Beside, RSs identify relay destination before caching it, and

246
do not store data irrelevantly. To validate the outcomes, next section evaluates the
performance of burst switch among various scenarios and compares the results with the
other forwarding schemes.
4. Simulation results
This study investigates the issues on data forwarding during relay in terms of process and
storage efficiency. Two metrics are introduced for the evaluations. Processing overhead, the
first metric, is the total bits that RS shall handle through all the relay process. The second
metric, storage overhead, is the memory space in which RS needs to store bursts during
relaying. Besides, five forwarding schemes are evaluated in the simulations: (1)
CID_w_MAP, (2) CID_w/o_MAP, (3) Tunnel_w_MAP, (4) Tunnel_w/o_MAP, and (5) Burst
Switch. Schemes (1) and (3) represent CID-based and tunnel-based forwarding schemes with
MAP support while scheme (2) and (4) denote the relay schemes without MAP indications.
For the schemes enabling MAPs, the CID in a MAP-IE directs the station in a hop-by-hop
manner, which means RSs swap CIDs for each hop during the relaying. The proposed burst
switch scheme employs a burst CID to indicate an end-to-end relay path for each access RS.
Table 1 lists detailed parameters, including system and traffic, and the four traffic types
defined in 802.16e are also used as inputs to show the performance differences between
services.

System Parameters Traffic Parameters
Channel Bandwidth 20 MHz Arrival Time - UGS 30 ms
Frame Duration 5 ms MPDU Size - UGS 200 bytes
Symbols Per Frame 48 Arrival Time - RTPS 30 ms
FFT Size 2048 MPDU Size - RTPS 1000 bytes
Code Rate 1/2 Arrival Time - NRTPS 100 ms
Modulation QPSK MPDU Size - NRTPS 1500 bytes
RS Number 20 Arrival Time - BE 300 ms
SS Number 100 MPDU Size - BE 800 bytes
Table 1. Simulation Parameters
Figure 8 illustrates the statistics of processing overhead for RSs in 2-hop relay case. It is
observed that CID_w/o_MAP makes severe processing overheads. Moreover, the curve
grows dramatically as the number of traffic flows increases. On the contrary, the overheads
for the other forwarding schemes rise slightly. The figure tells that disabling MAPs in two-
hop case would cause excess overhead for RSs and the results come from processing all the
MPDUs in a frame. Although MAP is also disabled in scheme (4), Tunnel_w/o_MAP
performs best by introducing tunnel header to avoid the drawback. When comparing with
scheme (4), Tunnel_w_MAP shows a little worse performance since MAPs and tunnel
headers are provided redundantly. It is worth to note that CID_w_MAP and proposed
scheme show identical results in the figure because the CIDs in MAPs both denote the
destination RS in two-hop cases.

247
0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
20 40 60 80 100 120 140 160 180 200
Number of Traffic Flows
P
r
o
c
e
s
s
i
n
g

O
v
e
r
h
e
a
d

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 8. Processing Overhead in 2-hop Scenario

0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
20 40 60 80 100 120 140 160 180 200
P
r
o
c
e
s
s
i
n
g

O
v
e
r
h
e
a
d

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 9. Processing Overhead in 5-hop Scenario
As the relay hop count increases, RSs shall handle more and more control information in the
relaying process, and figure 9 depicts the results in multi-hop relay environments. Generally
speaking, the processing overheads of all the schemes increase with traffic load in light
traffic load cases (i.e. less than 80 traffic flows). Tunnel-based schemes have less

248
computation overhead than CID-based schemes since tunnel headers can prevent RSs from
processing all the MPDUs in a frame. Nevertheless, the curse of tunnel-based schemes also
rises as increased traffic in heavy traffic load conditions (i.e. more than 100 traffic flows),
and it is because more and more tunnels are established as traffic load increases. Besides,
high traffic load also increases the MPDU amount, and, unfortunately, CID_w/o_MAP still
obtains the worst results. The key worth to be noticed here is that the curve of CID_w_MAP
and Birth Switch scheme rises limitedly in heavy load cases. Compared with the
CID_w_MAP, RSs employing Birth Switch save about 33 % overhead in processing MPDUs,
and the figure also shows the outcome. As a consequence, it is obtained that schemes (2), (3),
and (4) are impact by system load, and the processing overhead increases with the number
of traffic flows. CID_w_MAP and the proposed scheme can provide a stable outcome in RS
processing, especially when the relay network is loading high. The burst switch concept can
save more overheads than CID_w_MAP, and is suitable to forward data in multi-hop
environments.

0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
20 40 60 80 100 120 140 160 180 200
B
u
f
f
e
r
e
d

D
a
t
a

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 10. Storage Overhead in 2-hop Scenario
The second issue is storage overhead, and this study use figure 10 and figure 11 to discuss
the evaluations. Both the two simulations confirm the impact resulting from traffic loads. In
light load conditions, RSs in 2-hop case do not buffer as much data as those in 5-hop case
because the amount of relayed data is small. When the traffic loads high, RSs in both cases
have to forward and store more data. Note that the curve becomes saturated for high load
cases in figure 10 due to limited radio resource. Moreover, it is easy to find that the schemes
without MAP indication, e.g. (2) and (4), cache much more data than the other three
schemes do. It is because RSs cannot identify MPDUs before receiving the entire frame, and
hence check the MPDUs one by one. On the contrary, RSs in other schemes can filter
MPDUs by either frame or tunnel headers so that data would not be buffered inadequately.

249

0
10000
20000
30000
40000
50000
60000
70000
80000
90000
100000
20 40 60 80 100 120 140 160 180 200
B
u
f
f
e
r
e
d

D
a
t
a

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 11. Storage Overhead in 5-hop Scenario

0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
2 3 4 5
Hop Counts
P
r
o
c
e
s
s
i
n
g

O
v
e
r
h
e
a
d

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 12. Processing Overhead in Light Load Cases
This study further evaluates the impact brought from long path lengths. Figure 12 and 13
show the results in light and heavy traffic load conditions respectively. In the simulations of
figure 12, the scheme, CID_w/o_MAP, performs worst in all the cases. With the MAP
assistance, CID_w _MAP can perform better than tunnel-based solutions. However, the
tunnel solution without MAP outperforms the other tunnel-based scheme because of
avoiding extra processing for both MAP and tunnel headers. Furthermore, the RS

250
employing burst switch also avoids redundant and unnecessary processing by burst CIDs so
that the proposed scheme also has superior performance in light load conditions.

0
20000
40000
60000
80000
100000
120000
140000
160000
180000
200000
2 3 4 5
Hop Counts
P
r
o
c
e
s
s
i
n
g

O
v
e
r
h
e
a
d

(
b
i
t
s
)
CID_w_MAP
CID_w/o_MAP
Tunnel_w_MAP
Tunnel_w/o_MAP
Burst Switch

Fig. 13. Processing Overhead in Heavy Load Cases
In figure 13, these forwarding schemes show similar result with that in previous experiment.
CID-based forwarding schemes show the frustrated results in heavy load occasions if the
network disables MAP support. The other four mechanisms obtain more stable performance
in overhead saving when relay hop count increases. Regarding tunnel-based solutions, it is
found that enabling MAP does not increase as much overheads as applying tunnel headers
because it is tunnel header to dominate the overhead in heavy load conditions. The
forwarding schemes without introducing any extra MPDU headers perform better in the
simulation. The simulations also suggest the best performance of Burst Switching in multi-
hop environments, especially for heavy load cases.
Processing complexity and storage usage are critical issues for RS developments in wireless
relay networks. The evaluation results depict overall comparison of the two metrics in figure
14 and figure 15. We can see the performance of overhead saving among the data
forwarding schemes. This study uses the results of Tunnel_w/o_MAP as basics to show
following observations. (1) Without MAP indications, CID-based forwarding scheme
performs worst in saving processing and storage overheads, no matter in 2-hop or 5-hop
cases. (2) In tunnel-based schemes, enabling MAP indication can reduce the amount of
cached data with some penalty from extra MAP-IE processing. (3) Comparing the benefits
coming from applying MAP and tunnel headers, CID_w_MAP outperforms
Tunnel_w/o_MAP in storage efficiency since unnecessary MPDU buffering is avoided. (4)
CID_w_MAP scheme introduce irrelevant processing overhead in 5-hop and light load case
although it shows more adequate results than the other three conventional schemes. (5) The
proposed burst switch scheme performs better than all other forwarding schemes in both

251
heavy and light traffic load conditions. (6) Considering the performance among different
traffic loads, burst switch scheme saves approximately 8 % computation and about 60~73 %
storage costs from the results of evaluations.

Fig. 14. Overall Comparison for Processing Overhead

Fig. 15. Overall Comparison for Storage Overhead

252
2-Hop Multi-Hop
Scenario
Schemes
Processing Storage Processing Storage
CID_w_MAP Good Good Bad Good
CID_w/o_MAP Bad Bad Bad Bad
Tunnel_w_MAP Good Good Bad Good
Tunnel_w/o_MAP Relevant Bad Relevant Bad
Burst Switch Good Good Good Good
Table 2. Evaluation Summary
After comparing all the results from the view point of efficiency in processing and storage,
the suggestion for employing data forwarding mechanism can be obtained in TABLE II. In
CID-based mechanisms, it is suggested that adopting MAP helps on reducing process and
storage overheads. However, both schemes do nothing with the processing problem in
multi-hop circumstances. For tunnel-based solutions, network operator shall consider
deployment scenario to enable MAP or not since tunnel-based scheme with MAP support
performs irrelevantly while disabling MAP for tunnel-based solution causes storage
problem in multi-hop cases. After all, CID_w_MAP, Tunnel_w_MAP, and Burst Switch are
preferred to be used in 2-hop relay networks since the advantage of low processing and
buffering overhead were seen. In multi-hop scenario, all the other forwarding schemes are
not suggested because these schemes do not perform as superior as Burst Switch. The
proposed forwarding can solve all the overhead problems in 2-hop and multi-hop scenarios.
5. Conclusions
In this paper, issues for data forwarding in wireless relay networks are discussed and
highlighted. It is found that multi-hop relay has the problem of inefficient processing and
buffering by analyzing conventional data forwarding schemes in 802.16j MR system.
Through the discussions, introducing tunnel headers can prevent RSs to process data
redundantly while reusing MAPs can avoid buffering data unnecessarily. However, none of
the convention schemes solve both the problems at the same time.
A new simple forwarding scheme is brought out to tackle the addressed issues. The concept
of burst switch is provided with end-to-end relay indications, and RSs can filter data
efficiently by the proposed burst CID. Furthermore, data associated with the CID are
switched between RSs with simple CID matching. The performance evaluations also prove
stable and superior performances of the proposed scheme in various network and traffic
environments. Comparing with prior mechanisms, burst switch scheme also shows benefits
in both 2-hop and 5-hop cases. As a result, it can conclude that relaying with the proposed
method provides simple and efficient data forwarding in wireless multi-hop relay networks.
6. Future research
Apart from the overhead issues, forwarding data in a wireless relay network will suffer
from high packet error rate. The issue is also crucial for a wireless multi-hop network since
packets during relay will be corrupted due to unreliable wireless medium. It is expected that
data forward in multi-hop environments would make the situation worse. Although

253
Automatic Repeat reQuest (ARQ) mechanism is proposed to maintain the wireless
communications reliability, it has been shown that conventional ARQ (Fairhurst & Wood,
2002) cannot overcome the packet error problem efficiently in wireless multi-hop
circumstances. Moreover, how a RS participates in error control would impact data
forwarding performance directly. Based on error control behaviours, the relay ARQ
mechanisms can be classified into three types and are shown in figure 16. The first type is
End-to-End Relay ARQ, in which RSs relay packets toward a destination and involve
nothing on controlling errors between BS and SSs. The second one is Hop-by-Hop Relay
ARQ, and RSs in the scheme trigger feedbacks and retransmissions in per-hop basis. The last
type is Two-Link Relay ARQ, which divides an end-to-end transmission into two links and
maintains the communications in the links separately. To utilize limited radio resource and
provide enhanced relay services, relay ARQ transmissions shall be investigated in the
future.

BS RS RS SS RS
Internet
L3
L2
L1
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L3
L2
L1
L2
L1
L2
L1
L2
L1
L2
L1
L2
L1
L2
L1
L2
L1
L2
L1
(a) End-to-End ARQ
(b) Two-Link ARQ
(c) Hop-by-Hop ARQ

Fig. 16. Wireless Relay ARQ Transmissions
7. References
Nosratinia, A. ; Hunter, T.E. ; Hedayat, A. (2004). Cooperative communication in wireless
networks, IEEE communication magazine, Vol. 42, Issue 10, Oct. 2004, pp. 74-80.
Mohr, W. (2005). The WINNER (wireless world initiative new radio) project-development of
a radio interface for system beyond 3G, Proceedings of Personal, Indoor and Mobile
Radio Communications, pp. 501-504, July 2005, Berlin, Germany.
Doppler, K. ; Osseiran, A. ; Wodczak, M. ; Rost, P. (2007). On the Integration of Cooperative
Relaying into the WINNER System Concept, Proceedings of Mobile and Wireless
Communications Summit 2007, pp. 1-5, July, 2007, Budapest, Hungary.
Soldani, D. ; Dixit, S.(2008). Wireless relays for broadband access [radio communications
series], IEEE communication magazine, Vol. 46, Issue 3, March 2008, pp. 58-66.

254
Tao, Z. ; Teo, K.H. ; Zhang, J.(2007). Aggregation and concatenation in IEEE 802. 16j mobile
multihop relay (MMR) networks, Proceedings of IEEE Mobile WiMAX Symposium,
pp. 85-90, March 2007, Orlando, Florida.
IEEE Std 802.16e
TM
-2005. IEEE Standard for Local and Metropolitan Area Networks Part
16: Air Interface for Fixed Broadband Wireless Access System, Amendment 2:
Physical and Medium Access Control Layers for Combined Fixed and Mobile
Operation in Licensed Bands, IEEE Computer Society and the IEEE Microwave Theory
and Techniques Society, February 2006.
IEEE Std 802.16j
TM
-2009. IEEE Standard for Local and Metropolitan Area Networks Part
16: Air Interface for Fixed Broadband Wireless Access System, Amendment
1 :Multiple Relay Specification, IEEE Computer Society and the IEEE Microwave
Theory and Techniques Society, June 2009.
Fairhurst, G.; Wood, L. (2002). Advice to Link Designers on Link Automatic Repeat reQuest
(ARQ), RFC3366, IETF Network Working Group, August 2002.
Wang, S.H. ; Yin, H.C. ; Sheu, S.T. ; Chang, C.S. (2009). Efficient Data Foorwarding Schemes
for IEEE 802.16j Multi-hop Relay Networks, Proceedings of Mobile WiMAX
Symposium 2009, pp. 24-29, July 2007, CA, US.
Chun, Nie.; Korakis, T. ; Panwar, S. (2008). A Multi-Hop Polling Service with bandwidth
Request Aggregation in IEEE 802.16j Networks, Proceedings of Vehicular Technology
Conference, 2008, pp. 2172-2176, May 2008, Singapore.
Te, Z. ; Hua, Y. ; (2005). On Link Layer Policies of Data Forwarding over Wireless Relays,
Proceedings of Military Communications Conference, pp. 2138-2144, Oct. 2005, Atlantic,
US.
Wirth, T.; Venkatkumar, V.; Haustein, T.; Schulz, E.; Halfmann, R.; (2009). LTE-Advanced
Relaying for Outdoor Range Extension, Proceedings of Vehicular Technology
Conference, pp. 1-4, September. 2009, Anchorage, Alaska, USA.
Kyungmi, P.; Kang, C.G.; Chang, D.; Song, S.; Ahn, J.; Ihm, J.; (2009).Relay-enhanced
Cellular Performance of OFDMA-TDD System for Mobile Wireless Broadband
Services, Proceedings of Computer Communication and Networks, pp. 430-435, August
2009, Hawaii, US.
Zhao, Z. ;Fang, X. ;Long, Y. ;Hu,Z. ; Zhao, Y. ;Liu. Y. ;Chen, Y. ; Qu, H. Xu, L. ;(2010). Cost
Based Local Forwarding Transmission Schemes for Two-Hop Cellular Networks,
Proceedings of Vehicular Technology Conference, pp. 1-5, May 2010, Teipei, Taiwan.
14
Experiments of In-Vehicle Power Line
Communications
Fabienne Nouvel
1
, Philippe Tanguy
1
, S. Pillement
2
and H.M. Pham
2

1
Laboratory IETR Rennes,
2
University of Rennes 1, Irisa Labs,
France
1. Introduction
The omnipresence of ECU (electronic control units) in vehicles has lead the automotive
industry to face a great challenge in its transition from mechanical engineering towards
mechatronical products. The X-by-wire and X-tainment applications involve efficient
networks that allow bus sharing while reducing both cabling costs, number of wires and
connectors.
This chapter deals with the embedded in-vehicle networks and the use of emerging
technologies combining different communication systems like power line communications
(PLC) and/or wireless communications and pushing to a dynamic configuration of both
networks and ECU.
The ECUs that replace mechanical or hydraulic systems require secure and specific bus for
communication. In order to exchange information between sensors and actuators, different
networks have been proposed, from low data rate up to high data rate namely LIN, CAN,
FlexRay. In section 2, these networks are presented identifying their strengths and possible
drawbacks. As a result of using these fieldbuses, the cost of advanced systems should
plummet. Furthermore, X-by-wire systems do not depend on conventional mechanical or
hydraulic mechanisms. In (Len & Hefferman, 2001), the authors demonstrate the advantages
of X-by-wire and embedded networks.
Considering these specific domain embedded networks, we can observe that each solution
uses its specific wires and communication system. The growth of the complexity leads to the
necessity to commit to a limited set of networks which answers to these multiple
applications. An attractive solution to reduce the wires is the power line communication
(PLC) using the power lines (12/42V) to transmit both the power and the messages without
functional barriers domain. It can answer the vehicles requirements namely cost, decrease of
the amount of wires, flexibility and bandwidth. Section 3 is dedicated to PLC systems.
Nowadays, this technique is already proposed for domestic uses (Ribeiro et al., 2006). In
vehicle PLC seems to be a promising technology and has numerous advantages; it could
reduce the weight of wires, the amount of splicing, and simplify cables bundle and the
networks between ECU. The background and current studies are first addressed in Section
3. Although high data rate and flexibility obtained for indoor domestic PLC are proven, it is
not possible to apply them directly to cars because the geometrical characteristics and wires

256
topologies are totally different. Moreover, the in-vehicle PLC channels are affected by the
variable activation schedules of electrical functions, such as brakes, indicators, etc, which
produce sharp modifications in the circuit's load impedances over brief time intervals, as
presented in (Lienard et al., 2008). To optimize high bit-rate communication, the PLC
channel transfer function must be carefully studied because it is frequency selective.
Previous studies show that the promising techniques are based on wide bandwidth
transmission, such a spread spectrum or orthogonal frequency division multiplex (OFDM).
Section 4 provides a description of the experimentations using the existing DC electrical
wires. The discussion is framed by describing in particular the area of PLC applications.
Results for different cars configurations are presented and demonstrate the feasibility of
PLC for automotive.
Section 5 will complete the chapter by describing other alternative solutions for in-vehicle
communication based on wireless communications such as ZigBee, Wireless USB, Wifi .
These technologies have been adopted for V2V, R2V and can be extended to in-vehicle
communication. Different solutions are proposed, giving a new perspective for ECU
communications, both for cars to cars and/or intra-cars communications.
In order to propose more flexibility with these different communication networks, new
electronic architecture need to be adopted to reduce the size of ECU. One solution is based
on dynamic reconfiguration. Section 6 will analyze this new concept for our embedded
system. Reconfigurable systems are already proposed for video driver assistance (Claus &
Stechele, 2010). The reconfiguration can increase both safety and flexibility. An ECU can
migrate tasks from one node to another. Furthermore, this functionality can be extended to
network architecture: according to the channel, the ECU loads, the modem can be
dynamically reconfigured to offer seamless communication between ECUs.
2. In-vehicle networks: overview of embedded solutions
Various vehicle buses for different tasks of communications between ECU are used today
according to their area of application (Navet, 2008). These embedded networks have both
increased the functionality and decreased the amount of wires. However, the usage of
different wires for the different networks still has the disadvantage of heavy, complex and
expensive. Among these networks, three of them are prevailing:
- the local interconnect network (LIN, the lowest data rate) : proposed by manufacturers,
the local interconnect network is used in on-off devices such as car seats, door locks,
rain sensors, ..
- the control area network (CAN, medium data rate) developed by BOSCH, is currently
the most widely used vehicular network. A typical vehicle can contain two or three
separate CAN networks operating at different transmission rates, from 125 Kbps up to a
higher-speed at 1 Mbps for more real-time-critical functions.
- the FlexRay is proposed for X-by-Wire applications which require higher data rate(10
Mbps) and safety. FlexRay is a fault-tolerant protocol designed for high-data-rate,
advanced-control applications. X-by-wire systems replace the mechanical control
systems with electronic component.
Figure 1 illustrates the embedded network architecture. We can observe that these networks
have a hierarchical structure.
Experiments of In-Vehicle Power Line Communications

257

Fig. 1. Network architecture (from http:// www.freescale.com)
2.1 The Local Interconnect Network: LIN
Conceived in 1998, the LIN network (LIN, 2003) is an inexpensive slow and serial bus used
for distributed body control electronic systems in vehicle. It enables effective
communication for sensors and actuators where bandwidth, speed and versality are not
required (i.e. inside mechatronic based subsystems generally made of an ECU and its set of
sensors and actuators). LIN is commonly used as a sub bus for CAN and FlexRay. A LIN
network is based on one master node and LIN slaves (up to 16 over 40 meters line length).
The master node decides when and which frame shall be transmitted according to the
schedule table containing the transmission order. At the moment a frame is scheduled for
transmission, the master sends the header inviting a slave node to send its data in response.
Any node interested can read a data frame transmitted on the bus. The reliability of LIN is
high but it does not have to meet the same levels as CAN. The LIN can be implemented
using just a single wire, while CAN needs two. The physical layer (PHY) supports a data
rate equal to 20 Kbps (due to electromagnetic limitations) but other transmission supports
enabling higher data rates are possible. LIN is widely used in middle range cars but it can
not support high data rate as required by devices like portable DVD players or multimedia
applications.
2.2 The Control Area Network: CAN
The CAN (CAN, 2009) is a widely communication fieldbus used in automotive and other
real time applications. It is a serial communications protocol which efficiently supports
distributed realtime control with a middle level of security. In automotive ECUs, sensors,
anti-skid-systems, etc. are connected using CAN with bit rates up to 1 Mbps. However, in
todays car, CAN is used as an SAE (Society of Automotive Engineers) class C (classification
defined in J2056/2 Survey, 1994)) network for real time control in the powertrain and

258
chassis domains (at 250 or 500 kbps). It is also implemented as an SAE class B network for
the electronics in the body domain (at 125 kbps).
In CAN, data are transmitted in frames containing up to 8 bytes of data and a number of
control bits. A CAN frame is labelled by an identifier whose numerical value determines the
frame priority. Depending on the CAN format (standard CAN 2.0A or extended CAN 2.0B)
the size of the identifier is 11 bits (CAN 2.0A) or 29-bits (CAN 2.0B). Between CAN frames
sent on the bus, there is also a 3 bit inter-frame space. The standard CAN frame format is
depicted in Figure 2.

Fig. 2. The CAN frame (from (Nolte, 2006))
Regarding the CAN MAC layer, CAN is a collision-avoidance broadcast bus (CSMA/CA for
carrier sense multiple access with collision avoidance), which uses deterministic collision
resolution to control access to the bus. It implements a fixed-priority based arbitration
mechanism that can provide real time guarantees and that is amenable to timing analysis.
As distributed real time systems become more and more complex, the computing power is
steadily growing, and the number of ECUs attached to CAN buses is growing. Thus CANs
maximum speed of 1 Mbps can lead to performance bottlenecks. Hence, methods for
increasing the achievable utilisation are needed, e.g., novel analysis methods that allow
increased utilisation while guaranteeing timing requirements to be fulfilled, and novel
approaches to schedule CAN.
2.3 The FlexRay protocol
X-by-wire systems need fault-tolerant communications with predictable message
transmissions and low jitter. This is traditionally solved using TDMA technologies, thanks
to their predictable nature. FlexRay (FlexRay Consortium, 2009) is a TDMA communication
system developed by a consortium founded in 2000, including both car and semiconductors
manufacturers. FlexRay is a fault-tolerant protocol designed for high-data-rate, advanced-
control applications. FlexRay is considered by manufacturers as the backbone network for
the other networks like CAN or LIN Currently, FlexRay can handle communications at 10
Mbps. An overview of the FlexRay frame format is given in Figure 3. The frame consists of
three segments: the header segment, the payload segment (up to 254 bytes of data), and the
trailer or CRC segment.
Communication is done in a communication cycle consisting of a static part and a dynamic
part, where each of the parts may be empty. The sending slots are represented through the
identifier (ID) numbers that are the same on both channels. The sending slots are used
deterministically (pre-defined TDMA strategy) in the static part. In the dynamic part there
can be differences in the phase on the two channels. Nodes that are connected to both
channels send their frames in the static part simultaneously on both channels. An interesting
feature of FlexRay is that it can provide scalable dependability i.e., the ability to operate in
configurations that provide various degrees of fault tolerance..

259

Fig. 3. FlexRay frame (from (FlexRay Consortium, 2009))
However, this network still uses specific wires, which do not achieve compatibility with
other networks. So, gateways are necessary to transfer information from ECUs connected to
a domain network to ECUs connected to other networks. Those gateways could introduce
latency, errors, bottlenecks, and so on.
Other fault-tolerant networks have been developed, namely TTP and TT-CAN, but it seems
they are not the best choice for automotive manufacturers due to limited flexibility, high
costs and conflicting interests (Nolte, 2006).
Considering these specific embedded networks, we can observe that each solution uses its
specific wires and communication system. We can see the wide diversity of the solutions
and the necessity to find a limited set of networks which answers to the growing of the
multiple applications and requirements. Furthermore, gateways are necessary to switch
from one network to another one. This increases the propagation time and does not
guarantee real time. One idea to avoid the growth of wires would be to use the PLC
technology that is currently developed for indoor AC networks to transmit information over
the 12V power distribution (Rubin, 2002). The possible applications of automotive PLC are
very wide, extending from low-speed data buses for activating actuators to high-speed
multimedia applications.
3. Power Line Communication (PLC)
Many studies are carried out on in-vehicle PLC and focus both on channels, impulsive
noises, drivers and protocols. The in-vehicle networks have reduced the number of wires,
allowing communication between different typologies of electronic systems (safety devices,
entertainment devices, and power train electronics). Additional cost reduction can be
accomplished by adopting PLC approach. PLC can be considered to provide the physical
layer for serial communications among ECU using for example LIN or CAN transceivers. In
this case, the dedicated bus is eliminated. However, PLC can provide both the physical and
MAC layers, allowing full compatibility between any ECU. This section considers first the
different indoor PLC solutions. Then we focus of in-vehicle PLC driver solutions.
3.1 Indoor PLC
In 2000, a coalition of manufacturers has established a new protocol HomePlug 1.0 that
enables the establishment of an Ethernet-IP class network over power line channels (Home
Plug V1.0, 2009). The HomePlug process is based on an OFDM technique (Bahai et al., 2004)
whose major advantage for the embedded PLC application is to cope with the frequency

260
selectivity of the power line channel caused by the multiple reflections of the loads
connected to the power grid and by the coupling to the other cables placed in the same
bundle. The modulation is based on 128 subcarriers equally spaced from 4.3 MHz to 25
MHz, in conjunction coding applied before differential encoding. HomePlug uses
CSMA/CA protocol to access to the network. Figure 4 represents the PHY frame format.

Fig. 4. HomePlug V1.0 PHY frame format (from (HomePlug, 2009))

Table 1. PHY and MAC layers of current PLC solutions (from (OMEGA, 2008))
More recently, the HomePlug AV (HPAV) has been introduced and will be the second major
standard released by the HomePlug Powerline Alliance (Gavette, 2006) (Afkhamie et al.,

261
2005). The main HomePlug AVs objective is to distribute multi-media content within the
house as well as data. The PHY layer still operates in the frequency range of [2 28] MHz
and provides a 200 Mbps PHY channel rate (150 Mbps net information rate).
Long OFDM symbols with 917 usable carriers (tones) are used in conjunction with a flexible
guard interval. Modulation densities from BPSK) to 1024 QPSK are independently applied
to each carrier based on the channel characteristics between the transmitter and the receiver.
Experimental systems of HPAV have been field tested in houses, suggesting that on average
HomePlug AV system achieves 10 times the data rate of a HomePlug 1.0 system.
At the same time, The HD-PLC (High Definition Power Line Communication) (Galli, 2008)
solution has been proposed by Panasonic. It is based on a specific OFDM modulation called
Wavelet-OFDM which exploits the Wavelet transform combined with 2 to 16 PAM
modulations. The Wavelet OFDM achieves highly efficient transmission with characteristics
that exceed even FFT-based OFDM systems. Wavelet OFDM features greater speed
efficiency and forms a deeper "flexible notch" that prevents interference with shortwave and
other broadcasts. No guard interval is included. The MAC layer uses the hybrid TDMA and
CSMA/CA protocols synchronized with the AC line cycle. Table 1 summarizes the current
PLC solutions.
Todays, the HomePlug Alliance, HD-PLC alliance and the IEEE (IEEE P1901, 2008) are
strongly committed to delivering a single mature solution that will be endorsed by the IEEE
P1901 work group as the baseline standard. A standard for high speed (over 100 Mbps at the
physical layer) communication will be proposed and will use transmission frequencies up to
100 MHz. These PLC solutions could be investigated in an automotive environment.
3.2 In-vehicle PLC
Although high data rate and flexibility obtained for indoor domestic PLC are proven, it is
not possible to apply them directly to cars because the geometrical characteristics and wires
topologies are totally different. PLC can be considered for the PHY layer only or for the
MAC and PHY layers. These two configurations are considered below.
In (Benzi et al., 2008), the authors focus on the issues that need to be addressed when
introducing PLC in vehicle. Three main domains need to be covered: the physical (PHY)
layer, the data link layer and the performances. In order to answer to them, the properties of
the automotive in board PLC supply networks have been investigated (Huck, 2005) (Arabia
et al., 2006), (Degardin et al., 2006), (Mohammadi et al., 2009). The results show the insertion
losses over the [0-30] MHz bandwidth are about -15 dB and -36 dB in the frequency range
[0.500-30] MHz. The noise measurements show an increasing background noise in the
frequency ranges [0-100] MHz, especially at frequencies less than 10 MHz, the peaks of the
noise could be in the range [-90 dBm/Hz; -40 dBm/Hz]. Varying space between cabling and
car body results in changing behaviour of the whole system. A new wiring harness structure
is proposed in (Benzi et al., 2008), based on a star structure using active star couplers. The
transfer function of this wiring seems to be more flat in the range between 150 and 250 MHz.
However this solution needs to re-organize all the harness, which is different from one
vehicle to another one.
Considering first the PLC for the PHY layer, a power line communication system over 12 to
42 V wires combining the LIN protocol and a PLC driver has been proposed ( De Caro,
2009). In order to avoid interferences between the master and slaves LIN nodes, two
different transmission modes have been adopted, one based on BPSK for master to slave

262
transfers, while slave mode exploits a BASK modulation. The modulation must support data
transfer at 10 Kbps, while the accepted conducted emission limits need to be less than 53
dBv in the [1-30] MHz band. Two carriers have been selected, one at 100 KHz for low
power modules and one at 2 MHz for high power modules. Although this LIN and PLC
transceiver is an attractive solution, the data rate remains under 10 Kbps that is not
convenient for X-by-wire applications.
A similar approach has been proposed for CAN protocol by many authors (Yamar, 2009),
(Silva et al., 2009) (Beikirch et al., 2000). The Yamar solution implements CAN and PLC
using the DC-BUS technology with different bit rates up to 1.7 Mbps. It uses narrow band
channels with a center frequency between [2-12] MHz. The DC-BUS protocol uses the
CSMA/CA multiplex mechanism allowing bidirectional communication up to 16 nodes. In
addition, this CAN-PLC solution can be used as a redundant channel for the CAN protocol.
However, this solution still does not answer to data rate over 10 Mbps.
Additional PLC drivers combining MAC layers have been presented in (Benzi, 2008). The
commercial solutions are available for automotive but to our knowledge not implemented
yet in vehicles.
More recently, PLC in electric vehicles has been studied in (Bassi et al., 2009). One can think
that the requirements of such communication system within an electrical car differ from a
fuel car. An experimental setup has been built. It uses 2 ECUs and 2 DCB500 transceivers to
modulate the DC-line. The DCB500 transceivers feature PLC communication over DC-line
with a bit rate up to 500 Kbps. The conducted and irradiated emissions show substantial
compatibility, except for the lower end frequencies (under 1 MHz) where significant peaks
are highlighted. In addition, different channel measurements in electric cars have been
carried out in (Barmada et al., 2010). Different cases are considered (front to/from rear part)
with different vehicles configuration (position key, battery,). As for fuel vehicle, the
channels are very frequency selective in the [0-30] MHz. We can conclude that the fuel and
electric vehicles seem to have similar behaviours in term of frequency channel and noise for
PLC applications.
Another solution for PLC is to consider both the MAC and PHY layers. Considering the
channel measurements, the candidate techniques for in-vehicle PLC are spread spectrum
combined with code division multiple access (CDMA) (Nouvel et all, 1994) and OFDM.
OFDM allows high data rate and outperform CDMA performances in term of throughput.
and frequency selectivity.
Experimentations using indoor OFDM PLC modems have been carried out and presented in
detail in previous studied presented in (Gouret et al., 2006), (Gouret et al., 2007), (Nouvel et
al., 2008), (Degardin, 2007) and more recently in (Nouvel et al., 2009A). The results are very
promising. Data rate up to 10 Mbps/s can be achieved in the [0-30MHz] bandwidth. The
solutions are based on HPAV standards. In (Nouvel et al., 2008) two PLC modems have
been tested: SPIDCOM (Spidcom, 2008) and DEVOLO modems. In the SPIDCOM modems,
the OFDM modulation is based on 896-carriers from 0 to 30 MHz divided into 7 equal sub-
bands. The MAC layer provides a mechanism based on TDMA and CSMA/CA is also
available. The PHY and MAC layers are similar to the HPAV ones but differ in some points:
number of sub-bands, equalization, and synchronization. With these SPIDCOM modems, an
8 Mbps is achieved with a transmitted power of -50 dBm. With a higher level (-37 dBm), we
achieve about 12 Mbps. For multi-media applications, this rate can be sufficient, but
decreases rapidly according to the loads. Then measurements have been carried out with

263
DEVOLO PLC modems. They comply with HPAV and support data speed of up 200 Mbps
in a range of 200 meters within a household grid. For intra-car communications, the power
supply and the coupling have been modified to take into account the DC channel.
Additional measurements are presented in next section. Figure 5 illustrates the spectrum of
the transmitted signal over the DC line.

0
10
20
30
40
50
60
70
80
90
100
0 5 10 15 20 25 30 35 40
f (MHz)
d
B
A
Devolo
Spidcom
classe 1
classe 2
classe 3
classe 4
classe 5

Fig. 5. HPAV and Spidcom spectrum over DC line
Beyond these promising results, the choice of the modulation parameters will be driven by
the PLC channels and optimized with regards to the bandwidth, the modulation technique,
the coding rate, the guard interval, and so on. This discussion is presented in the next
section.
4. In-vehicle measurements
In this section we deal with in-vehicle PLC measurements. In a first time we show some
results about real PLC transmissions. Indeed, we have decided to test the feasibility to adapt
indoor PLC modems in car. Then, we study in more details the in-vehicle PLC channel with
different measurements about the transfer function and the noise. To achieve the capacity of
the channel through the cables for PLC, many transfer functions between nodes in the
vehicle have been measured. Noises have also been considered.
4.1 In-vehicle PLC transmissions
4.1.1 Data rates measurements testbed
We have tested two indoor PLC modems complying with the standards HPAV and HD-
PLC in one car. We have measured throughputs at different points on a gasoline Peugeot
407 SW.
The Figure 6 illustrates the different points used during the throughput measurement.
Several scenarios have been used:
1. Car with engine-turned off

264
2. Car with engine-turned but not moving
3. Car with engine-turned but not moving and effects of lightning, warnings, radio,
windscreen wiper, electric windows
4. The car in motion and the effects of the equipments like in 3)

Fig. 6. Measurement scheme: the different uppercases represent the measurement points
The measurements have been achieved with two PLC modems and two computers which
have been plugged into the different points shown Figure 6. Therefore, we have measured
the TCP throughput between two points with two modems and two PC. The measurement
between points A and D has been called path AD. The throughputs are measured associated
with the payload ignoring headers. The throughput is also called Goodput according the
definition in section 3.7 of (Newman, 2009).
4.1.2 Results and discussion
Throughputs for different points have been studied and we can first observe a difference
between scenario 1) and the others. Figure 7 to 9 represent the throughput we obtain with
the two modems. Throughputs in Figure 7 are higher than 35 Mbps, and in Figures 8 and 9
more than 15 Mbps are achieved for all paths.

Fig. 7. HPAV and HD-PLC throughputs comparison

265
For scenario 2), 3) and 4) we remark that the HPAV has the best performances. Moreover,
we can observe short variations between the scenarios for the two indoor standards.
Furthermore there is a throughput difference according to the path in-vehicle. Indeed, we
can see that the path HD has throughput higher than all the others.
Indoor PLC standards have been designed according indoor channel characterization.
Moreover, the power level of the transmitted signal has been chosen according the indoor
CEM constraints. In fact, to respect the vehicle CEM it has been said in (Degardin et al.,
2007) that the power level of transmitted signal should be between -60 dBm/Hz and -80
dBm/Hz. This specific point must be taken into account for next PLC in-vehicle
transmission. That's why measurements on several vehicles have been achieved and are
discussed in the next subsection.

Fig. 8. HPAV throughputs for different paths in-vehicle for scenario 2), 3) and 4)

Fig. 9. HD-PLC throughputs for different paths in-vehicle for scenario 2), 3) and 4)

266
4.2 In-vehicle channel measurements
In order to design a future PLC modem it is necessary to study the PLC in-vehicle channel.
Here the transfer function and the background noise is studied.
Additional measurements have been performed on recent vehicles for two classes of paths:
front to front and rear to front (Tanguy et all, 2009). Figure 10 and 11 illustrate the results
according to our testbed (Figure 6). In order to analyze the DC PLC architectures, additional
transfer functions are measured on four different vehicles. The vehicles are classified
according to: the number and type of ECUs, the length of wires, the combustion engine.
4.2.1 Measurement testbed
The S-parameters are recorded using a full 4 ports Vector Network Analyzer (VNA) and a
PC interfaced to remote the device. We record the S-parameters during about 10 minutes
while the car is moving. The S-parameters are recorded about every 10 seconds for the 3
different paths: GF, GH and HD. Compared with the previous subsection we have introduce
a new measurement point called G which is for the most of vehicle tested a cigar lighter
receptacle. These paths have been chosen in order to analyze the differences between front
to front and rear to front.
Regarding the noise, the same points have been considered: G, D, F and H. Two different
noise studies have been carried out. The first consists of the measurement of the power
spectrum at each point during 10 minutes every 10 seconds with the vehicle moving. The
second is a measurement in the time domain. In fact, a digital storage oscilloscope (DSO) has
been used to record at each point the signal over the DC line. With this testbed we are able to
record two signals at two different points in the same time. Thus, we can observe the level of
noise at two different points simultaneously. Finally, the measurements have been performed
on a Peugeot 407 SW gasoline and diesel, a Renault Laguna II Estate and a Citron C3.
4.2.2 Results & discussion
Figure 10 and Figure 11 show an example of time and frequency responses for the three
paths GF, GH and HD and for a measurement bandwidth of [1-31] MHz. The impulse
responses have been calculated with the inverse Fourier transform of complex parameter S21.

Fig. 10. Impulse response for 3 paths GF,GH,HD on 407SW gasoline

267

Fig. 11. S21 for 3 paths GF, GH and HD on 407 SW gasoline

Min Max Mean Std
407 gasoline 391.8 KHz 832.6 KHz 533.8 KHz 89.9 KHZ
407 diesel 538.7 KHz 881.6 KHz 666 KHz 81.6 KHz
Laguna II 4.3098 MHz 4.8976 MHz 4.7163 MHz 142.8 KHz
BC0.9 GF
C3 440.8 KHz 1.3713 MHz 1.1587 MHz 143.9 KHz
407 gasoline 1.3713 MHz 2.1059 MHz 1.7578 MHz 190.3 KHz
407 diesel 97.9 KHz 1.0775 MHz 748.3 KHz 227.3 KHz
Laguna II 1.0775 MHz 1.2734 MHz 1.1443 MHz 45.6 KHz
BC0.9 GH
C3 489.8 KHz 1.5182 MHz 1.0591 MHz 331 KHz
407 gasoline 1.8121 MHz 2.057 MHz 2.006 MHz 40.8 KHz
407 diesel 685.7 KHz 734.6 KHz 712.6 KHz 24.5 KHz
Laguna II 685.7 KHz 832.6 KHz 744 KHz 31.9 KHz
BC0.9 HD
C3 881.6 KHz 1.0775 MHz 995.8 KHz 46.4 KHz
Table 2. Coherence bandwidth (BC0.9) for 3 paths (GF, GH and HD) and for 4 different
vehicles
In a previous study on in-vehicle PLC (Lienard et al., 2008) a delay spread under 380 s and
a coherence bandwidth greater than 400 KHz has been found. Moreover, in Table 2, we
observe the coherence bandwidths are different from one vehicle to another and from one
path to another. This means that the modulation must be adaptive.
Regarding the average attenuation we can also observed differences between the different
paths. For example, the Renault Laguna II Estate has a mean average attenuation of 9 dB for
the path GF, 31.6 dB for GH and 31.5 for HD. But the 407 SW gasoline has a mean average

268
attenuation of 40.1 dB for the path GF, 40.4 dB for GH and 24.4 for HD. Otherwise, we have
a maximum average attenuation of 69.3 dB for the path GH of the 407 SW diesel and a
minimum average attenuation of 5.8 dB for the path GF of the Laguna II.

Fig. 12. Noise measured with a spectrum analyser for 4 different paths on a Peugeot 407 SW
gasoline

Fig. 13. Spectrogram computed with the DSO recording at point G measured on a Peugeot
407 SW gasoline
To optimize the modulation parameters, we have to consider the noise. Figure 12 represents
an example of noise measurement with a spectrum analyzer for 4 different points in a
Peugeot 407 SW gasoline. We observe an increase of noise for some frequencies in the
bandwidth [0 5] MHz. Moreover we can see narrowband noises. Like in (Yabuuchi et al.,

269
2010) we have applied to noise recordings in-vehicle a time frequency analysis. In Figure 13
we show an example of spectrogram computed with the DSO recording at point G
measured on the same vehicle. We have computed the spectrogram with short-time Fourier
transform where an Hamming window of length equal to the length of HPAV OFDM
symbol (40.96 s) and an FFT size of 3072 points like in HPAV standard.
In Figure 13 we can observe that in the bandwidth [0 5] MHz the noise is constant during
the time of the recording. Therefore, in the case of a multi-carriers modulation transmission
in the bandwidth [2-30] MHz some subcarriers will be affected during all the transmission
time.
We have observed that the average attenuation, the coherence bandwidth and the RMS
delay spread are very different according the vehicles, the paths in-vehicle and the paths
between vehicles. We verified the capacity for each paths of each vehicles with the
parameters of the Table 3 according to

1
2
0
log (1 )
N
i
C f SNR
= +
(1)
with
f
the subcarrier bandwidth, SNR_{i} = (H_{i}
2
.Pe/Pn) the signal to noise ratio per
subcarrier , Pe is the PSD of the emitted signal and Pn is the PSD of the AWGN noise.

Parameters Values
Fmin 1 MHz
Fmax 31 MHz
Subcarrier N=1228
FFT/IFFT 3072
f
24.414 KHz
PSD of noise (Pn) - 120 dBm/Hz
PSD of signal (Pe) -60 dBm/Hz
Table 3. Simulation parameters: FFT/IFFT and
f
values are parameters used by the HPAV
standard
The results show the minimum of the average capacity is about 190 Mbps for the path GH
of the Peugeot 407 SW diesel and the maximum is about 507 Mbps for the path GF of the
Laguna II. We observed also differences between the paths and the vehicles.
The vehicles have not the same electrical topology. In fact, it depends on car manufacturer,
the size of vehicles, the number of ECUs ... Therefore the load on the electrical network, the
length of wires and the junctions between cables are different. We have several channels
which are different according the paths and the vehicles like we have shown with the
coherence bandwidth, the time delay spread, the channel gain and the capacities.
The multicarrier modulation seems to achieve good performances like we have seen during
the throughput measurement of HPAV and HD-PLC standards. In this study, only the

270
channel function transfer and the background noise have been studied. The impulsive noise
is an other important aspect to take into account (Umehara et al., 2010) and (Degardin et al.,
2008) for powerline communication. According to us the MAC/PHY layers must be
designed to take into account the differences between vehicles and the differences between
paths in-vehicle. Future work will be focus on the integration in a simulator of all the
channel measurements (transfer function, background noise, narrowband interference and
impulsive noise) in order to optimize the modulation scheme.
5. In-vehicle wireless communications
The interest in wireless networking has grown significantly due to the availability of many
wireless products. Looking at in-vehicle communications, more and more portable devices,
e.g., mobile phones, laptop computers and DVD player can exploit the possibility of
interconnection with the vehicle. Wireless communication could be an attractive solution to
reduce the number of cables and disturbances in cars. We have reviewed potential wireless
solutions, specifically two of them in (Nouvel et al., 2009A). We have performed tests similar
to PLC tests in order to qualify the channel in the 2.4 GHz band. Data rate measurements
show it is possible to achieve more than 10 Mbps/s in the vehicle, using also OFDM
technology. Additional studies have been carried out in (Nolte et al., 2009). The authors in
(Zhang et al., 2009) have conducted measurements in the [0.5 16] GHz band. One can
observe the different delay profile, different clusters, different paths and the impact of
passengers. Due to lake of space, it is not possible to describe all the measurements. And we
invite the interested readers to look at the papers and chapters.
6. From static to dynamic ECU and communication networks
Taking into account all these networks, from specific network up to PLC or wireless
combined with the constraint of flexibility and security, one attractive idea is to be able to
switch from one network to another one, without additional cost. If the main
communication fails, the ECU ( modem) can switch to the secondary protocol and continue
to run. Reconfigurable architectures based on FPGA may offer very flexible links inside a
vehicle. A dynamically reconfigurable system allows changing parts of its logic resources
without disturbing the functioning of the remaining circuit. This property can applied for
networks, in order to allow changing from one protocol to another one according to the
channel behaviour, errors, load, etc. This section will discuss about this new concept and
demonstrates how it can be integrated in vehicle.
Certain modern FPGAs offer dynamic and partial reconguration (DPR Dynamically and
Partially Reconfigurable) capability that allows to change dynamically one portion of the
FPGA without affecting the rest of the circuit. Currently, the Xilinx Virtex FPGAs (Xilinx,
Inc, 2008) are the only commercially available circuits supporting the DPR paradigm and
large applications implementation. Internal structure of a Xilinx Virtex5 is presented in
Figure 1. The main resources dispatched in the FPGA matrices are slices, DSP blocks
(DSP48E), memory blocks (BRAM), input/output (IO) banks, and Clock Management Tiles
(CMTs) as well as the reconfiguration interfaces, so called ICAP. Slices are the smallest
configurable elements constituted of LUTs (Look-Up Table), registers and logic gates. DSP
blocks offer a powerful set of processing elements for data applications.

271
The dynamic reconfiguration takes place in Partially Reconfigurable Region (PRR) which
can be partially reconfigured independently. Designing a dynamically self-reconfigurable
system always require the declaration of PRRs. A PRR is implemented statically despite the
fact that its content is dynamic. Thus, at runtime, dynamic reconfiguration can only take
place into the PRR. Communications between a dynamic task and its static environment is
assured through the bus macro interfaces. Bus macros are also specified statically.

Fig. 14. View of the Virtex5 5VSX50T captured from Xilinx PlanAhead design tool
The FPGA fabric is partitioned into one static logic and one or more partially recongurable
regions (PRRs). This fabric partitioning enables reconguration of a single PRR without
system interruption (the static region and other PRRs continue execution while only the
recongured PRR halts). Each PRR has a related partial bitstream and the reconguration
process can be done by sending this partial bitstream to the reconfiguration port. In modern
FPGAs, the reconfiguration is stored in SRAM based memory, leading to a weakness from a
reliability point of view.
Modern FPGAs, besides customary high-density reconfigurable resources, offer the
designers the possibilities of implementing programmable processors having features of
Commercial Off-The-Shelf (COTS) components (no need to modify processor architecture or
application software). Processors play the role of processing units, and one particular is used
as coordination units in the embedded system. Besides, processors are in charge of collecting
the data from peripherals and from the memory, process the data and send them to the
memory and to the peripherals. Also, processors manage the memory and initialize the
peripherals. Xilinx FPGA devices include two categories of processors: the hardcore
embedded processor (PowerPC) (Xilinx, Inc, 2004) and softcore processors (MicroBlaze,
PicoBlaze) (Xilinx, Inc, 2009). Hardcore embedded processors are hard-wired on the FPGA
die and their number is limited on each device. On the other hand, softcore processors use
recongurable resources, so the number that can be actually implemented depends on the
device size only. The MicroBlaze tasks can be classified into 2 types: hardware tasks and
software tasks. Hardware tasks are peripherals connecting to MicroBlaze. Software tasks are
software programs running inside MicroBlazes. Generally, hardware tasks are designed
using High Description Language HDL like VHDL, Verilog and software tasks are
programmed using C. Regardless of their design methods they are presented in our system

272
in compiled forms of binary files called bitstreams. A bitstream is the set of binary data
describing the circuits implemented on the FPGA, or in PRRs (partial bitstream). Shorten
term bitstream 1 will refer to all the bitstreams of FPGA1, idem with bitstream 2 for
FPGA2, .
The processor software context is a set of information needed to uniquely define the state of
the processor at a given moment. It could include the states of the processor registers, the
cache, the memory, etc. Saving and restoring all relevant values allow for processor context
switching and error recovery. The softcore processor MicroBlaze context is represented by
the 32-bit values of 32 General Purpose Registers and two Special Registers: the Program
Counter (PC) and the Machine Status Register (MSR).
A MicroBlaze task migration consists in migrating hardware task, software task and
restoring the software context. Hardware task migration requires the appropriate peripheral
to be added using dynamic reconfiguration. Software context is also migrated by dynamic
reconfiguration. And copying the saved software context into the related MicroBlaze
program memory does the software context recovery process.
Due to their exibility, FPGAs are attractive for mission-critical embedded applications like
automobile domain, but their reliability could be insufficient unless some fault-tolerance
techniques capable of mitigating soft errors are used. Dynamic partial reconguration
provide not only the flexibility in both hardware and software, but also further solutions
dealing with reliability problem in critical domains. The dynamic reconfiguration allows the
reloading of the defected module to the correct state and the re-execution of the attributed
tasks. It cans also re-distribute defected tasks in the faulty module to other processing units
in the system.
We present here the feasibility of integrating dynamic reconfiguration features into
automotive-aimed applications in which certain fault-tolerance degrees should be
maintained. In case of a fault occurrence, the system must be capable of react in real-time to
ensure the safety for driver as well as pedestrian. The reaction in this case can be the fast
fault detection and correction by loading the original configuration to put the faulty module
to the state as at start-up. It can also be the critical task migration from the defected module
to another module.
To define a new embedded automotive platform based on reconfigurable architecture, in
CIFAER (CIFAER, 2008) we advocate for the use of Radio Frequency and Power-Line
Communication for intra-vehicle communications (Nouvel et al., 2008). The communication
can be switched from one to the other by dynamically reconfiguring a defined
communication zone on an FPGA. These two modes offer very flexible links inside a vehicle.
Figure 15 shows the fault-tolerant multi-FPGA platform. The system consists of four FPGAs
connected together using two Ethernet communication (in future development one will be
based on PLC interface, while the other will be constructed on RF connections). The first
network is routed via a network switch, the other network form a ring topology for the
fault-tolerance purpose. The Ethernet protocol is built by Ethernet controller as MicroBlaze
hardware peripherals and LightWeight IP (LightWeight) as the software library. The lwIP is
an open-source stack using TCP/IP protocol, which can be easily adapted to PLC and
wireless modem. Each FPGA contain a fault-tolerant dynamic multi-processors system,
consisting of several MicroBlaze (Figure 16). Further details about this system architecture,
called FT-DyMPSoC, as well as the fault-tolerance schemes implemented can be found in
(Pham et al., 2009) and (Pham et al., 2010) for interested readers.

273

Fig. 15. Fault-tolerant multi-FPGA platform. Two communications networks are supported
for reliability purpose

Fig. 16. FT-DyMPSoC Architecture. Included in each FPGA this architecture insert fault-
tolerant mitigation schemes
On the overall system, each FPGA is interfaced with a memory that can be accessed by all
the processors inside the same FPGA. This memory is partitioned into three segments
(Figure 15):
- One for saving all the bitstreams and the software contexts of all the processors of this
particular FPGA.
- One for saving all the bitstreams of the next FPGA in the ring network.
- One reserved and used in case of failure occurrence in the system. This segment helps
to transfer the bitstreams and contexts between different FPGAs.
The memory segmentation guarantees the existence of at least one copy of all the bitstreams
over the whole network.
As we can see in Figure 17, the bitstream of each FPGA is present in its local memory and
also in the local memory of the previous FPGA in the ring topology. For example, FPGA1
stores its own bitstream 1 and and the bitstream 2, FPGA2 stores bitstream 2 and bitstream
3 These copies will be used in case of system failure, and permit fast context switching.

274

Fig. 17. Fault recovery strategies. Once the faulty FPGA is identified, the copies of the
bitstreams are exchanged in order to keep a valid copy of all the configurations
The fault-tolerance degrees are maintained at two levels in the system. The Intra-FPGA level
corresponds to the fault-tolerance strategy inside each FPGA, and is related to the design of
the FT-DyMPSoC system. The fault-mitigation strategy is realized using the connection
matrices algorithm (Pham, 2009), and fault are mitigated by using dynamic reconfiguration
at the processors level. The second level called Inter-FPGA level corresponds to the overall
system presented in Figure 15. To detect error in the overall network, all the FPGAs
exchange frequently among them detection frames. These frames contain the software
contexts of the four MicroBlazes of each FPGAs. On one hand, this helps detecting error in
the network. On the other hand, including the contexts within the detection frame will help
to resume the tasks of a faulty FPGA on another FPGA. During the exchange if the contexts
of one FPGA (i.e. FPGA3 in Figure 17) are not received by the others circuits, the FPGA3 is
declared faulty. There are 2 possibilities: the MicroBlaze 1 (supporting the interface to the
network) of FPGA3 is faulty, causing the communication lost of this FPGA, or the whole
FPGA3 is faulty. In order to distinguish these 2 possibilities, the secondary ethernet links is
used. FPGA2 and FPGA4 try to communicate with MicroBlaze 2 and 3 of FPGA3. If these
communications fails, the whole FPGA3 is declared defected, if not, only the MicroBlaze 1 is
defected.
If only one MicroBlaze inside one FPGA fails, we can manage this error thanks to dynamic
reconfiguration of this processor or by using task migration within the MPSoC system. The
error is managed at the FPGA level. If the whole FPGA fails the task migration concerns the
overall circuit. In this case the task of the FPGA3 needs to be dispatched across the
remaining circuits. If the system cannot manage all the tasks with one missing FPGA
priority needs to be defined and used to maintain critical services for example. In this case,
arbitration on the running tasks needs to be executed, and reconfiguration of the remaining
FPGA is launched.
If one FPGA is lost, we need to maintain the two bitstreams copy stored in the faulty FPGA.
For example, if the FPGA3 is lost (Figure 17), the copies of bitstream 3 and 4 are inaccessible
requiring a clone of bitstream 3 and bitstream 4. We propose here 2 strategies delivering the
bitstream 3 and 4 to other FPGAs.
1. The first strategy uses only the secondary communication media. We need to use
FPGA1 reserved segment as intermediate medium. First the bitstream 4 is copied from
FPGA 4 to FPGA1 reserved segment, then to FPGA2. Afterwards, bitstream 3 is copied
from FPGA2 to FPGA1, then to FPGA4.

275
2. The second strategy requires both communication media. Bitstream 4 is copied from
FPGA4 to FPGA1 using direct Ethernet link. Simultaneously, bitstream 3 is copied from
FPGA2 to FPGA4 using the primary Ethernet via the switch.
In case the Ethernet switch fails, all the primary Ethernet connections are defected; This
leads to a connection loss between all the FPGAS. At this moment all circuits switch to the
ring topology. The second network will then ensure proper operation of the overall system.
The use of redundancy of the network, coupled with the new dynamically reconfigurable
paradigm permits to construct highly reliable system.
7. Conclusion
In this chapter, an initial foreseeable solution with PLC has been presented to allow
communications and interoperability between embedded applications with different
requirements. PLC network answers both to cost, flexibility and throughput requirements.
Future work should be devoted to optimize both PLC modulations and ECU architectures in
order to minimize the number of cables and ECU etc. This implies rethinking the DC
bundles as rethinking the implementation of networks as independent domain.
Furthermore, it is possible to build a reconfigurable ECU for both application and
communication. This new concept will allow combining different network technologies. It
will answer to fault tolerance constraints, required in X-by-wire applications
This work has been carried out by the CIFAER project, supported by the ANR and by the
French Premium Cars competitiveness Cluster ID4car.
8. References
Afkhamie, K.H.; Katar, S.; Yonge, L. & Newman R. (2005). An overview of the upcoming
HomePlug AV standard, Proceedings of the IEEE International Symposium on Power
Line Communications and Its Applications, pp. 400404, 0-7803-8844-5, 6-8 Apr. 2005,
Vancouver, BC, Canada.
Arabia, E.; Ciofi, C.; Consoli, A.; Merlino, R. & Testa A. (2006). Electromechanical Actuators
for Automotive Applications Exploiting Power Line Communication, Proceedings of
SPEEDAM, pp. 909-914, 1-4244-0193-3, Taormina, 26-26 May 2006
Bahai, A.; Saltzberg, R. & Ergen, M (2004). MultiCarrier Digital Communications, ISBN: (HB)
0-387-22575-7, Springer NewYork
Barmada, S.; Raugi, M.; Tucchi, M. & Zheng, T. (2010). Powerline communication in a full
electric vehicle, Proceedings of IEEE International Symposium on Power Line
Communications and Its Applications, pp. 331-336, 978-1-4244-5009-1, Rio de Janeiro ,
28-31 March 2010
Bassi, E.; Benzi, F.; Almeida, L. & Nolte, T. (2009). Powerline communication in electric
vehicles, Electric Machines and Drives Conference, pp. 17491753, 978-1-4244-4251-5,
Miami, 3-6 May 2009
Beikirch, H. & Voss, M. (2000). CAN-transceiver for field-bus power line communication,
Proceedings of the International Symposium on Power Line Communications and Its
Applications, Limerick, pp. 257264, April 2000
Benzi, T.; Facchinetti, T. Nolte & Almeida L. (2008). Towards the powerline alternative in
automotive applications, Proceedings of Factory Communication Systems, pp. 259-262,
978-1-4244-2349-1, Dresden, 21-23 May 2008

276
CAN. (2009), Retrieved from official web site http://www.can-cia.org/ ( 2009)
CIFAER (2008), Communications Intra-vhicule et Architecture Embarque Reconfigurable,
French ANR Project 2008-2011. Available at web site www.agence-nationale-
recherche.fr/AAPProjetsOuverts
Claus, C.; Stechele W. (2010), AutoVision - Reconfigurable Hardware Acceleration for
Video-Based Driver Assistance, In: Platzner, Teich, Wehn (Editors): Dynamically
Reconfigurable Systems, ISBN 978-90-481-3484-7, Springer, 2010
De Caro S. & Testa, A. (2009). A Power Line Communication approach for body electronics
modules, Proceeding of Power Electronics and Applications, pp. 1-10, 978-1-4244-4432-
8, Barcelona, 8-10 September 2009
Degardin, V.; Laly, P.; Linard, M. & Degauque, P. (2006). Impulsive Noise on In-Vehicle
Power Lines: Characterization and Impact on Communication Performance,
Proceedings of IEEE International Power line Communications and Its Applications, pp.
222-226, 1-4244-0113-5, Orlando, 26-29 March 2006.
Degardin, V.; Lienard, M.; Degauque, P. & Laly, P. (2007). Performances of the Homeplug
PHY layer in the context of in-vehicle powerline communications, Proceedings of
Power Line Communications and Its Applications, pp. 93-97, 1-4244-1090-8, Pisa, 26-28
March 2007
FlexRay Consortium. (2009). FlexRay Communication System, Protocol Specification,
Version 2.0. Retrieved from: http://www.flexray.com
Galli, S.; Koga, H. & Kodama, N. (2008). Advance signal Processing for PLCs: Wavelet
OFDM, Proceedings of Power Line Communications and Its Applications, pp. 187-192,
978-1-4244-1975-3, Jeju city, Jeju Island, 2-4 April 2008
Gavette, S. & al. (2006). HomePlug AV Technology Overview, Technical report. Sharp
Laboratories of America. Retrieved fromdownload.microsoft.com/download.
Gouret, W.; Nouvel, F. & El Zein, G. (2006). Additional Network Using Automotive
Powerline Communication, Proceedings of International Conference on Intelligent
Transport Systems Telecommunications, pp. 1087-1092, 0-7803-9587-5, Chengdu, 26-29
March 2006
Gouret, W.; Nouvel, F. & El Zein, G. (2007), High Data Rate Network Using Automotive
Power Line, Proceedings of International Conference on Intelligent Transport Systems
Telecommunications, pp. 1-4, 1-4244-1178-5, Sophia Antipolis, 26-28 March 2007
IEEE P1901 (2008), IEEE P1901 Working group. Retrieved from web official web site
http://grouper.ieee.org/groups/1901/
Home Plug (2009). HomePlug Powerline Alliance. Retrieved from official web site
www.homeplug.org
Huck, T., Schirmer, J. & Dostert, K. (2005). Tutorial about the implementation of a vehicular
high speed communication system. in IEEE International Power line Communications
and Its Applications IEEE ISPLC , pp 162-166, 6-8 April, 2005.
J2056/2 Survey. (1994). J2056/2 survey of known protocols. In SAE Handbook. Warrendale,
PA: Soc. Automotive Eng. (SAE), vol. 2.
Leen, G; Hefferman, D. (2001). Vehicles without wires. Computing & Control Engineering
Journal, Volume. N 12(Issue 5), October (pp. 205-21).
Lienard, M.; Carrion, M.; Degardin, V. & Degauque, P. (2008). Modeling and Analysis on In-
vehicle power line communication channels, Proceeding of IEEE Transaction on
Vehicular Technology, vol 57, N2, pp. 670-679, 0018-9545

277
LIN Consortium. (2003). LIN Specification Package, Revision 2.0.Retrieved from
http://www.lin~subbus.org/( 2009)
Mohammadi, M.; Lampe L.; Lok, M.; Mirabbasi, S.; Mirvakili, M.; Rosales, R. & Van Veen, P.
(2009). Measurement study and transmission for in-vehicle power line
communication, Proceedings of IEEE Power Line Communications and Its Applications,
pp. 7378, 978-1-4244-3790-0, Dresden, 29-March -1 April 2009
Navet (2008), The Automotive Embedded Systems Handbook, Industrial Information Technology
series, Taylor and Francis , CRC Press, ISBN 978-0849380266, 2008.
Newman, D. (2009). Benchmarking Terminology for Firewall Performance, RFC 2647.
Nolte,T. ( 2006). Share-Driven Scheduling of Embedded Networks. University, Dissertation,
May 2006. Department of Computer Science and Engineering Mlardalen University
Vsteras, Sweden, Printed by Arkitektkopia, Vsteras, Sweden Distribution , 2006
Nolte,T. & Hansson H.( 2009). Wireless Automotive Communications, Internal Report
available at
http://ant.comm.ccu.edu.tw/course/97_ITS/1_HW1/0.Wireless%20Automotive%
20Communications.pdf ,2009
Nouvel, F.; El Zein, G. & Citerne, J. (1994). Code division multiple access for an automotive
area network over power-lines, Proceedings of IEEE Vehicular Technology Conference,
pp. 525529, 0-7803-1927-3, Stockholm, 8-10 June 1994.
Nouvel, F. & Maziro, P. (2008). X-by-Wire and intra-car communications: power line
and/or wireless solutions, Proceedings of International Conference on Intelligent
Transport Systems Telecommunications, pp. 443-448, 978-1-4244-2857-1, Phuket, 24
October 2008.
Nouvel, F; Gouret, W; Maziro (2009a), Automotive Network Architecture for ECUs
Communications, in Automotive Informatics and Communicative Systems: Principals in
Vehicular Networks", 28 pp, 2009, ISBN 978-160566338-8, 2009
Nouvel F.; Tanguy, P. & Maziero P. (2009b), What is about next high speed power line
communication systems for in -vehicle networks, Proceedings of ICICS, pp. 533-537,
978-1-4244-4656-8, Macau, 8-10 December 2009
OMEGA (2008), Deliverable D3.1 State of the art, application scenario and specific
requirements for PLC, OMEGA ICT-213311 Project, FP7, Available on web site
http://www.ict-omega.eu/fileadmin/documents/deliverables/Omega_D3.1.pdf
Pham, H.-M.; Pillement S. & Demigny, D. (2009) A Fault-Tolerant Layer For Dynamically
Reconfigurable Multi-Processor System-on-Chip, in Proc. Int. Conf. on
ReConFigurable Computing and FPGAs, pp. 284289, Cancun, Mexico, Dec. 2009.
Pham, H.-M.; Pillement, S. & Demigny, D. (2010) Evaluation of Fault-Mitigation Schemes for
Fault-Tolerant Dynamic MPSoC, in Proc. Int. Conf. on Field Programmable Logic and
Applications, 2010
Rubin, A. (2002). Implementing automotive protocols for communications over noisy
battery power lines, Proceedings of the IEEE Conference on Conv. Elect Electron. Eng.,
pp. 306, 1 December 2002
Ribeiro. et al. (2006) Power Line communications : a promising communication system
paragidm for last miles and meters applications, Telecommunications : Advances and
trends in transmissions, pp. 134-154 , ISBN 85-98876-18-6.
Silva, P.; Almeida, L.; Caprini, D.; Facchinetti, T.; Benzi, F. & Nolte, T. (2009). Experiments
on timing aspects of DC-powerline communications, Proceedings of IEEE

278
international Conference on Emerging Technologies & Factory Automation, pp. 1674-
1677, 978-1-4244-2727-7, Palma de Mallorca, Spain, September 22-25 2009
Spidcom (2008). Spidcom Inc., Retrieved from official web Official web site
http://www.spidcom.com.
Tanguy P.; Nouvel, F. & Maziro, P. (2009b), Power Line Communication standards for in-
vehicle networks, pp. 533-537, 978-1-4244-5347-4/09, Lilles, 20-22 October 2009
Tanguy, P.; Nouvel, F. (2010). In-Vehicle PLC Simulator Based on Channel Measurements,
in next Proceedings of International Conference on Intelligent Transport Systems
Telecommunications, Kyoto, 9-11 November 2010.
Umehara, D.; Morikura, M.; Hisada, T.; Ishiko S. & Satoshi, H. (2010). Statistical Impulse
Detection of In-Vehicle Power Line Noise Using Hidden Markov Model,
Proceedings of Power Line Communications and Its Applications, pp. 341-346, March
2010.
Valo (2006). Valo Inc., Electrical and Electronic Distribution Systems: Focus on Power Line
Communication. Retrieved from official web http://www.valeo.com
Xilinx, Inc (2004). PowerPC 405 Processor Block Reference Guide, 2004
Xilinx, Inc (2008). Early Access Partial Reconguration User Guide UG208, September 2008.
Xilinx, Inc (2009). MicroBlaze Processor Reference Guide UG081 (v10.3), 2009
Yamar, 2009, http://www.yamar.com, last access 20.09.2009
Yabuuchi, Y.; Umehara, D.; Morikura, M.; Hisada, T.; Ishiko ,S. & Horihata, S. (2010).
Measurement and Analysis of Impulsive Noise on In-Vehicle Power Lines,
Proceedings of Power Line Communications and Its Applications, pp. 325-330,
March 2010.
Zhang, N.; Zhu, X.; Liu, L.; Yu, C.; Zhang, Y.; Dong, Y.; Zhang, H.; Kuai, Z. & Hong, W.
2009. Measurement and characterization of wideband channel for in-vehicle
environment. In Proceedings of the 4th international Conference on Radio and Wireless
Symposium (San Diego, CA, USA, January 18 - 22, 2009). IEEE Press, Piscataway, NJ,
183-186.
15
Kinesthetic Cues that Lead the Way
Tomohiro Amemiya
NTT Communication Science Laboratories
Japan
1. Introduction
Wayfinding is of vital importance to pedestrians walking in unfamiliar areas. Generally,
pedestrians rely on directional information, street names, and landmarks [Bradley &
Dunlop, 2005]. Recently, many mobile devices, such as mobile smart phones, can provide us
with detailed digital maps, global positioning information, and navigational information.
These location-based data and services are usually presented on visual displays. However,
the visual displays in mobile devices are very small, which makes it hard to see and use the
data. With the increasing complexity of information, and the variety of contexts of its use, it
becomes important to consider how other non-visual sensory channels, such as audition and
touch, can be used to communicate necessary and timely information to users. Additionally,
there are a number of user groups, such as visually impaired people and the emergency
services, who also require non-visual access to geographical data.
Kinesthetic stimulation, such as that for pulling or pushing the hand, has the potential to be
more intuitive and expressive than cutaneous stimulation, such as rumbling vibration, in
conveying direction information because force feedback devices can indicate a one-
dimension direction directly. Although a substantial number of force feedback devices have
been developed in the last twenty years, most of them use either mechanical linkage to
establish a fulcrum relative to the ground (Massie & Salisbury, 1994), use a huge air
compressor (Suzuki et al., 2002; Gurocak et al., 2003), or require wearing a heavy device
(Hirose et al., 2001). Physical constraints mean that none of them can be used in portable
information devices. Some portable torque displays have been proposed, based on the
gyro effect (Yano et al., 2003) or angular momentum change (Tanaka et al., 2001) have been
proposed; however, they can produce neither a constant force nor a translational force
without also producing a reaction force; they can generate only a transient rotational force
since they use a change in angular momentum. Recently, there have been a number of
proposals for generating both constant and directional forces without an external fulcrum
by using two oblique motors whose velocity and phase are controlled (Nakamura & Fukui,
2007), by shifting the center-of-mass of a device dynamically to simulate kinesthetic inertia
(Swindells et al., 2003), and by producing an air pressure field with airborne ultrasound
(Iwamoto et al., 2008).
In contrast, our idea is to exploit the characteristics of human perception to devise a new
force perception method for portable information devices that can generate a translation
force sensation with a long duration (Amemiya et al. 2005; Amemiya and Maeda 2009). The
method uses an asymmetric oscillation, where brief intense pulses of acceleration alternate

280
with longer periods of low-amplitude recovery. Although the net acceleration is zero,
humans perceive a sustained force sensation in the direction of the pulses. This is attributed
to the nonlinear relationship between perceived acceleration and physical acceleration. We
built a handheld prototype that generates periodic motion through asymmetric acceleration,
in which asymmetric oscillation is generated by a swinging slider-crank mechanism. Our
previous findings indicated that the pulse frequency determines the effective generation of
the kinesthetic illusion of being pulled. In this chapter, we present a new hybrid
configuration comprising a swinging slider-crank mechanism and a cam mechanism as an
approach to fabricating a smaller force feedback system for portable information devices
and describe an experiment in which we conducted an empirically examined turn-by-turn
navigation with the device used by pedestrians with visual impairments. The results show
the device intuitively conveys turning instructions and has potential to be used by untrained
users.
2. Pseudo-attraction force
2.1 Haptic sensory illusion
The study of illusions can provide valuable insights into not only human perceptual
mechanisms but also the design of new human interfaces. To generate a sustained
translational force without grounding, we focused on the characteristics of human
perception, which until now have been neglected or inadequately implemented in haptic
devices. Although we human beings always interact with the world through human sensors
and effectors, the perceived world is not identical to the physical world. For instance,
different spectra can elicit the same color in human perception. When we watch television,
the images on TV (a combination of RGB colors) we see are different from what we see
through windows, i.e., a natural image is a composition of all wavelengths of light.
Furthermore, animation actually consists of a series of still pictures in a flip book. Different
stimuli can produce almost the same percept, which, though it may seem strange, is normal
for humans. Since some illusions are very stable independent of individual variation, we can
apply those illusions in practice, such as in designing human interfaces, if we can figure out
ways to convert them to subjectively equivalent percepts. Hayward has pointed out that
illusions are at the basis of virtually all technological displays (Hayward 2008), mentioning
this also includes haptic interfaces.
2.2 Principle
The kinaesthetic illusion of being pulled or pushed, discovered by the authors (Amemiya et
al. 2005), can be used to design haptic interfaces. Using different acceleration patterns for
two directions to create a perceived force imbalance, the method exploits the characteristics
of human perception to generate a force sensation and thereby produce the sensation of
directional pushing or pulling. Specifically, a quicker acceleration (stronger force) is
generated for a very brief time in the desired direction, while a slower acceleration (weaker
force) is generated over a longer period in the opposite direction. The internal human haptic
sensors do not detect the slower acceleration (weaker force) , so the original position of the
mass is washed out. The result is that the user is tricked into perceiving a unidirectional
force. This force sensation can be made continuous by repeating the motions. If the
acceleration patterns are well considered and designed, a kinesthetic illusion of being pulled
can be created because of this nonlinearity.

281
2.3 Requirements
There are still many aspects of the manifestation of the kinaesthetic illusion of the pseudo-
attraction force that are not well understood, but putative mechanisms have been
accumulating. We know that no directional force is felt if the mass is merely moved back
and forth, but that using different acceleration patterns for the two directions to create a
perceived force imbalance produces the perception of a pseudo-attraction force (Amemiya &
Maeda, 2009). The frequency of the oscillation plays an important role in eliciting the
perception of a pseudo-attraction force. Oscillations with high frequency might create a
continuous force sensation, but previous experimental results have shown that the
performance decreases steadily at frequencies over ten cycles per second (Amemiya et al.,
2008). In contrast, oscillations with low frequency tend to be perceived as a discrete knocked
sensation. If we wish to create a sustained rather than a discrete force sensation, such as the
sensation of being pulled continuously, the frequency should be in the five to ten cycles per
second range. In addition, changes in the gross weight and the weight of the reciprocating
mass affects the perceived force sensation. The threshold of the ratio of the gross weight and
the weight of the reciprocating mass is 16%, which is a rough standard for effective force
perception in the developed prototype (Amemiya & Maeda, 2009).
3. Hardware design
Our first prototype used a swinging-block slider-crank mechanism to create an asymmetric
oscillation [Fig. 1(a)]. In the mechanism, a circular motion of constant speed (crank OB) is
transformed into a curvilinear motion since a swinging linkage BC slides and turns around
point A. The end point on the curvilinear motion (point C) is connected with a rod (point D),
which slides along a linear slider with asymmetric acceleration back-and-forth. Because of
the length of the linkages, especially linkage CD (rod), the overall length of the mechanism
tends to be large at about 175 mm (Amemiya and Maeda 2008; 2009). Figure 1(b) shows the
new mechanism, which is the equivalent mechanism of the previous one but with the length
of linkage CD decreased to virtually zero. As in the previous prototype, a circular motion of
constant speed (crank OB) is transformed into a curvilinear motion by a swinging-block
slider-crank mechanism. The difference is that the end point on the curvilinear motion
(point C) slides along a grooved cam, whose shape is a circular arc with a radius of CD, with
point D as the center of the arc. This produces a reciprocating motion with asymmetric
oscillation.
The mechanisms in Fig. 1 have a single DOF (degree of freedom). We previously developed
a prototype of a two-dimensional force display by having one module based on the slider-
crank mechanism mounted on a turntable. The direction of the force display module was set
by driving the turntable with a belt drive system. Turntable rotation, however, took
considerable time, which meant that immediacy was lost. To overcome the problem of
turntable rotation, we adopted the summation of linearly independent force vectors. We
then fabricated a new 2-DOF prototype to generate a force sensation in at least eight
cardinal directions by the summation of linearly independent force vectors. To create the
force display, we stacked four layers, each containing a single module and two of which
were orthogonal. By combining the force vectors generated by each module, the force
display can create a force sensation on a two-dimensional plane more quickly than the
turntable approach. The force display has the potential to create a force sensation in any
arbitrary direction on a two-dimensional plane if the amplitude of the force vector can be
changed. The asymmetric oscillation (F(t)) is given by

282

Fig. 1. Mechanisms for generating asymmetric oscillation in first prototype (a) and new one
(b). The two mechanisms move identically

n
j
j j
j
d x t
F t m e
dt
2
2
1
( )
( )
=
=
(1)
where m
j
is the weight in module j, n is the number of modules, and d
2
x
j
/dt
2
is the
acceleration generated by module j. The acceleration d
2
x
j
/dt
2
is given by the second
derivative with respect to time of the motion of the weight x
j
. The equation for the motion of
the weight in module j is

j 1 j j 1 j 3 1 j j
x t l cos t d l cos t l l t
2
( ) ( ) { ( 1)sin } = + + (2)
where

j
j
l
l d l d t
2
2 2
1 1
2 cos
=
+
(3)
and x
j
(t) = OD, d = OA, l
1
= OB, l
2
= BC, l
3
= CD, and
j
t = AOB in Fig. 1. The
j
is the
constant angular velocity, and t is time. In the prototype, d = 28 mm, l
1
= 15 mm, l
2
= 60 mm,
l
3
= 70 mm, and n = 4, and the unit vectors are <e
j
, e
j+1
> =0, ||e
j
|| =1. All m
j
and
j
values
are identical, with m
j
= 40 g,
j
/2= 5 Hz.
In the developed 2-DOF prototype, the output shaft of each motor (DC 6.0 V, 2232R006S;
Faulhaber) is mounted in a roller made of nylon. The roller drives the crank wheel by
friction. The external diameter of the motor roller is 4 mm, and that of crank wheel is 44 mm.
The reduction ratio is basically 1:11, but it changes slightly as a result of changes in factors
such as temperature or pressure. The prototype weighs approximately 430 g. The diameter
of the base is the same as that of a compact disc (i.e., 120 mm). The prototype is 36-mm thick.
Each weight on the slider is equipped with a photo-interrupter (PM-R24; SUNX Ltd.) to
detect its position. The speed of each crank is stabilized at five counts per second by closed

283
feedback loop control (P control) with a microprocessor so that the signal intervals of the
photo-interrupters are close to 200 ms. When combining two orthogonal force vectors, the
phases of the cranks are synchronized by another closed feedback loop (PID control) so that
the onset intervals of the photo-interrupters are close to zero.

Fig. 2. Structure of the new prototype for generating pseudo-attraction force on a two-
dimensional plane
4. Hardware evaluation
Figure 3 shows the measured acceleration profile generated by the device at five counts per
second. Acceleration values were calculated from the position data of each weight, which
were acquired with a laser sensor (Keyence Inc., LK-G150, 10 kHz sampling) with the
bottom of the device fixed to a base. The acceleration profile of the top layer differed slightly
from the others, due its distance from the fixed base. The effect of oscillation was augmented
by the principle of leverage, leading to some degree of measurement error. The acceleration
amplitude reached about 50% of the theoretical acceleration peak. We assume that the

284
friction drive transmitted less torque than the previous gear drive. This clarified that there is
a trade-off between the torque transmission efficiency and noise level when we select the
friction drive or the gear drive.
Figure 4 shows examples of the response profiles of phase synchronization. The onset
intervals between pairs consisting of two orthogonal modules were acquired from the
photo-interrupters when the bottom of the device was fixed to the base. An onset interval of
zero means that the two orthogonal modules are synchronized. The phases were
synchronized within five seconds, which showed that the force display created a force
sensation in the eight cardinal directions on a two-dimensional plane.

Fig. 3. Acceleration profile calculated from the position data measured with a laser sensor.
The data were processed by a seventh order LPF Butterworth filter with a cut-off frequency
of 100 Hz
To drive the crank, the first prototype used a pinion gear and crown gears, whose axes were
relatively displaced. The relative displacement of the gear axes caused gear noise, which
annoyed the users. In fact, many people, including people who are blind, who have held the
previous prototype, have complained about the noise. The sound pressure level of the noise
generated by the previous prototype was measured in an anechoic room at NTT
Communication Science Laboratories. A sound level meter (Rion, Inc., NL-31 Class 1) was
used to measure the noise with the A-weighted sound pressure level (SPL). The sound level
meter was fixed to a tripod at a height of 1.0 m from the ground and 30 cm from the
prototype. The measured noise SPL showed that with the gear drive in the prior prototype,
the noise level exceeded 60 dB(A), whereas environmental noise level was 15 dB(A). In
contrast, the new prototype uses friction drive, and its noise level at frequencies of 3, 5, 10
counts per second does not exceed 50 dB(A), which shows that the friction drive emits much
less noise.

285

Fig. 4. Examples of measured phase synchronization responses for combinations of each
module. The onset intervals obtained by the photo-interrupters were controlled to be close
to zero by changing the angular velocity of the motor in the force display
5. User studies
We performed two user studies to investigate the applicability of the force display to
pedestrian navigation. One was a psychophysical pilot study, which revealed that people
who held the force display clearly sensed a directed force. The other was a navigation
experiment related to predefined route guidance, in which people with visual impairments
used the force display.
5.1 Directional perception
Five people without visual impairments (three right-handed men and two right-handed
women, aged 22-34 years, 25.45.3), who were paid volunteers, participated in the
psychophysical pilot study. None of the participants had any previous experience with force
display prototypes. They were required to reply with one of eight cardinal directions after a
five-second oscillating stimulus (north was defined as 0 degrees and the forward direction,
and east as 90 degrees and right) without any prior training. Each participant experienced
eight conditions five trials, for a total of 40 trials (randomized). Visual and auditory effects

286
were suppressed by having the participants wear eye-masks and earmuffs. The results
showed that the force display provided a well-perceived directed force sensation (pseudo-
attraction force sensation) even though the measured acceleration peak was only half the
expected value, and all responses from participants were almost identical to the direction of
stimuli.
5.2 Pedestrian navigation
For the empirical experiment (Fig. 5), to verify the navigation system, we recruited people
with visual impairments who were untrained. We measured the time required to complete a
walking task, the level of ease of use and, the users expectations of the guidance on a
subjective scale, and collected user feedback. We built a human-size maze in the gymnasium
of the Kyoto Prefectural School for the Visually Impaired, Japan. Since the streets and
avenues in Kyoto are mainly laid out in a grid, the routes of the maze were designed in a
checkerboard shaped.

Fig. 5. Overview of kinesthetic navigation for pedestrians with visual impairments. After
location data received from a satellite system or embedded sensors and orientation data had
been collected and processed, a haptic cue of direction was presented to the pedestrian
5.2.1 Method
There were twenty-three visually impaired participants, 19 males and 4 females, who were
volunteers from the Kyoto Prefectural School for the Visually Impaired. Ages ranged from
17 to 62 years (30.014.7). Thirteen are totally blind, and the other ten are partially sighted.
The participants all reported that they had no irregularities with their hands in terms of
tactile perception at the time of the experiment, and they were all untrained. The research
protocol was approved by local ethics committees. All participants provided written
informed consent prior to testing. No participants underwent any prior training.
Participants who usually used a cane when walking held a cane with one hand and the
haptic direction indicator with the other. Others held the haptic direction indicator with the
dominant hand. They carried a small shoulder bag (less than 600 g), which contained a
notebook computer (OQO model 02; OQO Inc.), a custom-build circuit with a
microprocessor (PIC18F252; Microchip Tech. Inc.), and a 12-V battery (ENAX Corp.).
Bluetooth 2.1 allowed the portable computer to communicate with the remote computer
(ThinkPad X60s; Lenovo Corp.) within 100 meters. Since the global positioning system (GPS)
did not work inside the gym with satisfactory accuracy, we used nine infrared sensors
installed at the junctions of the maze as a local positioning system; they were connected to

287
the remote computer. To produce white noise, a clip-shaped music player (iPod shuffle 2nd
generation; Apple Inc.) and noise-cancelling headphones (Quiet Comfort 3; Bose Corp.)
were used. To measure the yaw angle of the force display, a motion sensor (MDP-A3U9S;
NEC TOKIN Corp.) was attached to it. The human-sized experimental labyrinth was formed
with a series of foam panels (1,800 mm 900 mm) occupying a space of 9 m 15 m. The
pads were soft enough not to cause injury in collisions. The route and journey duration were
automatically logged by the system with infrared sensors placed at the corners of the walls,
captured by digital video cameras from the second floor (about 4 m from the ground), and
manually noted by the experimenters.
A participant holding the haptic directional indicator was brought to one of three different
departure points in the labyrinth. An experimenter walked behind the participant to ensure
his/her safety. First, the force display presented a sensation of pushing forward. The
participant then started walking from the departure point. The participant was guided by
the force display to turn left or right at a certain turning point, and there were totally nine
possible turning points. At that point, the infrared sensor detected the arrival of the
participant. In response, the remote computer connected to the infrared sensors sent the turn
instruction to the notebook computer in the participants bag. The direction of the force
vector (go straight; turn left or right) was initially determined by the predefined route at
each turning point and automatically updated to one of the eight cardinal directions
according to the orientation of the participant. In a similar way, they were then guided to
the second, third, and fourth turning points and finally to the destination point.
Experimenters indicated that the destination was reached after the participants had arrived
at the destination point.
The force display was always on during the navigation. If the participant made a wrong
turn (and was about to begin walking the wrong route), the experimenter changed the
direction of the haptic stimuli manually to return the participant to the predefined route and
sent them via the remote computer. The same haptic stimuli for the eight cardinal directions
were used for the revision. Nevertheless, if this correction failed or the participants did not
notice the stimulus change, the experimenter walking behind intervened by touching their
backs, giving verbal information about the correct turn and taking note of any incorrect
actions. Note that the participants were made aware of a wrong turn in an interruption but
not in a revision. The participants had to go back one block to return to the correct path
when there was an interruption.
The turn-by-turn navigation task consisted of walking predefined routes in the human-size
maze under two auditory information conditions (audio masked and audio unmasked).
Under the audio-masked condition, all the audio information was masked by noise-
cancelling headphones with white noise. Under the audio-unmasked condition, the
participants were able to hear ambient sounds, but no artificial sounds were presented. We
expected that people with visual impairments would utilize ambient audio information to
identify obstacles or walls in front of them. Since the foam panels along the route were 900
mm high, the audio cues would be different from those provided when walking along an
ordinary hallway. However, the wooden wall forming part of the building was over five
meters high, and it provided a clear acoustic echo that assisted localization. Each participant
completed one trial per one condition. All predefined routes were determined so that the
departure and destination points were incongruent, and there were four turns. The routes
were different for the two conditions and were randomly selected to reduce the learning
effect, for example to prevent the distance to a turn being remembered or guessed. The

288
participants were instructed to walk as fast and as accurately as possible. All participants
were invited to complete our two-item questionnaire and to comment freely about what
they felt during the task after the experiment. The statements were presented in a different
randomized order for each participant. Each statement was rated by the participants on a
seven-point Likert scale, with -3 meaning `totally disagree and +3 `totally agree. The
questions were:

Q1. The guidance was easy to understand.
Q2. I expect it would be useful in disaster situations.

Our intention with Q1 was to gain some insight into the usability of the force display in the
experiment for users with visual impairments. The aim with Q2 was to gain some insight
into the feasibility of using it during disasters, such as in heavy smoke (visual cues
unavailable) or when there is a loud siren noise (audio cues unavailable). In Q2, the
experimenter explained that a disaster meant a typical situation that lacked some audiovisual
information, and which required quick and safe directional navigation. These questions
focused on obtaining a subjective rating of the understandability and reliability of the force
display and the navigation system.
5.2.2 Results
The experimental results show that our proposed system successfully enabled the
participants to follow predefined routes. Twenty-one of twenty-three participants (i.e., 91%)
successfully walked from the entry point to the endpoint with four turns under both
conditions. Figure 6 shows examples of walking trajectories. Note that trials revised by the
experimenter were also counted as successes, whereas trials interrupted by the experimenter
were counted as failures. The same two participants failed to complete the tasks under both
conditions. The experimenter intervened twice under the audio-masked condition, and
three times under the audio unmasked condition. No other participants required
intervention during the navigation task. Specifically, one of those two participants always
seemed to judge left stimuli as right stimuli, and the other interpreted right stimuli as left
stimuli. This tendency appeared to prevent them from returning to the predefined route
even if the stimuli were changed manually. Seven participants under the audio-masked
condition and six under the audio- unmasked condition failed to perceive the force
sensation indicating a turn. Two of them could not recover the route at all. The other
participants were able to recover the route when the force display was driven by the
experimenter. Under the audio-masked condition, five participants made five mistakes.
Under the audio- unmasked condition, four made five mistakes. The average time required
to recover from an incorrect turn was about five seconds, and the longest time was about ten
seconds.
Almost all the participants rated both statements highly. The medians of the questionnaire
results were +2 and +2. The quartile ranges were 1 and 1. No outliers were observed.
Moreover, we analyzed the questionnaire results very conservatively by considering only
the high scores (+2 or +3) for each statement as indicating the usefulness; in other words, the
seven-point questionnaire responses were converted into a binary response (`high score or
`not high score). The scientific motivation for this is that we wanted to perform an analysis
that only took account of people who felt strongly about the usefulness of this force-based
navigation. By chance alone, the probability of a high score is 2/7. The ratings for the two

289
statements have frequencies that are much higher than would be expected by chance (21
and 18 respectively out of 23, with corresponding p<.0001 and .0001, using the binomial
distribution).
Some participants commented after performing the task that they felt the device to be very
useful and made it easy to comprehend the direction. A negative comment was that it was
hard to keep the force display horizontal and to maintain the direction indicated by the
display for a long period of time. Another negative comment from one participant was that
his hand felt numb because of the vibration and weight of the force display.

Fig. 6. Examples of a visually impaired participants walking trajectories. Even if auditory
information was masked, they could follow predefined routes
6. Discussion and design implications
The results provide clear evidence of the usefulness of force-based navigation for
pedestrians with visual impairments. Some participants collided with the walls because the
developed navigation system had no function for the recovery to a straight path from a
meandering walk. Due to the lack of such a function, the participants' direction tended to
diverge from the center line after they had turned a corner. To avoid such collisions, haptic
signals for mid-course corrections are required. There is a trade-off between increasing the
number of haptic signals and intuitive comprehension (or mental load), which is a version of
the haptic icon problem (Enriquez and MacLean 2008). Nonetheless, this should be
considered in future implementations. Our proposed system would also benefit from
information displays for other sensory modalities. For instance, adding an audio navigation
aid would make it relatively easily to provide complex and semantically rich information,
such as categories of landmarks and street names. This semantic richness and complexity is
very difficult to achieve with just a force display. Ross and Blasch pointed out that the
combination of speech (auditory) and tapping (tactile) information would be useful as
orientation aids (Ross and Blasch 2000). Future work will include investigating the effect of
such additional meaningful audio information such as street names, and landmarks.

290
Our haptic navigation system can easily be expanded to support sighted pedestrians. In
particular, travellers with a different mother tongue visiting an area would reap significant
benefits because haptic instruction is nonverbal. The proposed system would also enable
users with reading disabilities to travel independently. Iconic symbols such as left or right
arrows would be sufficiently clear and accurate to indicate directions. However, haptic cues
not only indicate direction (at the sensory phase) but also make bodies move directly (at the
motor phase), which requires less response time. Therefore, haptic cues would be more
suitable for navigation applications. Future work will include the integration of satellite
navigation instruction, which would expand the available area of our technique.
Manually changing the force vector assisted participants who made a wrong turn to return
to the original route. Since the haptic stimuli were identical for the automatic and manual
operations, one of the main reasons manual intervention was required was the incomplete
algorithm for changing the direction of the force vector. The algorithm depended on the
accuracy and reliability of the local position sensors (infrared sensors placed at the maze
corners). The detection onset for the local position sensors and turning a corner varied as the
walking pace changed.
Our experimental setup has some limitations with regard to generalizing the findings for all
areas of pedestrian navigation since we only examined turn-by-turn navigation task in
orthogonal-crossed route. Since most streets and avenues, with the exception of certain cities
such as Kyoto, are not laid out in a grid, a navigation task in relation to complicated routes
must be the next step. In addition, we have to clarify whether the same angular resolution
(i.e., the eight cardinal directions) is sufficient for arbitrary routes, in particular when
avoiding obstacles. Pielot et al. have reported deficiencies in the human perception of body
orientation when small angles are involved (Pielot et al. 2008). Our previous study also
showed that there is a systematic error with respect to the human perception of force
orientation among people with visual impairments when they are not moving, which is less
than 15 degrees when the force vector is provided in the eight cardinal directions (Amemiya
2009). A dynamic exploration of the force vector would be useful when walking complicated
routes, which is similar to the guidance a human or a guide dog might provide and depends
on the active or passive perception of haptic feedback while walking. Therefore, the
systematic error could be minimized with a closed feedback loop, which could lead to
precise routing. However, debate continues about the best way to understand the user's
orientation for dynamic exploration, and about the most suitable part of the body to which
to attach the orientation sensors. Moreover, we must consider whether the force display
should be held in the hand, carried in a bag, or worn.
In the experiment, the force display always generated force during navigation, because in a
pilot study we received feedback that the absence of stimuli made the participants anxious.
However, it is well known that continued vibration tends to cause perceptual fatigue (i.e.,
adaptation) if it is presented for too long, as with all other sensations (Coren et al. 2003).
Nevertheless, there were no differences in the number of collisions with the walls between
the L-shaped walking paths of the same trial: one is from the entry point to the second turn,
the other is from the second to the last turn to the exit point. The average numbers of
collisions were 0.10 and 0.12 (audio unmasked condition), and 0.07 and 0.09 (audio masked),
respectively. The average lengths of the former and latter paths were 11.2 meters and 13.2
meters (audio unmasked), and 9.4 meters and 14.2 meters (audio masked), respectively. In
addition, all but one of the participants reported that they did not feel any subjective
perceptual difference between the start and the end as mentioned above. We conjecture that

291
the effect of vibrotactile adaptation may not appear during short-term navigation, since the
asymmetrically oscillating stimuli involve not only cutaneous but also proprioceptive
sensations. The muscle spindles or the Golgi tendon organs, which are receptors generating
the proprioceptive sensation, are slowly adapting units, while the Pacinian corpuscles,
receptors that detect skin vibration, are rapidly adapting units.
Many of the participants made similar positive comments. The questionnaire rating clearly
revealed high confidence levels and high expectations. Their medians and quartiles
confirmed that the force sensation was well perceived. Responses to Q1 indicated that the
participants were aware of the direction information and found the information provided by
our system to be very intuitive. Those to Q2 confirmed that many of the participants
realized the importance of force feedback in emergency navigation. The participants also
commented on the quietness of the force display. They commented that the noise level of the
force display would be acceptable in daily life, such as in public spaces. People with visual
impairments sometimes use the information provided by acoustic echoes to gain awareness
of the environment. The proposed device was so quiet that this information could still be
used.
It is crucial to miniaturize and lighten the force display so that it can be carried more easily.
However, the amplitude of the kinesthetic stimuli, which should be large enough to be
perceived, is proportional to the mass of the weight and the amplitude of acceleration. The
trade-off between the mass of the weight and strength of perception limits the amount by
which the weight can be reduced. However, the size of the force display could be reduced
by using other mechanisms to generate similar asymmetric oscillation, such as linear
actuators. We speculate that a miniaturized version of the force display could be embedded
in canes for people with visual impairment. It is true that some people feel that no device
should be attached to the cane, but only a relatively small amount of retraining would be
needed.
7. Related work
Recent research has addressed a range of issues concerning pedestrian navigation aid.
Pedestrian navigation systems on mobile devices, such as mobile phones with a satellite
positioning function, that employ different sensory channels (i.e., visual and/or audio
channels) are being increasingly developed. Since people with visual impairments cannot
use vision-based navigation aids, many handheld auditory-feedback devices for people with
visual impairments have been developed, such as Talking Signs (Crandall et al. 1999) or
similar acoustic information output devices (Loomis et al. 2005). Although such audio
interfaces help users move in the right direction by providing sound cues, they can be
problematic when they conflict with other sounds or speech around the users, making it
difficult for them to distinguish and interpret the sounds generated by the system (Wilson et
al. 2007). In addition, pedestrians with visual impairments often rely on information
contained within the ambient sounds for navigation purposes. Wearing headphones
prevents them from hearing these ambient sounds thereby making navigation less safe.
Furthermore, auditory information cannot be used in noisy situations, such as on busy city
streets.
Tactile interaction may help overcome such navigation issues for people with visual
impairments. It has been reported that tactile interaction can effectively assist pedestrians
with visual impairments when crossing the street (Ross and Blasch 2000). As tactile-based

292
navigational aids, vibrotactile stimulation systems that use several vibrators in the shape of
a cap (Cassinelli et al. 2006), rings (Amemiya et al. 2004), a vest (Erp et al. 2005), belt (Tan et
al. 2003; Heuten et al. 2008), and glove (Zelek et al. 2003) have been proposed.
Unfortunately, these tactile approaches require that users learn how to convert stimuli to
information. This is not intuitive and requires training since the tactile stimuli employed are
basically non-directional.
8. Acknowledgements
We thank Dr. Ichiro Kawabuchi for his technical assistance. We also thank Hisashi
Sugiyama, the staff of Kyoto City Fire Department, and the staff of the Kyoto Prefectural
School for the Visually Impaired for their kind cooperation. This study was supported by
Nippon Telegraph and Telephone Corporation and was also partially supported by the
sponsorship of the Fire Defence Agency, Japan.
9. References
Amemiya, T.; Ando, H. & T. Maeda T. (2005). Virtual force display: Direction guidance
using asymmetric acceleration via periodic translational motion, Proceedings of
World Haptics Conference, IEEE Computer Society, pp. 619-622.
Amemiya, T.; Ando, H. & T. Maeda T. (2008). Lead-Me interface for pulling sensation in
hand-held devices, ACM Transactions on Applied Perception, Vol. 5, No. 3, pp. 1-17.
Amemiya, T. & Maeda, T. (2008). Asymmetric oscillation distorts the perceived heaviness of
handheld objects, IEEE Transactions on Haptics, Vol. 1, No. 1, pp. 9-18.
Amemiya, T. & Maeda, T. (2009). Directional force sensation by asymmetric oscillation from
a double-layer slider-crank mechanism, Journal Computing Information Science in
Engineering, Vol. 9, No. 1, 011001.
Amemiya, T.; Maeda, T. & Ando, H. (2009). Location-free Haptic Interaction for Large-Area
Social Applications, Personal and Ubiquitous Computing, Vol. 13, No. 5, pp. 379-386,
Springer.
Amemiya, T. & Sugiyama, H. (2009). Haptic Handheld Wayfinder with Pseudo-Attraction
Force for Pedestrians with Visual Impairments, Proceedings of 11th ACM Conference
on Computers and Accessibility (ASSETS 2009), Pittsburgh, PA, pp. 107-114.
Amemiya, T. & Sugiyama, H. (2010). Orienting Kinesthetically: A Haptic Handheld
Wayfinder for People with Visual Impairments, ACM Transactions on Accessible
Computing, Vol. 3, No. 2, pp. 1-23.
Amemiya, T.; Yamashita, J.; Hirota, K. & Hirose, M. (2004). Virtual Leading Blocks for the
Deaf-Blind: A Real-Time Way-Finder by Verbal-Nonverbal Hybrid Interface and
High-Density RFID Tag Space, In Proc. of IEEE Virtual Reality Conference 2004 (VR
2004), pp. 165-172.
Bradley, A. & Dunlop, D. (2005). An experimental investigation into wayfinding directions
for visually impaired people, Personal Ubiquitous Computing, Vol. 9, No. 6, pp. 395-
403.
Cassinelli, A.; Reynolds, C. & Ishikawa, M. (2006). Augmenting spatial awareness with
haptic radar. In Proc. International Conference on Wearable Computing. IEEE Computer
Society, pp. 61-64.

293
Coren, S.; Ward, L. M. & Enns, J. T. (2003). Sensation and Perception. John Wiley and Sons,
Inc.
Crandall, W.; Brabyn, J.; Bentzen, B. & Myers, L. (1999). Remote infrared signage evaluation
for transit stations and intersections. Journal of Rehabilitation Research and
Development Vol. 36, pp. 341-355.
Enriquez, M. & MacLean, K. (2008). The role of choice in longitudinal recall of meaningful
tactile signals. In Proc. of 16th IEEE Symposium on Haptic interfaces for virtual
environment and teleoperator systems. pp. 49-56.
Erp, J. B. F. V.; Veen, H. A. H. C. V.; Jansen, C. & Dobbins, T. (2005). Waypoint navigation
with a vibrotactile waist belt. ACM Transactions on Applied Perception, Vol. 2, No. 2,
pp. 106-117.
Foulke, E. (1996). The roles of perception and cognition in controlling the mobility task.
International Symposium on Orientation and Mobility.
Golledge, R. G. (1992). Place recognition and wayfinding: making sense of space. Geoforum
Vol. 23, No. 2, pp. 199-214.
Gurocak, H.; Jayaram, S.; Parrish, B. & Jayaram U. (2003). Weight sensation in virtual
environments using a haptic device with air jets, Journal of Computing and
Information Science in Engineering, Vol. 3, No. 2. ASME, pp. 130-135.
Hayward, V. (2008). A Brief Taxonomy of Tactile Illusions and Demonstrations That Can Be
Done In a Hardware Store, Brain Research Bulletin, Vol. 75, pp. 742-752.
Heuten, W.; Henze, N.; Boll, S. & Pielot, M. (2008). Tactile wayfinder: a non-visual support
system for wayfinding. In NordiCHI. ACM International Conference Proceeding Series,
Vol. 358. ACM Press, pp. 172-181.
Hirose, M.; Hirota, K.; Ogi, T.; Yano, H.; Kakehi, N.; Saito, M. & Nakashige, M. (2001).
HapticGEAR: The Development of a Wearable Force Display System for Immersive
Projection Displays, Proceedings of Virtual Reality 2001 Conference, pp. 123130.
Hoshi ,T.; Takahashi, M.; Iwamoto, T. & Shinoda, H. (2010). Noncontact Tactile Display
Based on Radiation Pressure of Airborne Ultrasound, IEEE Transactions on Haptics,
Vol. 3, No .3, pp. 155-165.
Loomis, J.; Marston, J.; Golledge, R. & Klatzky, R. (2005). Personal guidance system for
people with visual impairment: A comparison of spatial displays for route
guidance. Journal of Visual Impairment and Blindness Vol. 8, No. 5, pp. 61-64.
Massie, T. & Salisbury, J. K. (1994). The phantom haptic interface: A device for probing
virtual objects, Proceedings of the ASME Winter Annual Meeting, Symposium on Haptic
Interfaces for Virtual Environment and Teleoperator Systems, Vol. 55-1, pp. 295-300.
Nakamura, N. & Fukui, Y. (2007). Development of Fingertip Type Non-grounding Force
Feedback Display, Proceedings of World Haptics Conference 2007, pp. 582-583.
Pielot, M., Henze, N., Heuten, W., & Boll, S. (2008). Evaluation of continuous direction
encoding with tactile belts. In Proc. the 3rd international workshop on Haptic and Audio
Interaction Design, Springer, LNCS, pp. 1-10.
Richard, C. & Cutkosky, M. (1997). Contact Force Perception with an Ungrounded Haptic
Interface, Proceedings of the ASME Dynamic Systems and Control Division, pp. 181
187.
Ross, D. & Blasch, B. (2000). Wearable interfaces for orientation and wayfinding. In Proc.
ACM Conference on Assistive Technologies. ACM Press, pp. 193-200.

294
Suzuki, Y.; Kobayashi, M. & Ishibashi, S. (2002). Design of force feedback utilizing air
pressure toward untethered human interface, Proceedings of CHI 02 Extended
Abstracts on Human Factors in Computing Systems. ACM Press, 2002, pp. 808-809.
Swindells, C.; Unden, A. & Sang, T. (2003). TorqueBAR: an ngrounded haptic feedback
device. Proceedings of the 5th international conference on multimodal interfaces. ACM
Press, pp. 52-59.
Tan, H. Z.; Gray, R., Young, J. J. & Traylor, R. (2003). A haptic back display for attentional
and directional cueing. Haptics-e: The Electronic Journal of Haptics Research Vol. 3,
No. 1.
Tanaka, Y.; Masataka, S.; Yuka, K.; Fukui, Y.; Yamashita, J. & Nakamura, N. (2001). Mobile
torque display and haptic characteristics of human palm. Proceedings of 11th
international conference on augmented tele-existence, pp. 115-120.
Wilson, J.; Walker, B.; Lindsay, J.; Cambias, C. & Dellaert, F. (2007). Swan: System for
wearable audio navigation. In Proc. International Conference on Wearable Computing.
IEEE Computer Society, pp. 91-98.
Yano, H.; Yoshie, M. & Iwata, H. (2003). Development of a nongrounded haptic interface
using the gyro effect, Proceedings of 11th international symposium on Haptic Interfaces
for Virtual Environment and Teleoperator Systems. IEEE Computer Society, pp. 32-39.
Zelek, J. S.; Bromley, S.; Asmar, D. & Thompson, D. (2003). A haptic glove as a tactile-vision
sensory substitution for wayfinding. Journal of Visual Impairment and Blindness Vol.
97, No. 10, pp. 621-632.
Part 2
Transmission Technologies and Propagation
16
Technological Trends of Antennas in Cars
John R. Ojha, Ren Marklein and Ian Widjaja
Germany
1. Introduction
Antennas have become a commonplace in automotive applications. These are broadly
classified as wire and patch antennas which are used in cars for inter-vehicle
communication. Besides its use in the automotive sector, these antennas are also used as
arrays in the aviation sector e.g. fuselage integrated microstrip phased antenna arrays. These
wire and patch antennas can either be modeled analytically e.g. using the Greens function,
derived from Eigen functions or numerically using various approaches e.g. MoM, FDTD,
FEM etc. Besides the common usage of wire and patch antennas of various shapes,
integrated antennas are also widely used. Antennas starting from the traditional monopole
antenna followed by patch antennas on car roof tops and mesh antennas on car windscreens
will be discussed in this chapter.
2. Figures of merit
This section lists and explains some salient figures of merit of antennas. The input
impedance and the radiated fields (near and far) are termed as the primary figures of merit
since they form the basis on which other secondary figures of merit such as VSWR,
bandwidth, and directivity etc. are determined. Section 2.1 elaborates on the primary figures
of merit viz. input impedance. Section 2.2 explains some secondary figures of merit which
are obtained from the input impedance. The theory of how the effective radiating power is
calculated from the far-field gain patterns is explained in section 2.3.
2.1 Input impedance
The input impedance Z
in
is defined as the impedance presented by an antenna at its input
terminals a b, as shown in Fig. 1. In other words, the input impedance of an antenna is the
ratio of the voltage to the current or the ratio of the electric to the magnetic field measured at
the input terminals (feeding point). The input impedance of an antenna is expressed in
terms of its real and imaginary parts as

in in in
Z R jX , = + (1)
where Z
in
is the antenna impedance at the input terminals a b,
R
in
is the antenna resistance at the input terminals a b,
and X
in
is the antenna reactance at the input terminals a b.

298

Fig. 1. Block diagram of a transmitting antenna
The imaginary part X
in
of the input impedance represents the power stored in the near field
region of the antenna. The resistive part R
in
of the input impedance consists of two
components, the radiation resistance R
r
and the loss resistance R
l
. The power associated with
the radiation resistance R
r
is the power actually radiated by the antenna and the loss
resistance R
l
represents the dielectric or conducting losses resulting in power dissipation.
The input impedance is of great importance in wire and patch antennas and is therefore
discussed here. The input impedance is used as a foreboding of unwanted radiation for
EMC related aspects especially in the automotive sector. However, in the case of antennas,
the input impedance with the source impedance is used as an intermediate parameter for
determining the S11 parameter, return loss, Voltage Standing Wave Ratio (VSWR), and
bandwidth. This is explained in more detail in section 2.2, where the matching
characteristics of a patch antenna and its bandwidth are explained.
2.2 Reflection coefficient / S11 / VSWR / return loss
Antennas are commonly used in various type of smart antenna systems. In order for any
given antenna to operate efficiently, the maximum transfer of power must take place
between the feeding system and the antenna. Maximum power transfer can take place only
when the input impedance of the antenna (Z
in
) is matched to that of the feeding source
impedance (Z
S
). According to the maximum power transfer theorem, maximum power can
be transferred only if the impedance of the source is a complex conjugate of the impedance
of the antenna under consideration and vice-versa. If this condition for matching is not
satisfied, then some of the power may be reflected back. This is expressed as

1 | |
,
1 | |
VSWR
+
=

(2)
with
,
r in S
i in S
V Z Z
V Z Z
= =
+
(3)

299
where is called the reflection coefficient, V
r
is the amplitude of the reflected wave, and V
i

is the amplitude of the incident wave. The VSWR is basically a measure of the impedance
mismatch between the feeding system and the antenna. The higher the VSWR, the greater is
the mismatch. The minimum possible value of VSWR is unity and this corresponds to a
perfect match. The return losses (RL), obtained from equations (2) and (3), indicate the
amount of power that is transferred to the load or the amount of power reflected back. In the
case of a microstrip-line-fed antenna, where the source and the transmission line
characteristic impedance or the transmission line and the antenna edge impedance do not
match, waves are reflected. The superposition of the incident and reflected waves leads to
the formation of standing waves. Hence the RL is a parameter similar to the VSWR to
indicate how well the matching is between the feeding system, the transmission lines, and
the antenna. The RL is
20log| | RL = (dB). (4)
To obtain perfect matching between the feeding system and the antenna, = 0 is required
and therefore, from equation (4), RL = infinity. In such a case no power is reflected back.
Similarly at = 1, RL = 0 dB, implies that all incident power is reflected. For practical
applications, a VSWR of 2 is acceptable and this corresponds to a return loss of 9.54 dB.
Usually return losses ranging from 10 dB to 12 dB are acceptable.
The bandwidth could be defined in terms of its Voltage Standing Wave Ratio (VSWR) or
input impedance variation with frequency. The VSWR or impedance bandwidth of an
antenna is defined as the frequency range over which it is matched with that of the feed line
within specified limits. The BW of an antenna is inversely proportional to its quality factor Q
and is expressed as

1 VSWR
BW
Q VSWR
= . (5)
The bandwidth is usually specified as the frequency range over which the VSWR is less than
2 (which corresponds to a return loss of 9.5 dB or 11 % reflected power). Sometimes for
stringent applications, the VSWR requirement is specified to be less than 1.5 (which
corresponds to a return loss of 14 dB or 4 % reflected power). In the case of a patch antenna,
the input impedance with the source impedance is used as an intermediate parameter for
determining the S11 parameter (a measure of the reflection coefficient ), return loss,
Voltage Standing Wave Ratio (VSWR), and bandwidth. The return loss is expressed in dB in
terms of S11 as the negation of the return loss. The bandwidth can also be defined in terms
of the antennas radiation parameters such as gain, half power beam width, and side-lobe
levels within specified limits.
2.3 Effective radiating power
For every other antenna, the directivity is defined as the ratio of the radiation intensity in a
given direction from the antenna to the radiation intensity U
0
averaged over all directions. If
the direction is not specified, the direction of maximum radiation intensity is implied.
Hence mathematically the directivity is

max max
0
4
o rad
U U
D
U P
= = , (6)

300
where
max
,
rad
U P are the maximum radiation intensity and total radiated power, expressed
in Watts / solid angle and Watts respectively.
The antenna gain is directly associated with the directivity of an antenna and is therefore
associated with only the main lobe. The term K is the radiation efficiency expressed in terms
of the conduction efficiency K
c
and dielectric efficiency K
d
as

c d
K K K = , (7)
Gain and directivity extraction are based on the source power. Let us assume that P
t
is the
source power and P
v
are some losses in the structure (e.g. dielectric losses), then a power P
t

P
r
=P
t
- P
v
will be radiated. The directivity (as compared to an isotropic point source) is then
defined as
D = 4R2 * (S
s
/P
r
), (8)
where S
s
= (1 /2)
F0
(|E 2 E |2 / Z ) +

Z
F0
denotes the wave impedance of the surrounding medium.
From the equation the gain is extracted from the directivity as
G K D = , (9)
where G is the gain and D is the directivity. (For an antenna with 100% efficiency, K = 1.)
The far field gain is determined from the electric far-field components E
and E
and the
source power. The electric field components E
and E
are calculated from the surface

electric current densities. The effective radiating power is extracted from the gain by
removing the effect of the losses in the form of metallic or /and dielectric losses.
Effective Radiating Power Gain Power loss = (10)
3. Numerical approaches for determining figures of merit
The numerical analysis e.g. MoM can be carried out either in the spectral or in the time
domain. A patch antenna comprising metallic and dielectric parts with a feeding pin or
microstrip line is solved using the traditional MoM by decomposing the antenna as
discretized surface parts
wire parts
attachment node of the wire to the surface element.
Metallic surfaces contain different basis functions as shown in Fig. 2. The MoM uses surface
currents to model a patch antenna. In the case of ideal conductors, the boundary condition
of E
tan
= 0 is applied.
The most commonly used basis functions for line currents through wires are stair case
functions, triangular basis functions, or sine functions. The MoM code uses triangular basis
functions. In contrast to wires, two-dimensional basis functions are employed for surfaces.
The current density vectors have two-directional components along the surface. Figure 2
shows the overlapping of so-called hat functions on triangular patches. An integral equation
is formulated for the unknown currents on the microstrip patches, the feeding wire /
feeding transmission line, and their images with respect to the ground plane. The integral
equations are transformed into algebraic equations that can be easily solved using a

301
computer. This method takes into account the fringing fields outside the physical boundary
of the two-dimensional patch, thus providing a more exact solution. The coupling
impedances Z
ik
are computed in accordance with the electric field integral equation.

Fig. 2. Hat basis functions on discretised triangular elements on patches
The MoM uses either surface-current layers or volume polarization to model the dielectric
slab. In the case of dielectric materials we have to consider 2 boundary conditions

1 2
n E n E =
G G
G G
, (11)

1 2
n H n H =
G G
G G
. (12)
The traditional full-model applied in the MoM code uses a surface-current approach which
is categorised as
double electric current layer approach or
single magnetic and electric current layer approach.
4. Various type of antennas
Various type of antennas are described here. Antennas e.g. the conventional monopole,
which is of historical importance is still widely used due to its simplicity in construction.
The following sections deal with technological trends with respect to the monopole family
of antennas as well as patch antennas.
4.1 Wire antennas (monopole antenna)
Monopole antennas are commonly used in automotive applications where range is
important. A brief description of how a monopole antenna is characterised will be
illustrated e.g. a monopole antenna is suitably placed on a car and then meshed effectively
for numerical simulation. These antennas are also very easy to design and tune simply by
slightly varying the length. It is assumed the antenna is a quarter wavelength long, which is
typical of monopole antennas in the UHF band. The radiation characteristics are linearly
polarized, either horizontally or vertically, depending on antenna orientation. Radiation
resistance of a quarter wave monopole is approximately 37, and does not vary much with
presence or absence of ground plane. The radiation resistance of monopole antennas is

302
length dependent. Resonance of a quarter-wavelength monopole occurs when its length is
slightly less than a quarter-wavelength. The appropriate length for a quarter-wave
monopole at 433.92MHz would be 2808 433.92 = 6.47 inches. Sophisticated antenna
measurements are generally not necessary unless a highly optimized design is desired. This
makes the monopole very popular and easy to apply. The bandwidth of the antenna can
either be broadened by providing an LC circuit or by providing a parasitic element near the
wire part connected to the source. Fig. 3 shows a simple sketch of the traditional monopole
antenna.

Fig. 3. The traditional monopole antenna
Some salient features are
- To increase the resonant frequency, decrease the monopole height.
- To increase the bandwidth, increase the wire thickness. Variation in wire thickness will
have a small effect on the resonant frequency of the antenna. The resonant frequency of
the antenna should be corrected for by adjusting the length.
- To decrease the impedance variation versus frequency, increase the element size.
4.2 Monopole antenna with sleeve
Monopole antennas have problems of low bandwidths. The aim of this section is to show a
scheme to broaden the bandwidth by providing a sleeve as shown in Fig. 4. The cylindrical
sleeve acts as a parasitic element. The advantage of the monopole antenna with the
provision of a sleeve is clear from Fig. 5. If the diameter of the wire is not large a wire can
still be used instead of cylinder.

Fig. 4. Monopole antenna with sleeve

303

Fig. 5. Comparison of sleeve monopole antenna and the traditional monopole antenna
The design approach is to adjust the exterior dimensions of the antenna to achieve pattern
stability and then to use the region within the sleeve for impedance matching [Poggio et al.].
- To increase the operating frequency, decrease the monopole height.
- To increase the bandwidth, increase the wire thickness. (Note that changes in wire
thickness will have a small effect on the operating frequency of the antenna. This
should be corrected for by adjusting the length according to the previous guideline).
- To decrease the impedance variation versus frequency, increase the element diameter.

Fig. 6. Monopole antenna (inclined) mounted on a car

304
Fig 5 shows the characteristics of a traditional monopole antenna on an infinite ground
plane. The far-field gain, antenna efficiency, and matching characteristics change with
change in location of a monopole antenna in positions A, B, and C shown in Fig. 7. Fig 8
shows variation in the far- field gain patterns for change in the antenna location. There is
also a variation in the far-field gain, shown in Fig. 9 when the monopole antenna is upright
and inclined. In todays world the antenna is mounted inclined on a car as shown Fig. 6 and
Fig. 7 (scheme D). The determination of antenna efficiency and matching characteristics
(VSWR) is left as an exercise to the reader.

Fig. 7. Monopole antennas mounted at various locations

Fig. 8. Far-field gain patterns of antennas at various locations

305

Fig. 9. Far-field gain patterns of antennas at orientations (vertical/tilted)
4.3 Patch antennas
The most common patch antennas in todays world are primitives such as squares, triangles,
etc, metallised on a substrate backed by a ground plane. The next section gives a brief
overview of a rectangular, a circular, and an elliptical patch antenna.
4.3.1 Rectangular patch antenna
From the cavity model point of analysis, the wave numbers k
x
,, k
y
,, k
z
in the corresponding

x , y , z directions are

0 1 2
0 1 2
0 1 2
x
y
z
m
k , m , , ,.....
L
n
k , n , , ,.....
W
p
k , p , , ,....
H

= =

= =

= =

, (13)
where m, n, p represent the number of half-cycle field variations along the x, y, and z
directions respectively. The primed cylindrical co-ordinates x , y , z are used to represent
the field within the cavity. The resonant frequency for such a patch or cavity is

2 2 2
1
2
r mnp
p m n
( f )
L W H

= + +

, (14)

306
where W, L, H represent the width, length and height of the patch antenna. Since the
substrate height H is very small ( )
0
H << the electric field along the z direction is assumed
constant and hence 0 p = and 0
z
k = and consequently the last term in equation (14)
disappears.
Some design guidelines for a rectangular patch antenna shown in Fig. 10 are
- To increase (decrease) the resonant frequency, decrease (increase) the patch length.
- To increase bandwidth, increase the substrate height and/or decrease the substrate
permittivity (this will also affect resonant frequency and the impedance).
- The bandwidth may be increased (decreased) by increasing (decreasing) the patch
width.
- To increase (decrease) the input impedance decrease (increase) the pin inset.
Note: Antennas on very thin substrates have high copper-losses, while thicker and higher
permittivity substrates may lead to performance degradation due to surface waves and
feed-pin impedance. The maximum impedance that can be realised is governed by the
impedance seen at the edge of the patch. The minimum realisable impedance is zero, at the
centre of the patch. However, the practical minimum is governed by the rapid impedance
variation as the centre is approached. A typical patch antenna similar in nature is mounted
on a car as shown in Fig. 11 with no substrate.

Fig. 10. Square/ rectangular patch antenna

Fig. 11. Patch antenna (no substrate) mounted on a car

307
4.3.2 Circular patch antenna
From the cavity model point of analysis, the wave numbers
z
k , k , k

in the corresponding
, , z directions are

0 1 2
0 1 2 3
0 1 2
mn
z
k , m , , ,.....
A
k , n , , .....
p
k , p , , ,....
H

= =

= =

= =

, (15)
where m, n, p represent the number of half-cycle field variations along the , , and z
directions respectively. The primed cylindrical coordinates , , z are used to represent
the field within the cavity. Taking into account the condition 0
z
k = , as in the case of the
rectangular structure, the resonant frequency for such a circular patch is
( )
1
2
mn
r
mnp
f
A

, (16)
where A represents the radius of the disk and
mn
represents the zeros of the derivatives
of the Bessel function ( )
m
J x whose values are given in table 1.

Mode (m,n)
mn

(1,1) or 1
st
mode 1.8412
(2,1) or 2
nd
mode 3.0542
(0,1) or 3
rd
mode 3.8318
(3,0) or 4
th
mode 4.2012
Table 1. Zeros of the derivatives of ( ) 0
m
J x = at mode (m, n) of order m at the
th
n zero
cross-over point
The use of electrically thick substrates in designs will have degraded matching due to
increased feed pin inductance.
- Increasing the patchs diameter will decrease the resonant frequency and vice versa.
- Increasing the substrate height will increase the bandwidth, but will decrease the
resonant frequency slightly.
- Increasing the substrate height will also result in a more inductive reactance due to the
feed pin.
- To increase/decrease the input impedance, increase/decrease the feed offset.
- The circular patch antenna may be fine tuned for both impedance and centre frequency
by the use of trimming stubs, as for the rectangular patch.
Note: Antennas on very thin substrates have high copper losses, while thicker and higher
feed-pin impedance. The maximum impedance that can be realised is governed by the
variation as the centre is approached. The patch antenna is on the x-y plane.

308

Fig. 12. Circular patch antenna
4.3.3 Elliptical patch antenna
Elliptical antennas are used for single-fed circular polarized antennas, especially in
automotive applications. These antennas are characterized analytically making use of the
Mathieu function in the case of elliptical antennas. A circular patch antenna could also be
used however 2 feeds are necessary with the physical angle and electrical angle displaced by
90 degrees, namely
feed 1: y = 0 and V = 1 at phase angle = 0 degrees.
feed 2: x = 0 and V = 1 at phase angle = 90 degrees.
A typical substrate would have an
r
of 2.48 and a substrate height approximately 1.5% of a
free-space wavelength.
- To increase the operating frequency, reduce the patch dimensions while keeping the
ratio of the major to the minor ellipse axes constant.
- To improve the axial ratio at the centre frequency, increase or decrease the ratio of the
major to the minor ellipse axes.
- To increase the bandwidth, try increasing the substrate height and/or decreasing
r
.
- To increase/decrease the input impedance, the feed offset should be increased/decreased.

Fig. 13. Elliptical patch antenna
Note: Antennas on very thin substrates have high copper losses, while thicker and higher
feed-pin inductance. The maximum impedance that can be realised is governed by the
variation as the centre is approached. Furthermore the best performance is achieved when

309
the ratio of the minor axis to the major axis is almost unity. As in section 4.2.1 and 4.2.2 the
antenna is on the x-y plane.
Properties of patch antennas of rectangular and circular geometries on planar surfaces were
listed briefly. Such patch antennas also exist on cylindrical and spherical surfaces. Other
patch antenna shapes (besides rectangular, circular and elliptical) widely used are triangular
and annular in nature. One of the most widely used triangular shaped patch antennas is the
bow-tie antenna. Annular antennas are used in applications where a broader bandwidth is
required. In some cases, the inner radius of the annulus is short circuited.
5. Design guidelines for patch antenna arrays
For a given center frequency and substrate relative permittivity, the substrate height should
not exceed 5% of the wavelength in the medium. The following guidelines are a must for
designing a patch antenna and its arrays fed by microstrip lines.

Fig. 14. Automotive patch antenna selected from Antenna Magus
- The length of the patches may be changed to shift the resonances of the centre
fundamental frequency of the individual patch elements. The resonant input resistance
of a single patch can be decreased by increasing the width of the patch. This is
acceptable as long as the ratio of the patch width to patch length (W/L) does not exceed
2 since the aperture efficiency of a single patch begins to drop, as W/L increases
beyond 2.

310
- To increase bandwidth, increase the substrate height and/or decrease the substrate
permittivity (this will also affect resonant frequency and the impedance matching).
- To increase the input impedance, decrease the width of the feed lines attached directly
to the patches as well as the width of the lines attached to the port. The characteristic
impedance of the quarter-wave sections should then be chosen as the geometric mean
of half the impedance of the feed lines attached to the patches and the impedance of the
port lines.
Antenna Magus (see Fig. 14) is a software tool that helps choose the appropriate antenna for
a given application and estimates the S11 / VSWR and the far field gain characteristics.
Caution: Antennas on very thin substrates have high copper-losses, while thicker and
higher permittivity substrates may lead to performance degradation due to surface waves.
Although arrays are not directly used in cars, they are used in base stations for car to car
communication.
6. Modelling of a strip / mesh antenna on a windscreen
The proliferation of communication devices that are required in modern automobiles,
require automobile designers to include more and more antennas into their vehicle designs.
Requirements include FM/AM antennas, TV antennas, etc. Aesthetically speaking, this is a
problem that can only be overcome by including such antennas into vehicle designs in
unobtrusive ways. A prominent modern development is to include these antennas into the
windscreens of a vehicle. These windscreens include multiple layers of glass and wiring
that form the antenna. As with other antenna designs, engineers require the ability to
simulate new designs to evaluate many antenna operating characteristics, including:
Efficiency
Impedance bandwidth
Far-field radiation characteristics
FEKO includes a solution method based on the MoM that can be used for rigorous analysis
of windscreen antennas. The method meshes only the metallic antenna elements, so the
resource requirements that are devoted to modelling of the dielectric layers of the glass is
almost negligible. Features of the method include:
Boundaries of the dielectric interfaces between different layers of glass are accurately
accounted for.
Coupling between closely spaced antenna elements are taken into account.
Finite size glass antennas can be integrated into a full car model.
Curvature and rotation of the window is considered.

Fig. 15. Integrated windscreen antennas

311

Fig. 16. Current distribution due to integrated windscreen antennas
The windscreen can consist of one or more layers and the different layers do not have to be
meshed and thus simulation time is greatly reduced when compared to conventional
methods. Fig. 15 and Fig. 16 show a 3D representation of the car and windscreen being
simulated e.g. for current distribution, the input impedance / S11, etc. Besides the use of
integrated antennas on windscreens, these are also integrated to car tires, mirrors, and
bumpers for collision avoidance at the 76 GHz band.

Fig. 17. Comparison of Antenna reflection co-efficient (simulation /measured results)
The currents are calculated
Based on MoM solution with the incorporation of planar green function
Full consideration of:
Boundaries between dielectric layers of glass
Coupling between closely spaced antenna elements
Curved/rotated windscreens
Multiple windscreens

312
Using the aforesaid approach only the metallic parts need to be meshed and not the
dielectric parts of the windscreen elements of a car. Alternately the dielectric material i.e. the
windscreen can be modelled using various methods e.g. FEM. However this approach is
more time consuming as even the windscreen has to be meshed.
Results show that the planar Green's function approach (windscreen analysis - WA) is in
good agreement with the measured results. Fig.17 shows a fairly good agreement between
the simulation and the measured results.
7. References
Balanis, C. A. (2005). Antenna Theory, Analysis and Design, Wiley & Sons, Artech House, ISBN
978-0471603528, USA.
Kumar, G. & Ray, K.P. (2003). Broadband Microstrip Antennas, Artech House, ISBN 978-
1580532440, USA.
Garg, R. (2001). CAD for Microstrip Antennas Design Handbook, Artech House, ISBN, Artech
House
Sainati, R. A. (1996). CAD for Microstrip Antennas for Wireless Applications, Artech House,
Publisher, ISBN 978-0890065624, Boston.
Bancroft, R. (1996). Understanding Electromagnetic Scattering Using the Moment Method
A Practical Approach, Artech House, ISBN 978-0890068595, Boston.
Refer to FEKO by using the following information:
Author: EM Software & Systems - S.A. (Pty) Ltd
Title: FEKO (www.feko.info)
Suite: (the suite number reported by FEKO)
Publisher: EM Software & Systems - S.A. (Pty) Ltd
Address: PO Box 1354, Stellenbosch, 7599, South Africa
Refer to Antenna Magus by using the following information:
Author: Magus (Pty) Ltd
Title: Antenna Magus (www.antennamagus.com)
Version: (the version number reported by Antenna Magus)
Publisher: Magus (Pty) Ltd
Address: PO Box 1354, Stellenbosch, 7599, South Africa
17
Link Layer Coding for
DVB-S2 Interactive Satellite Services to Trains
Ho-Jin Lee
1
, Pansoo Kim
1
, Balazs Matuz
2
, Gianluigi Liva
2
,
Cristina Parraga Niebla
2
, Nuria Riera Daz
2
and Sandro Scalise
2

1
Broadcasting and communication convergence research division, ETRI,
161 Gajeong-dong Yuseong-gu Daejeon,
2
Institute of Communications and Navigation,
German Aerospace Center (DLR), 82234 Wessling,
1
Republic of Korea
2
Germany
1. Introduction
The growing number of railway passengers represents an appealing market for
multimedia services. Satellites could be used to fulfill these demands due to the large
coverage area and the low cost of associated terrestrial infrastructure. However,
transmissions to mobile users through satellite links always pose a big challenge,
especially since line of sight connection is frequently interrupted by obstacles between the
satellite and the mobile receiver. The railroad satellite channel (RSC) in particular suffers
from severe fadings that can be described using a combined statistical/deterministic
model. In this paper, we will focus on the DVB-S2 [1] forward link providing service to
high-speed trains. Additional protection of the data on link layer (LL) has been taken into
account to mitigate the fading effects. The LL coding scheme investigated in this paper is
based on the adoption of an erasure correcting code whose symbols are packets of
constant size. Examples of erasure correcting codes applied in satellite communication
systems can be found in [2][4]. The effort for the LL code design is mainly focused on the
mitigation of the fade events due electrical trellises or power arches (PA) that are placed
aside the tracks in order to provide the electric power to the trains along many railways.
Such events are frequent and nearly periodic. In [5] it has already been shown that
without a proper mitigation technique they would lead to an unacceptable quality of
service. The rest of the paper is organized as follows. In Section II. we will provide an
overview of the railroad satellite channel, focusing on the effect of electrical trellises on
the received signal power level. In Section III the overall system architecture is described.
Some insights on the link layer code design are provided as well. Section IV shows a
performance comparison between the proposed link-layer coding approach and an
enhancement of the DVB-S2 physical layer (PHY layer) through a long inter-frame
interleaver. Moreover, a further, simplified model for the railroad satellite channel is
introduced to give a basic understanding of the performance for the different solutions.
Concluding remarks follow in Section V.

314
2. Railroad satellite channel model
An appropriate model for the propagation channel in a railway environment can be derived
using the land mobile satellite channel (LMSC) as a reference scenario [6]: in the first
instance it is sufficient to characterize the channel behavior by two different states, i.e. a line
of sight (LOS) state with relatively high received signal power and a non line of sight
(NLOS) state where the signal is shadowed or blocked by objects in the vicinity of the
receiver. In the former state, the received signal is composed of a direct and a multipath
component, with the instantaneous received signal power S obeying a Ricean probability
density function:
( ) ( ) ( ) ( ) 0
exp 1 2 .
Rice
p S c c S I c S = +
Here, c denotes the so-called Rice factor, i.e. the direct-tomultipath signal power ratio and
0
I is the modified Bessel function of order zero. In the NLOS state, with no direct signal path
present, the signal power shows Rayleigh behavior around a short-term mean value S
0
with
the PDF described by:
( ) ( )
0 0
0
1
exp .
Rayl
p S S S S
S
=
For the short-term mean
0
S a lognormal distribution is assumed:
( )
( )
2
0
0 2
0
10log
10 1
exp ,
2 2 ln10
dB
LN
dB dB
S
p S
S

=

with
dB
describing the average power level (in dB) and
dB
the variance of the power
level (in
2
dB ) due to large scale fading. The railroad satellite channel has some
peculiarities that have not been modeled properly by the previous description.
Measurement campaigns show that a constant attenuation of 2-3 dB is introduced by
catenaries above the tracks. Also, long fades occur mainly due to structures like bridges or
tunnels, and shorter but periodic ones that are caused by several metallic obstacles along
the railroad. Among them there are posts (with or without brackets), electrical trellises or
arches spanning over the tracks, but they will be simply referred to as power arches for
the rest of this paper. In the sequel, we will restrict ourselves on a RSC model
corresponding to the LMSC in the LOS case superimposed by short deep fades ascribed to
the power arches. Since these fades are nearly-periodic (and thus deterministic), they are
not suitable for a statistical characterization. In turn, the modeling approach proposed in
[5], here recalled for sake of clarity, will be adopted. The attenuation introduced by the
above-mentioned power arches can be accurately described using the knife-edge
diffraction theory. The knife-edge attenuation describes the ratio between the received
electro-magnetic field
D
E in presence of an obstacle and the received field under free
space conditions
0
E . For an object of two finite dimensions, it can be represented as sum
of two diffracted signal components:
Link Layer Coding for DVB-S2 Interactive Satellite Services to Trains

315
( ) ( )
2 2
1 2
2 2
0
1
,
2
d v
j t j t
v
D
h
v
MAX MAX
G G j E
e dt e dt
E G G

+
= +

where d is the width of the obstacle, h the height above LOS and ( ) /
MAX
G G denotes the
radiation pattern of the directive antenna. Moreover, the Fresnel parameter v can be
calculated out of h, the wavelength and the distance
1
d between the receiver and the
object, as well as the distance
2
d between the object and the satellite according to:
( )
1 2
1 2
2
.
d d
v h
d d
+
=
Following that, the railroad satellite channel which is the basis for all further investigations
looks like depicted in Figure 1. It comprises Ricean fading (with a Rice factor of 18 dB), as
well as periodic deep fades as a result of equally spaced power arches aside the railway. The
worse scenario, where additional NLOS is present according to the LMSC model, will not be
considered in the sequel.

Fig. 1. RSC realization with a power arch distance of 50m at a speed of 300km/h; periodic
deep fades occur every time the train passes by a PA
3. System description

Fig. 2. Transmission chain used for the simulation

316
Our approach to mitigate the impairments of the RSC is based on link layer coding. For
illustration purposes, a simplified block diagram of the link is depicted in Figure 2. In our
case LL coding is applied on MPEG-TS packets. The systematic link layer encoder receives
at its input K MPEGTS packets and produces at its output N packets, referred to as LL
codeword. Due to the systematic nature of the code, this codeword consist of the K input
MPEG-TS packets followed by M = N K parity packets. On the receiver side, the
decoder takes care of recovering lost MPEG-TS packets. Note that in the link layer coding
framework the code works on an erasure channel, where the erased units (i.e., the LL
codeword symbols) are whole packets. In the context of this paper, such feature is
guaranteed by the error detection performed after physical layer decoding on each
MPEGTS/parity packet. In other words, the PHY decoder attempts to protect individual
MPEG-TS packets, whereas the LL decoder is meant to recover lost MPEG-TS packets. The
LL decoder design is highly facilitated by the underling erasure channel, permitting for
some kind of codes the adoption of software based decoding up to several tens of Mbytes
per second. To allow this appealing feature, our investigation is focused on the adoption
of low-density parity-check (LDPC) codes [7] as erasure correcting codes. LDPC codes
provide capacity approaching performance on many communication channels [8], and in
the framework of LDPC codes some astonishing erasure correcting codes have been
developed [9][11]. The LDPC codes adopted for the simulations belong to the family of
the so-called irregular repeat-accumulate (IRA) codes [12]. IRA codes allow simple
efficient encoding while keeping nearcapacity performance. The design of the IRA codes
have been optimized through extrinsic information transfer (EXIT) analysis [13][15],
constraining the parity-check matrix of the code to a block-circulant form that would also
permit a simple hardware decoder design. Further performance enhancements with
similar encoding complexities are expected by adopting more sophisticated IRA-like
designs [16] [17]. After link layer coding is done, the MPEG-TS packets within each LL
codeword are interleaved (denoted by
1
in Figure 2) to break up channel correlations.
For our investigations we limited the maximum length of the link layer codeword in a
way that it spans over to 200 ms. For example, taking into account a symbol rate of 27.5
MBaud, QPSK modulation and physical layer code rate r = 1/2, the LL codeword length
would be N = 3400 packets. This constraint has been introduced to avoid long delays
which could affect real-time applications. The stream of MPEG-TS/parity packets is then
forwarded to a DVB-S2 transmitter, which takes care of the physical layer coding through
the serial concatenation of a BCH (Bose, Ray-Chaudhuri, Hocquenghem) code and LDPC
code according to the DVB-S2 standard (for details see [1]). The physical layer codeword
size corresponds to the large frame size of the DVB-S2 standard (64800 bits). Before
forwarding the data to radio frequency (RF) frontend, we allow a further (optional)
physical layer inter-frame block interleaver
2
that permutes the bits among several
frames. For the sake of comparison, in the following we will consider also the scenario
where the LL coding block is disabled. In such case, the diversity necessary to overcome
the short periodic fade events will be provided by the inter-frame interleaver only.
However, the interleaver latency will be constrained to be lower than 200 ms.
Furthermore, to keep the comparison fair, physical layer code rate will be lowered in a
way that the overall efficiency of the two systems is the same.

317
4. Outcomes
In section IV, we provide some numerical results obtained through Monte Carlo
simulations on the RSC described in Section II. The analysis is focused on the case of LOS
conditions with superimposed power arches. The performance is depicted in terms of
MPEG-TS packet error rate (PER) versus signal-to-noise ratio (SNR)
0
/
s
E N . Here,
s
E denotes the energy per modulated symbol and
0
N the one-sided noise power spectral
density. For the LL coded solution, the results are shown in terms of residual MPEG-TS
packet error rate at the output of the LL decoder vs.
0
/
s
E N . Besides the PA width and
the train velocity that have a high impact on the fade duration, also other factors have to
be taken into account for the simulations, such as the current latitude and the traveling
direction of the train. To simplify the simulations it is advisable to determine an effective
power arch width that takes into account these factors. For a PA width of 30 cm and a
latitude of 38
o
geometric considerations yield to an effective PA width of roughly 87cm.
Considering train speeds from 30 km/h to 150 km/h and a north-south traveling
direction, the resulting fade durations range from ~100 ms to ~20 ms. The distance
between two subsequent PA is constantly set to 50 m.
4.1 Simulation results
Assuming only PHY layer coding with a code rate of 1/4 combined with physical layer
interleavers of different lengths we obtain the plots in Figure 3 (at 30 km/h on the left, at 150
km/h on the right). For both speeds, intermediate error floors arise at error rates
proportional to the interleaver duration (in the charts, the performance with interleaver
lengths of 200 ms, 100 ms, 50 ms and no interleaving are depicted). However, at high speeds
the rate 1/4 code in combination with a sufficiently long interleaver is able to overcome the
floor, but the steepness of the curve in the subsequent waterfall region remains quite poor.
For low speeds the error floors remain, for all the investigated interleaver lengths and also
for relatively high SNRs. The outcomes for joint physical/link layer coding with rate 1/2
codes on both layers and with a LL interleaver duration of 200 ms combined with different
PHY layer interleaver durations are shown in Figure 4 (at 30 km/h on the left, at 150 km/h

Fig. 3. PER vs Es/No with only physical layer protection(DVB-S2, QPSK, r=1/4) and
different interleaving depths. Overall spectral efficiency of 0.5 bps/Hz. Speed of 30km/h
(left) and 150km/h (right)

318
on the right). The best results can be achieved by using no physical interleaver at all. In
this case the LL code is able to overcome the errorfloor at both speeds and ensures a steep
slope of the PER curve in the waterfall region. Compared to plain physical layer coding
with inter-frame interleaving, joint physical/link layer coding clearly shows an
improvement of performance. Note that for joint PHY/LL coding both interleavers
1
and
2
are synchronized, so that the overall delay that is experienced due to interleaving is
equivalent to the maximum of the delays introduced by
1
and
2
(in our case not more
than 200 ms).

Fig. 4. PER vs Es/No with joint physical layer protection(DVB-S2, QPSK, r=1/2) and link
layer Protections(link layer LDPC code, R=1/2) and different interleaving depths. Speed of
30km/h (left) and 150km/h (right). Overall spectral efficiency of 0.5bps/Hz
4.2 Insights on the use of PHY layer interleavers on the RSC

Fig. 5. Main concept of the AWGN/EC-AWGN channel model for the RSC channel
The behavior of a coding scheme including physical layer coding and long inter-frame
interleavers on the LOS channel with superimposed blockages can be easily understood by
splitting the contributions to the packet error probability
e
P (that is the stochastic equivalent
to the simulated PER) into two parts, the LOS error rate and the blockage error rate. The
LOS condition is referred as the good state (G). To simplify the analysis in such state, the
channel characteristic is approximated by an AWGN channel (recall the high Rice factor in
LOS). The blockage condition ascribed to PAs is referred to as the bad state (B). Here, the
channel is basically a bursty erasure channel. Assuming interleaving windows longer than a

319
PA fade duration (thus, spreading the erasures on a wider duration), the PHY layer decoder
deals with a combination of an erasure channel with AWGN (EC-AWGN). This state spans
over a whole interleaver window. As an illustration the overall channel model is depicted in
Figure 5. Denoting by X the channel state random variable, the stationary probabilities of
being in the good/bad state are given by
and
G B
G B
B G B G
T T
P P
T T T T
= =
+ +

where ( )
G B
T T represents the time spent in the good (bad) state. Assuming periodic fade
events due to the power arches,
B
T shall be replaced by the interleaver length L,
expressed in seconds, or by the fade duration
f
T , in case no interleaving is applied. For
sake of simplicity, lets summarize such parameter as
B
. The sum
B G
T T + has to be
replaced by the power arch periodicity , while
G
T becomes the time interval between
two interleaving windows affected by consecutive power arch fades,
G
. Lets define
Pr{ } E X G = and Pr{ } E X B = as the packet error conditional probabilities given the good
(bad) state, where E denotes the packet error event. The error probability
e
P can be
therefore expressed as

{ } { }
{ } { }
Pr Pr
Pr Pr .
e G B
G B
G B G B
P P E X G P E X B
E X G E X B
= = + =

= = + =
+ +
(1)
Note that, due to the parameters chosen for the simulations,
2
~ 10
B G
. Consequently,
B
P is in the order of magnitude of
2
10
. Thus, at low SNRs, where the error rates are high

even in LOS conditions, the first term in (1) dominates the summation. At high SNRs, the
error rate Pr{ } E X G = in LOS condition quickly decreases. The second term becomes
therefore dominant. The performance curve of the system in such conditions can be
composed in a two-fold fashion:
1. The error probability Pr{ } E X G = in LOS conditions is evaluated numerically down to
error rates that are negligible respect to
B
P . With the current scenario of
2
~ 10
B
P

the
simulation can be stopped once Pr{ } E X G = approaches
3
10

2. As a second step the probability Pr{ } E X B = for the EC-AWGN channel has to be
computed. In case no interleaving is applied,
B
is equal to
f
T . In this interval
Pr{ } E X G = can be reasonably set to 1. Otherwise, the problem of computing
Pr{ } E X B = reduces to the performance evaluation of the channel code on the
ECAWGN, where, assuming random interleaving, each bit soft-value is erased with a
probability , with /
f
T L = in the interval
B
L = .
The two error probabilities are then combined on the same chart following equation (1). This
is exemplified in Figure 6. The LL code employed for the simulation is a short (2048,1024)
LDPC code. The impact of the physical layer interleaver length becomes therefore quite
clear: large interleavers lower the erasure rate (recall that /
f
T L = ) of the codeword bits in
blockage conditions, increasing the steepness of the Pr{ } E X B = curve. At the same time,
high values of L rise up the intermediate floor, at the error rate given by /
B
P L = . This is

320
compliance with the results presented in charts 3 and 4: the higher the interleaver length, the
higher the intermediate error floor. As it can be seen, there are some exceptions, since the
same error floor arises for the 50 ms and 100 ms interleavers. This is due to the fact that PA
fades affect two interleaver windows for the 50 ms case (i.e., the interleaving window has a
length which comparable to the fade duration). In case of a scheme employing a link level
code on top of the physical layer, it is often advisable to abstain from the use of physical
layer interleavers to keep the intermediate error floors as low as possible (especially at low
train speeds). The packet recovery task in presence of the PA fades is then left to the LL
code.

Fig. 6. Packet error rate for the simplified RSC (hybrid AWGN/EC-AWGN model)
5. Concluding remarks
The main purpose of this work was to draw a comparison between physical layer coding
and interleaving and the innovative approach of joint physical/link layer coding for the
railroad satellite channel. It was shown that the system performance can be highly
improved for this type of channel by splitting redundancy on different layers. Employing
link layer coding shows some performance advantages with respect to the use of long
physical layer interleavers, especially in case of frequent short blockages. A simplified
model for the LOS railroad satellite channel with superimposed periodic fades was
introduced, with focus on the performance of a scheme employing physical layer coding
enhanced by long inter-frame interleavers. The proposed model allows a precise
calculation of the arising error floors, as well as simple explanation of the system behavior
for different interleaver lengths and train velocities. This knowledge turns out to be very
helpful for the code design.

321
6. Acknowledgments
This work was supported by the ETRI-DLR Collaborative Research under the name of
Communication Technologies for Satellite Broadband Mobile based on DVB-S2/RCS. The
work has been presented in part at VTC 2008 spring[18].
7. References
[1] Digital Video Broadcasting (DVB): Second generation framing structure,channel
coding and modulation systems for Broadcasting, Interactive Services, News
Gathering and other broadband satellite applications, ETSI Std. EN 302 307,
2004.
[2] C. Di, H. Ernst, E. Paolini, S. Coletto, and M. Chiani, Low-density parity-check codes for
the transport layer of satellite broadcast, in Proc. AIAA International Communications
Satellite Systems Conference(ICSSC 2005), Rome, Italy, Sep. 2005.
[3] M. Chiani, G. Liva, and E. Paolini, Investigation of long erasure codes for space
communication protocols, CCSDS, Rome, Tech. Rep., June 2006, spring Meeting.
[4] E. Paolini, G. Liva, M. Chiani, and G. Calzolari, Tornado-like codes: a new appealing
chance for space applications protocols?, in 3rd European Space Agency Workshop
on Tracking, Telemetry and Command Systems for Space Applications, TTC 2004, Sep.
2004.
[5] S. Scalise, R. Mura, and V. Mignone, Air Interfaces for Satellite Based Digital TV
Broadcasting in the Railway Environment, in IEEE Transactions on Broadcasting,
IEEE, Ed., vol. 52, no. 2, June 2006, pp. 158166.
[6] E. Lutz, M. Werner, and A. Jahn, Satellite Systems for Personal andBroadband
Communications. Springer Verlag, 2000.
[7] R. G. Gallager, Low-Density Parity-Check Codes. Cambridge, MA: M.I.T. Press, 1963.
[8] T. Richardson and R. Urbanke, The capacity of low-density parity check codes under
message-passing decoding, IEEE Trans. Inform Theory, vol. 47, 2001.
[9] M. Luby, LT-codes, in Proc. of the ACM Symposium on Foundations of Computer Science
(FOCS), 2002.
[10] E. Paolini, M. Fossorier, and M. Chiani, Analysis of doubly-generalized LDPC codes
with random component codes for the binary erasure channel, in Proc. of Allerton
Conf. on Communications, Control and Computing, Monticello, USA, Sep. 2006.
[11] A. Shokrollahi, Raptor codes, IEEE Transactions on Information Theory, vol. 52, no. 6, pp.
25512567, Jun. 2006.
[12] H. Jin, A. Khandekar, and R. McEliece, Irregular repeat-accumulate codes, in Proc.
International Symposium on Turbo codes and Related Topics, Sep. 2000, pp. 18.
[13] S. ten Brink, Convergence behavior of iteratively decoded parallel concatenated codes,
IEEE Trans. Commun., vol. 49, pp. 17271737, Oct. 2001.
[14] G. Liva and M. Chiani, Protograph LDPC codes design based on EXIT analysis, in Proc.
IEEE Globecomm, Nov. 2007.
[15] G. Liva, S. Song, L. Lan, Y. Zhang, W. Ryan, and S. Lin, Design of LDPC codes: A
survey and new results, J. Comm. Software and Systems, Sep. 2006.

322
[16] A. Abbasfar, K. Yao, and D. Disvalar, Accumulate repeat accumulate codes, in Proc.
IEEE Globecomm, Dallas, Texas, Nov. 2004.
[17] G. Liva, E. Paolini, and M. Chiani, Simple reconfigurable low-density parity-check
codes, IEEE Comm.Letters, vol. 9, pp. 258260, March 2005.
[18] B. Matus, Link Layer Coding for DVB-S2 Interactive Satellite Services to Trains, in Proc.
IEEE VTC, Sigapore, May. 2008
18
Mobility Aspects of Physical Layer in Future
Generation Wireless Networks
Asad Mehmood and Abbas Mohammed
Blekinge Institute of Technology Karlskrona
Sweden
1. Introduction
The demand from social market for high speed broadband communications over wireless
media is pushing the requirements of both the mobile and fixed networks. The past decade
has witnessed tremendous advancement in the blooming development of mobile
communications including mobile-to-mobile and mobile-to-fixed networks. Wireless fixed
and cellular networks of future generation will need to support new protocols, standards and
architecture leading to all IP-based networks. Different systems like digital video
broadcasting (DVB) via satellites have great success commercially as they provide ubiquitous
coverage and serve large number of users with high signal quality. Satellite communications
have proven to be attractive means to provide communication services such as broadband
communications (3G services), surveillance, remote monitoring, intelligent transportation
systems, navigation, traffic warnings and location-based information etc. to fixed and mobile
users. However, to meet the growing demands of mass market integration of satellites and
terrestrial networks seems to be inevitable for future generation wireless networks.
Due to technology advances and growing traffic demands, communication systems must
evolve to completely new systems or within themselves in order to provide broadband
services in a safe and efficient way. While enhancements continue to be made to leverage
the maximum performance from currently deployed systems, there is a bound to the level to
which further improvements will be effective. If the only purpose were to deliver superior
performance, then this in itself would be relatively easy to accomplish. The added
complexity is that such superior performance must be delivered through systems which are
cheaper from installation and maintenance prospect. Users have experienced an incredible
reduction in telecommunications charges and they now anticipate receiving higher quality
communication services at low cost. Therefore, in deciding the subsequent standardization
step, there must be a dual approach: in search of substantial performance enhancement but
at reduced cost. Long Term Evolution (LTE) is that next step and will be the basis on which
future mobile telecommunications systems will be built. LTE is the first cellular
communication system optimized from the outset to support packet-switched data services,
within which packetized voice communications are just one part.
In case of highly mobile scenarios, the effects of signal blockages and Doppler shifts
introduce more burdens on the receiver demodulator. The signals blockage is prominent in
the case of land mobile communications as compared to satellite communications. In
deciding the technologies to comprise in LTE, one of the key concerns is the trade-off

324
between cost of implementation and practical advantage. Fundamental to this assessment,
therefore, has been an enhanced understanding different scenarios of the radio propagation
environment in which LTE will be deployed and used.
The organization of the chapter is as follows. In section 2, different mobility aspects related
to the physical layer of future generation mobile communication networks are discussed.
Section 3 discusses the propagation scenarios in which LTE will be deployed. Section 4
describes space-time processing techniques to enhance the system performance. In section 5
LTE systems performance is evaluated at different mobile speeds. Finally, section 6
concludes the chapter.
2. Physical layer aspects
The high data rate multimedia broadcast/multicast services at cheap rates with appropriate
quality-of-service (QoS), fast handoff techniques and wide area seamless mobility pave the
way for future generation wireless communications. Wireless network operators require
different schemes for including new services to take benefits from new access technologies.
Fundamental to these strategies is to incorporate mobility that can bring unique advantages
to mobile users. In response to these requirements, the wireless industry is foreseen to shift
toward LTE and world wide interoperability of microwave access (WiMAX) technologies to
be able to support cost effectively the capacity required by mobile operators to meet mass
market demands of data services (Motorola, 2010). LTE must be able to provide superior
performance compared to the existing wireless network infrastructures which suffer from
cell-edge performance, spectral efficiency and desired QoS to end users. In order to provide
high data rates with high QoS in already crowded spectrum, LTE is susceptible to different
impairments: noise and interference. Therefore to mitigate these propagation impairments,
efficient and robust techniques need to be adapted to take full benefits of the technology. A
thoughtful design of physical layer aspects to mitigate these propagation impairments and
improve the system performance is thus crucial for successful operation and support of the
desired QoS.
2.1 Objectives of physical layer
The objectives of LTE physical layer are the significant increase in peak data rates up to
100 Mb/s in downlink and 50 Mb/s in uplink within 20 MHz spectrum leading to spectrum
efficiency of 5 Mb/s, increased cell-edge performance maintains site locations as in Wide
Band Code Division Multiple Access (WCDMA), reduced user and control plane latency to
less than 10 ms and less than 100 ms, respectively (Kliazovich1, et al.). LTE will be able to
provide interactive real-time services such as high quality video/audio conferencing and
multiplayer gaming with mobility support for up to 350 km/h or even up to 500 km/h and
reduced operation cost. It also provides a scalable bandwidth 1.25/2.5/5/10/20 MHz in
order to allow flexible technology to coexist with other standards, 2 to 4 times improved
spectrum efficiency the one in Release 6 HSPA to permit operators to accommodate
increased number of customers within their existing and future spectrum allocation with a
reduced cost of delivery per bit, low power consumption and acceptable system and
terminal complexity. The system should be optimized for low mobile speed but also support
high mobile speed as well. In this section we will discuss some of the features included in
LTE physical layer to mitigate propagation impairments.
Scalable OFDMA: Multiple access schemes are used in multi-user communications to
provide on-demand data rates to users by sharing the available resources in available finite
Mobility Aspects of Physical Layer in Future Generation Wireless Networks

325
bandwidth. The orthogonal frequency division multiple access (OFDMA) is used as
multiple access scheme in the downlink and single carrier frequency division multiple
access (SC-FDMA) is used in the uplink. OFDMA is OFDM based multiple access technique
used for LTE to facilitate the exploitation of multi-user diversity, frequency diversity and
flexible users scheduling to enhance the system capacity in challenging multi-user
communications with wide range of applications, data rates and QoS requirements. The
flexible structure of OFDMA allows efficient implementation of space-time processing
techniques, e.g., multiple-input multiple-output (MIMO) with reasonable complexity. The
scalable bandwidth with different FFT sizes and dynamic subcarrier allocation allows the
efficient use of spectrum in different regional regulations for mobile applications.
Frame Structure and Transmission Modes: LTE supports two types of frame structures:
type1 frame structure which is designed for frequency division duplex (FDD) and is valid
for both half duplex and full duplex FDD modes. Type 1 radio frame has a duration 10 ms
and consists of 20 slots each of 0.5 ms. A sub-frame comprises two slots, thus one radio
frame has 10 sub-frames. In FDD mode, half of the sub-frames are available for downlink
and the other half are available for uplink transmission in each 10 ms interval, where
downlink and uplink transmission are separated in the frequency domain (3GPP, 2008).
Type 2 frame structure is applicable for time division duplex mode (TDD). The radio frame
is composed of two identical half-frames having duration of 5 ms. Each half-frame is further
divided into 5 sub-frames having duration of 1 ms. Two slots of length 0.5 ms constitute a
sub-frame which is not special sub-frame. The special type of sub-frame is composed of
three fields Downlink Pilot Timeslot (DwPTS), GP (Guard Period) and Uplink Pilot Timeslot
(UpPTS). Seven uplink-downlink configurations are supported with both types (10 ms and 5
ms) of downlink-to-uplink switch-point periodicity. In 5 ms downlink-to-uplink switch-
point periodicity, special type of sub-frames are used in both half-frames but it is not the
case in 10 ms downlink-to-uplink switch-point periodicity, special frame is used instead of
are used only in first half-frame. For downlink transmission sub-frames 0, 5 and DwPTS are
always reserved. UpPTS and the sub-frame next to the special sub-frame are always
reserved for uplink communication (3GPP, 2009).
Mobility Support: One of the features of LTE is appropriate physical layer design to facilitate
users at high vehicular speeds to support delay sensitive applications (e.g., VOIP) with
appropriate QoS. The physical layer features such as power control, hybrid automatic repeat
request (HARQ), sub-channelization and pilot structure are used to mitigate the fluctuations
in the received signal caused by channel fast fading. In addition, link adaptation technique is
used to adjust system parameters according to channel dynamics, i.e, to select appropriate
parameters under available propagation conditions. This permits to optimize the spectral and
power sources of the system under poor propagation conditions.
Advanced Antenna Techniques: Multiple antenna systems based on space-time processing
algorithms have brought great benefits to wireless communications by exploiting the spatial
domain to use the resources in efficient way. Advanced antenna techniques such as
diversity techniques, spatial multiplexing and beamforming are employed to create
independent multiple parallel channels which result in overall system improvement in
terms of link reliability, high capacity, extended coverage and reduced transmitted power.
LTE uses advanced antennas techniques in both single-user and multi-user MIMO cases.
Link Adaptation and Channel Coding: Link adaptation is used to adjust the system
parameters in time varying propagation conditions to facilitate users at different data rates.
Thus link adaptation scheme is very closely related to channel coding schemes used for

326
forward error correction (Sesia, et al. 2009). LTE schedules down link data transmission and
selects modulation and coding schemes based on the feedback information in terms of
signal-to-interference plus noise ratio (SINR) provided by channel quality indicator (CQI) in
uplink direction. The LTE specifications define the signalling between user terminal and
eNodeB for link adaptation and switching between different modulation schemes and
coding rates that depend on several factors including cell throughput and required QoS.
Scheduling and Quality-of-Service: The purpose of scheduling is to manage the resources
in uplink and downlink channels while maintaining the desired QoS according to user
expectations. In LTE eNodeB performs this operation. The principle of scheduling algorithm
is to allocate the resources and transmission powers in order to optimize certain set of
parameters such as throughput, user spectral efficiency, average delay and outage
probability. The LTE MAC layer can support large number of users with desired QoS.
3. Radio propagation models
From the beginning of wireless communications there is a high demand for realistic mobile
fading channels. The reason for this importance is that efficient channel models are essential
for the analysis, design, and deployment of communication systems for reliable transfer of
information between two parties. Realistic channel models are also significant for testing,
parameter optimization and performance evolution of communication systems. The
performance and complexity of signal processing algorithms, transceiver designs and smart
antennas etc., employed in future mobile communication systems, are highly dependent on
design methods used to model mobile fading channels. Therefore, correct knowledge of
mobile fading channels is a central prerequisite for the design of wireless communication
systems (Rappaport, 1996; Ibnkahla, 2005; Ojanpera, et al., 2001).
The difficulties in modeling the wireless channel are due to complex propagation processes.
A transmitted signal arrives at the receiver through different propagation mechanisms as
shown in Figure 1. The propagation mechanisms involve the following basic mechanisms: i)
free space or line of sight (LOS) propagation ii) specular reflection due to interaction of
electromagnetic waves with plane and smooth surfaces which have large dimensions as
compared to the wavelength of interacting electromagnetic waves iii) Diffraction caused by
bending of electromagnetic waves around corners of buildings iv) Diffusion or scattering
due to contacts with objects having irregular surfaces or shapes with sizes of the order of
wavelength v) Transmission through objects which cause partial absorption of energy
(Oestges, et al., 2007; Rappaport, 1996). It is significant here to note that the level of
information about the environment a channel model must provide is highly dependent on
the category of communication system under assessment. To predict the performance of
narrowband receivers, classical channel models which provide information about signal
power level distributions and Doppler shifts of the received signals, may be sufficient. The
advanced technologies (e.g., UMTS and LTE) build on the typical understanding of Doppler
spread and fading also incorporate new concepts such as time delay spread, direction of
departures (DOD), direction of arrivals (DOA) and adaptive antenna geometry (Ibnkahla,
2005). The presence of multipaths (multiple scattered paths) with different delays,
attenuations, DOD and DOA gives rise to highly complex multipath propagation channel.
Figure 2 illustrates power delay profile (PDP) of a multipath channel with three distinct
paths.

327

Fig. 1. Signal propagation through different paths showing multipath propagation
phenomena

Power

1

2

3
Delay

Fig. 2. Power delay profile of a multipath channel
3.1 Propagation aspects and parameters
The behaviour of a multipath channel needs to be characterized in order to model the
channel. The concepts of Doppler spread, coherence time, delay spread and coherence
bandwidth are used to describe various aspects of the multipath channel.
3.1.1 Delay spread
To measure the performance capabilities of a wireless channel, the time dispersion or
multipath delay spread related to small scale fading of the channel needs to be calculated
in a convenient way. One simple measure of delay spread is the overall extent of path
delays called the excess delay spread. This is an appropriate way because different
channels with the same excess delay can exhibit different power profiles which have more
or less impact on the performance of the system under consideration. A more efficient
method to determine channel delay spread is the root mean square (rms) delay spread
(
rms
) which is a statistical measure and gives the spread of delayed components about the
mean value of the channel power delay profile. Mathematically, rms delay spread can be
described as second central moment of the channel power delay profile (Rappaport, 1996)
which is written as:

328

N 1
2
n n m
n 0
rms N 1
n
n 0
P ( )
P
=

=

(1)
where,
1
0
1
0
N
n n
n
m N
n
n
P
P
is the mean excess delay.

3.1.2 Coherence bandwidth
When the channel behaviour is studied in frequency domain then coherence bandwidth
c
f
is of concern. The frequency band, in which the amplitudes of all frequency components of
the transmitted signal are correlated, i.e., with equal gains and linear phases, is known as
coherence bandwidth of that channel (Ibnkahla, 2005). The channel behaviour remains
invariant over this bandwidth. The coherence bandwidth varies in inverse proportion to the
delay spread. A multipath channel can be categorized as frequency flat fading or frequency
selective fading in the following way.
Frequency flat fading: A channel is referred to as frequency flat if the coherence
bandwidth
c
f >>B, where B is the signal bandwidth. In this case frequency components of
the signal will experience the same amount of fading.
Frequency selective fading: A channel is referred to as frequency selective if the coherence
bandwidth
c
f B . In this case different frequency components will undergo different
amount of fading. The channel acts as a filter since the channel coherence bandwidth is less
than the signal bandwidth; hence frequency selective fading takes place (Fleury, 1996).
3.1.3 Doppler spread
The Doppler spread arises due to the motion of mobile terminal. Due to the motion of
mobile terminal through standing wave the amplitude, phase and filtering applied to the
transmitted signal vary with time according to the mobile speed (Cavers, 2002). For an
unmodulated carrier, the output is time varying and has non-zero spectral width which is
Doppler spread. For a single path between the mobile terminal and the base station, there
will be zero Doppler spread with a simple shift of the carrier frequency (i.e., Doppler
frequency shift) at the base station. The Doppler frequency depends on the angle of
movement of the mobile terminal relative to the base station.
3.1.4 Coherence time
The time over which the characteristics of a channel do not change significantly is termed as
coherence time. The reciprocal of the Doppler shift is described as the coherence time of the
channel. Mathematically we can describe coherence time as:

c
rms
1
T
2
=
(2)
where
rms
is root mean square vale of Doppler spread.

329
The coherence time is related to the power control schemes, error correction and
interleaving schemes and to the design of channel estimation techniques at the receiver.
4. Standard channel models
Standard channel models can be developed by setting up frame work for generic channel
models and finding set of parameters that need to be determined for the description of the
channel. Another method is to set up measurement campaigns and extracting numerical
values of parameters and their statistical distributions (Meinil, et al., 2004).
When designing LTE, different requirements are considered: user equipment (UE) and base
station (BS) performance requirements which are crucial part of LTE standards, Radio
Resource Management (RRM) requirements to ensure that the available resources are used
in an efficient way to provide end users the desired quality of service, the RF performance
requirements to facilitate the existence of LTE with other systems (e.g., 2G/3G) systems
(Holma, et al., 2009). The standard channel models play a vital role in the assessment of
these requirements. In the following section, some standard channel models are discussed
which are used in the design and evaluation of the UMTS-LTE system.
4.1 SISO, SIMO and MISO channel models
COST projects, Advanced TDMA (ATDMA) Mobile Access, UMTS Code Division Testbed
(CODIT) conducted extensive measurement campaigns to create datasets for SISO, SIMO
and MISO channel modeling and these efforts form the basis for ITU channel models which
are used in the development and implementation of the third generation mobile
communication systems (Sesia, et al., 2009). COST stands for the European Co-operation in
the Field of Scientific and Technical Research. Several Cost efforts were dedicated to the
field of wireless communications, especially radio propagation modeling, COST 207 for the
development of Second Generation of Mobile Communications (GSM), COST 231 for GSM
extension and Third Generation systems, COST 259 Flexible personalized wireless
communications (1996-2000) and COST 273 Towards mobile broadband multimedia
networks (2001-2005). These projects developed channel models based on extensive
measurement campaigns including directional characteristics of radio propagation (Cost 259
and Cost 273) in macro, micro and picocells and are appropriate for simulations with smart
antennas and MIMO systems. These channel models form the basis of ITU standards for
channel models of Beyond 3G systems. Detailed study of COST projects can be found in
(Molisch, et al., 2006; Corria, 2001).
The research projects ATDMA and CODIT were dedicated to wideband channel modelling
specifically channel modelling for 3
rd
generation systems and the corresponding radio
environments. The wideband channel models have been developed within CODIT using
physical-statistical channel modelling approach while stored channel measurements are
used in ATDMA which are complex impulse responses for different radio environments.
The details of these projects can be found in (Ojanpera, et al., 2001).
4.2 ITU multipath channel models
The ITU standard multipath channel models proposed by ITU (ITU-R, 1997) used for the
development of 3G 'IMT-2000' group of radio access systems are basically similar in
structure to the 3GPP multipath channel models. The aim of these channel models is to

330
develop standards that help system designers and network planners for system designs and
performance verification. Instead of defining propagation models for all possible
environments, ITU proposed a set of test environments in (ITU-R, 1997) that adequately
span the all possible operating environments and user mobility. In this chapter we use ITU
standard channel models for pedestrian and vehicular environments.
4.2.1 ITU Pedestrian-A, B
In both Pedestrian-A and Pedestrian-B channel models the mobile speed is considered to be
3 km/h. For Pedestrian models the base stations with low antennas height are situated
outdoors while the pedestrian users are located inside buildings or in open areas. Fading
can follow Rayleigh or Rician distribution depending upon the location of the user. The
number of taps in case of Pedestrian-A model is 3 while Pedestrian-B has 6 taps. The
average powers and relative delays for the taps of multipath channels based on ITU
recommendations are given in Table 1 (ITU-R, 1997).
4.2.2 ITU Vehicular-A (V-30, V-120 and V-350)
The vehicular environment is categorized by large macro cells with higher capacity, limited
spectrum and large transmit power. The received signal is composed of multipath
reflections without LOS component. The received signal power level decreases with
distance for which path loss exponent varies between 3 and 5 in the case of urban and
suburban areas. In rural areas path loss may be lower than previous while in mountainous
areas, neglecting the path blockage, a path loss attenuation exponent closer to 2 may be
appropriate.
For vehicular environments, the ITU vehicular-A channel models consider the mobile
speeds of 30 km/h, 120 km/h and 350 km/h. The propagation scenarios for LTE with
speeds from 120 km/h to 350 km/h are also defined in (Ericsson, et al., 2007) to model high
speed scenarios (e.g., high speed train scenario at speed 350km/h). The maximum carrier
frequency over all frequency bands is f=2690 MHz and the Doppler shift at speed v=350
km/h is 900 Hz. The average powers and relative delays for the taps of multipath channels
based on ITU recommendations are given in Table 2 (ITU-R, 1997).

Tap
No
Pedestrian-A Pedestrian-B Doppler
Spectrum
Relative Delay
(ns)
Average
Power(dB)
Relative Delay
(ns)
Average
Power(dB)

1 0 0 0 0 Classical
2 110 -9.7 200 -0.9 Classical
3 190 -19.2 800 -4.9 Classical
4 410 -22.8 1200 -8 Classical
5 NA NA 2300 -7.8 Classical
6 NA NA 3700 -23.9 Classical
Table 1. Average Powers and Relative Delays of ITU multipath Pedestrian-A and
Pedestrian-B cases

331

Tap No

Average Power(dB) 0 -1.0 -9.0 -10.0 -15.0 -20.0
Relative Delay(ns) 0 310 710 1090 1730 2510
Table 2. Average Powers and Relative Delays for ITU Vehicular-A Test Environment.
5. Multiple antenna techniques
Broadly, multiple antenna techniques utilize multiple antennas at the transmitter or/and
receiver in combination with adaptive signal processing to provide smart antenna array
processing, diversity combining or spatial multiplexing in a wireless system (Dahlman, et
al., 2007; Salwa, et al., 2007). Previously, in conventional single antenna systems the
exploited dimensions are only time and frequency whereas multiple antenna systems
exploit an additional spatial dimension. The utilization of spatial dimension with multiple
antenna techniques fulfils the requirements of LTE; improved coverage (possibility for
larger cells), improved system capacity (more user/cell), QoS and targeted date rates are
attained by using multiple antenna techniques as described in (3 GPP, 2008). Multiple
antenna techniques are an integrated part of LTE specifications because some requirements
such as user peak data rates cannot be achieved without the utilization of multiple antenna
techniques.
The radio link is influenced by the multipath fading phenomena due to constructive and
destructive interferences at the receiver. By applying multiple antennas at the transmitter or
at the receiver, multiple radio paths are established between each transmitting and receiving
antenna. In this way dissimilar paths will experience uncorrelated fading. To have
uncorrelated fading paths, the relative location of antennas in the multiple antenna
configurations should be distant from each other. Alternatively, for correlated fading
(instantaneous fading) antenna arrays should be closely separated. Whether uncorrelated
fading or correlated fading is required depends on what is to be attained with the multiple
antenna configurations (diversity, beamforming, or spatial multiplexing) (Dahlman, et al.,
2007). Generally, multiple antenna techniques can be divided into three categories (schemes)
depending on their benefits: spatial diversity, beamforming and spatial multiplexing which
will be discussed further in the following sections.
5.1 Spatial diversity
Conventionally, multiple antennas are exercised to achieve increased diversity to encounter
the effects of instantaneous fading on the signal propagating through the multipath channel.
The basic principle behind spatial diversity is that each transmitter and receiver antenna
pair establishes a single path from the transmitter to the receiver to provide multiple copies
of the transmitted signal to obtain an improved BER performance (Zheng, et al., 2003). In
order to achieve large gains with multiple antennas there should be low fading correlation
between the transmitting and the receiving antennas. Low value of correlation can be
achieved when inter-antenna spacing is kept large. Hence it is difficult to place multiple
antennas on a mobile device due size restrictions depending upon the operating carrier
frequency. An alternative solution is to use antenna arrays with cross polarizations, i.e.,
antenna arrays with orthogonal polarizations. The number of uncorrelated branches (paths)
available at the transmitter or at the receiver refers to the diversity order and the increase in

332
diversity order exponentially decreases with the probability of losing the signal. To achieve
spatial diversity for the enhancement of converge or link robustness multiple antennas can
be used either at the transmitter side or at the receiver side. We will discuss both transmit
diversity where multiple antennas are used at the transmitter (MISO-multiple-input signal-
output), and receive diversity using multiple receive antenna (SIMO signal-input multiple-
output). On the other hand, MIMO channel provides diversity as well as additional degree
of freedom for communication.
5.2 Transmit diversity
The transmit diversity scheme relies on the use of
t
N 2 antennas at the transmitter side in
combination with pre-coding in order to achieve spatial diversity when transmitting a single
data stream (Furht, et al., 2009; Jankiraman, 2004). Usually transmit diversity necessitates
the absolute channel information at the transmitter but it becomes feasible to implement
transmit diversity without the knowledge of the channel with space-time block coding
(Jankiraman, 2004). The simplest transmit diversity technique is Alamouti space-time coding
(STC) scheme (Alamouti, 1998). Transmit diversity configuration is illustrated in Figure 3.
The use of transmit diversity is common in the downlink of cellular systems because it is
easier and cheaper to install multiple antennas at base station than to put multiple antennas
on every handheld device. In transmit diversity to combat instantaneous fading and to
achieve considerable gain in instantaneous SNR, the receiver is being provided with
multiple copies of the transmitted signal. Hence transmit diversity is applied to achieve
extended converge and better link quality when the users experience hostile channel
conditions.
In LTE, transmit diversity is defined only for 2 and 4 transmit antennas and these antennas
usually need to be uncorrelated to take full advantage of the diversity gain.
LTE physical layer supports both open loop and closed loop diversity schemes. In open loop
scheme channel state information (CSI) is not required at the transmitter, consequently
multiple antennas cannot provide beamforming and only diversity gain can be achieved. On
the other hand, closed loop scheme does not entail channel state information (CSI) at the
transmitter and it provides both spatial diversity and beamforming as well.
By employing cyclic delay diversity and space frequency block coding, open loop transmit
diversity can be accomplished in LTE. In addition, LTE also implements close loop transmit
diversity schemes such as beamforming.

R
x

T
X

Fig. 3. Transmit diversity configuration

333
5.3 Space-Frequency Block Coding (SFBC)
In LTE, transmit diversity is implemented by using Space-Frequency Block Coding (SFBC).
SFBC is a frequency domain adaptation of the renowned Space-Time Block Coding (STBC)
where encoding is done in antenna/frequency domains rather than in antenna/time
domains. STBC is also recognized as Alamouti coding (Rahman, et al.). Thus, SFBC is
merely appropriate to OFDM and other frequency domain based transmission schemes.
The advantage of SFBC over STBC is that in SFBC coding is done across the subcarriers
within the interval of OFDM symbol while STBC applies coding across the number of
OFDM symbols equivalent to number of transmit antennas (Rahman, et al.). The
implementation of STBC is not clear-cut in LTE as it operates on the pairs of adjacent
symbols in time domain while in LTE the number of available OFDM symbols in a sub-
frame is often odd. The operation of SFBC is carried out on pair of complex valued
modulation symbols. Hence, each pair of modulation symbols are mapped directly to
OFDM subcarriers of first antenna while mapping of each pair of symbols to corresponding
subcarriers of second antenna are reversely ordered, complex conjugated and signed
reversed as shown in Figure 4.
For appropriate reception, mobile unit should be notified about SFBC transmission and
linear operation has to be applied to the received signal. The dissimilarity between CDD and
SFBC lies in how pairs of symbols are mapped to the second antenna. Contrarily to CDD,
SFBC grants diversity on modulation symbol level while CDD must rely on channel coding
in combination with frequency domain interleaving to provide diversity in the case of
OFDM.

S
0

Space
Frequency (OFDM subcarriers)
- S
1
*

S
0

*
-S
n+1
*

S
n
*
S
1

S
n
*
S
n+1
Frequency domain OFDM symbol
OFDM
modulation
OFDM
modulation
R
X

Fig. 4. Space-Frequency Block Coding SFBC assuming two antennas
The symbols transmitted from two transmitted antennas on every pair of neighboring
subcarriers are characterized in (Sesia, et al., 2009) as:

(0) 1
(0) 1
x (1) x (1)
X
x (2) x ( 2)

=

(3)
where
(P)
X (K) denotes the symbols transmitted from antenna port p on the k
th
subcarrier.
The received symbol can be expressed as:
y Hs n = + (4)

0 00 01 0 0
1 11 10 1 1
y h h S n
y h h S n

= +

(5)

334
where
ij
h is the channel response for symbol i transmitted from antenna j, and n is the
additive white Gaussian noise.
6. Performance comparison of channel estimation schemes
We simulate LTE down link using the SISO system with the parameters given in the
specifications (3GPP, 2009). The system bandwidth selected is 15 MHz with the numbers
of subcarriers 1536 out of which 900 subcarriers are used and the remaining are zero
padded. The sub frame duration is 0.5 ms which leads to a frame length of 1 sec. This
corresponds to a sampling frequency of 23.04 MHz or sampling interval of 43.4 ns. A
cyclic prefix of length 127 (selected from specification which is extended CP) is inserted
among data subcarriers to render the effects of multipath channel which completely
removes inter-symbol-interference (ISI) and inter-carrier-interference (ICI). In simulating
the SISO system, only one port of an antenna is considered and this antenna port is
treated as a physical antenna. We consider one OFDM symbol of size 900 subcarriers and
the reference symbols which are (total numbers of reference symbols are 150) distributed
among data subcarriers according to specifications (3GPP, 2009) transmitted from the
antenna during one time slot. The constellation mappings employed in our work are
QPSK, 16 QAM and 64 QAM.
The channel models used in the simulation are ITU channel models (ITU-R, 1997). At the
receiver end we used regularized LS and LMMSE estimation methods for the channel
estimation. All channel taps are considered independent with equal energy distribution. In
addition, frequency domain linear equalization is carried out on the received data symbols.
The performance of the system is evaluated by calculating the bit error rates using ITU
channel models with different modulation schemes.
The designed simulator is flexible to use. A scalable bandwidth is used, i.e., there is option
for using bandwidths of 5 MHz, 10 MHz, 15 MHz and 20 MHz. In addition, cyclic prefixes
of different lengths specified in (3GPP, 2009) can be easily selected in the simulation of the
system. We used single port of antenna which is taken as physical antenna however changes
can be easily made to include two ports antenna.
The performance of LTE transceiver is shown in terms of curves representing BER against
SNR values and is compared with AWGN for different channel models. Figures 5 and 6
show BER versus SNR for LMMSE and LS channel estimations, respectively, for different
ITU channel models using QPSK modulation. From these figures, it can be seen that
LMMSE channel estimation gives better performance than LS channel estimation. Figures 7
and 8 show BER plots for ITU channel models using 16QAM modulation format. It is seen
that by increasing the modulation order, the system performance degrades as compared to
QPSK modulation. This is due to the fact that higher modulations schemes are more
sensitive to channel estimation errors and delay spreads. For 16QAM, LMMSE still have
superior performance as compared to LS estimation but its performance also diminishes in
environments with high mobile speeds (Doppler spread) and large delay spreads. The LS
estimation gives poor performance for higher modulation schemes. Some interpolation
techniques can be employed to mitigate ISI effects which can enhance system performance.
Figure 9 illustrates the performance of transceiver for ITU vehicular-A channel model using
multiple antennas. The SISO system is also shown for comparision purposes.

335
0 2 4 6 8 10 12 14 16 18 20
10
-4
10
-3
10
-2
10
-1
10
0
SNR (dB)
B
E
R
BER vs SNR for ITU channel models using 4-QAM modulation

pedA_Z MMSE estimates
vehA120_Z MMSE estimates
VA350_Z MMSE estimates
AWGN

Fig. 5. BER performance of LTE transceiver for different channels using QPSK modulation
and LMMSE channel estimation

0 2 4 6 8 10 12 14 16 18 20
10
-4
10
-3
10
-2
10
-1
10
0
SNR (dB)
B
E
R

pedA_Z LS estimates
vehA120_Z LS estimates
VA350_Z LS estimates
AWGN

Fig. 6. BER performance of LTE transceiver for different channel models using QPSK
modulation and LS channel estimation

336
0 5 10 15 20 25 30
10
-4
10
-3
10
-2
10
-1
10
0
SNR (dB)
B
E
R

AWGN
pedA_Z LMMSE estimates
vehA_Z LMMSE estimates
VA350_Z LMMSE estimates

Fig. 7. BER performance of LTE transceiver for different channel models using 16 QAM
modulation and LMMSE channel estimation

0 5 10 15 20 25 30
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
SNR (dB)
B
E
R

AWGN
PedA_Z, LMMSE_estimates
VA120_Z, LMMSE_estimates
VA350_Z, LMMSE_estimates

Fig. 8. BER performance of LTE transceiver with multiple antennas for different ITU channel
models using 16 QAM modulation and LS channel estimation

337
0 2 4 6 8 10 12 14 16 18 20
10
-4
10
-3
10
-2
10
-1
10
0
SNR (dB)
B
E
R
BER vs SNR for ITU Vehicular-A channel using 4-QAM modulation

2x2_vehA_Z LMMSE estimates

Fig. 9. BER performance of LTE transceiver with multiple antennas for ITU Vehicular-A
channel model using 4-QAM modulation and LMMSE channel estimation
7. Conclusions
This chapter illustrates the physical layer aspects of future generation mobile
communication systems. Proper knowledge of propagation impairments and channel
models is necessary for the design and performance assessment of advanced transceiver
techniques employed to establish reliable communication links in future generation mobile
communication systems.
The results have been presented by means of simulations. The performance is evaluated in
terms of BER and SER and the obtained results are compared with theoretical values. The LS
estimator is simple and suitable for high SNR values; however its performance degrades
with higher constellation mappings for high mobile speeds. On the other hand, LMMSE
estimator is computationally complex and requires a priori knowledge of noise variance but
its performance is superior to LS estimates for higher modulation schemes and large delay
spreads. The performance of future generation mobile communication systems will be
highly dependent on different factors including operating frequency, elevation angles,
geographic location, climate etc.
8. References
3GPP (2008). TR 25.913:Requirements for Evolved UTRA (E-UTRA) and Evolved UTRAN
(E-UTRAN). Release 8, V 8.0.0.0.
3 GPP (2008). Overview of 3GPP Release 8: Summery of all Release 8 Features, V0.0.3.
3GPP (2009). Physical Channels and Modulation. TR 36.211 V8.7.0, Release 8.
Alamouti, S. M. (1998). A Simple Transmit Diversity Technique for Wireless
Communication. IEEE Journal Select. Areas Communications, pp. 1451-1458.

338
Correia, L., M. (2001). Wireless Flexible Personalized Communications (COST 259 Report),
John Wiley & Sons, Chichester, UK.
Cavers, J. K. (2002). Mobile Channel Characteristics. Kluwer Academic Publishers, New
York, Boston, Dordrecht, London, Moscow.
Dahlman, E., Parkvall, S., Skld, J., & Beming, P. (2007). 3G Evolution: HSPA and LTE for
Mobile Broadband, Elsevier Ltd.
Ericsson, Nokia, Motorola, and Rohde & Schwarz. (2007). R4-070572: Proposal for LTE
Channel Models. www. gpp.org, 3GPP TSG RAN WG4, meeting 43, Kobe, Japan.
Fleury, B. H. (1996). An Uncertainty Relation for WSS Processes and Its Application to
WSSUS Systems. IEEE Transactions on Communications, 44(12):16321634.
Furht, B., & Ahson, S. A. (2009). Long Term Evolution: 3GPP LTE radio and cellular
technology, published by Taylor & Francis Group, LLC.
Holma, H., & Toskala, A. (2009). LTE for UMTS: OFDMA and SC-FDMA Base Band Radio
Access. John Wiley & ISBN 9780470994016 (H/B) John Wiley & Sons Ltd.
Ibnkahla, M. (2005). Signal Processing for Mobile Communications. CRC Press, New York
Washington, D.C.
ITU-R (1997). M.1225. International Telecommunication Union: Guidelines for evaluation of
radio transmission technologies for IMT-2000.
Jankiraman, M. (2004). Space-Time Codes and MIMO Systems. Artech House Boston,
London.
Kliazovich1, D., Granelli, F., Redana, S., & Riato, N. (2007). Cross-Layer Error Control
Optimization in 3G LTE, IEEE Global Telecommunication Conference, Trento.
Meinil, J., Jms, T., Kysti, P., Laselva, D., El-Sallabi, H., Salo, J., Schneider, C., & Baum, D.
(2004). IST-2003- 507581 WINNER: Determination of Propagation, IST-2003- 507581
WINNER.
Molisch, A., F., Asplund, H., Heddergott, R., Steinbauer, M., & Zwick, T. (2006). The COST
259 vol.5, no. directional- channel model A-I: overview and methodology. IEEE
Transactions on Wireless Communications, 12, pp. 3421-3433.
Motrola. (2008). Inter-technology Mobility: Enabling Mobility between LTE and other
Access Technologies.
Oestges, C., & Clercks, B. (2008). MIMO Wireless Communications: From Real-World
Propagation to Space-Time Code Design, Elsevier Publishers.
Ojanpera, T., & Prasad, R. (2001). WCDMA: Towards IP Mobility and Mobile Internet.
Artech House Publishers, Boston, London.
Rahman, M. I., Marchetti, N., Das, S. S., Fitzek, F., & Prasad, R. Combining Orthogonal
Space-Frequency Block Coding and Spatial Multiplexing in MIMO-OFDM System.
for TeleInFrastruktur (CTiF), Aalborg University, Denmark.
Rappaport, T. (1996). Communications, Principles and Practice. Prentice-Hall, Englewood
Cliffs, NJ, USA.
Salwa, A. A., Thiagarajah, S. (2007). A Review on MIMO Antennas Employing Diversity
Techniques, Proceedings of the International Conference on Electrical Engineering
and Informatics Institute Technology Bandung, Indonesia.
Sesia, S., Toufik, I., & Bakkar, M. (2009). LTE- The UMTS Long Term Evolution, John Wiley
and Sons, Ltd, First Edition.
Zheng, L., & Tse, D. N. C. (2003). Diversity and Multiplexing: A Fundamental Tradeoff in
Multiple-Antenna Channels. IEEE Transactions on Information Theory.
19
Verifying 3G License Coverage Requirements
Claes Beckman
Center for RF-Measurement Technology, University of Gvle, and
Center for Wireless Systems, Wireless@KTH, Royal Institute of Technology,
Sweden
1. Introduction
In the beginning of the 21st century, the 3
rd
generation mobile phone systems, 3G, were
introduced all around the world. In most countries, spectrum for this technology was allocated
through some kind of licensing procedure. In Europe, the prevailing approach was to allocate
spectrum through auctions, a process which led to a situation where the European operators
found themselves committed to pay a staggering 130Bilion Euros for their 3G licenses.
However, in most European countries, the fee was not the only obligation put on the
licensee: A coverage, roll-out requirement was in many cases also connected to the license
(Northstream, 2002). Typically, these coverage requirements required that the licensees
cover a certain area at a certain point in time after that the licenses had been awarded.
In order for the regulators to verify that the licensees had met the coverage requirement and,
hence, complied with the regulation, a method for coverage verification was needed. Such
methods have therefore since then been developed by several European regulators (e.g. PTS
2004; ECC 2007). In this book chapter we describe some general underlying consideration for
the verification of radio coverage in UMTS systems and in particular we describe the Swedish
methodology developed by the Swedish Telecom regulator Post & Telestyrelsen (PTS).
2. Licensing of 3G in Sweden
In 2001, the Swedish Telecom regulator Post & Telestyrelsen (PTS) granted four licenses for
the operation of third generation mobile phone systems (PTS 2001). In contrast to most other
European countries, the Swedish licenses were granted through a beauty contest. When
acquiring the licenses, the licensees committed themselves to build networks that covered a
population of 8.860.000 inhabitants. This requirement implied that each operator would
cover some 99.98% of the Swedish population (as counted for in 1996). However, in order to
support the roll out, the regulator allowed the operators to build their networks in a
combination of self owned sites in the major cities (30% of population) and shared sites in
the countryside (70%) (Beckman and Smith, 2005). The roll-out of these 3G networks was
delayed several times and the coverage requirements somewhat modified, but in 2007 all
Swedish operators reported that they complied with the license requirements. Today
Sweden is unique in that more than 98% of the population and 48% of the of the national
territory (170.000 km
2
) has 3G service coverage (PTS 2008).
In contrast to many other European countries, the original Swedish 3G license defined
coverage by specifying a field strength requirement to be measured outdoors on the primary

340
common pilot channel, CPICH. The assumption was that depending on the environment
and the average building penetration pathloss, the pilot signal strength can be related to a
particular data service (rate) indoors.
In the original Swedish license requirement the operators where obliged to provide an out-
door signal strength that in the 3GPP release 99 standard of the UMTS system (3GPP 2000)
corresponded to an in-door packet switched data services, of 384 kbps in downlink and
144kbps in uplink. These requirements were then translated into a field strength for the
signal received from the base station. In the original license agreement coverage
requirement was that when measured outdoors at a height of 1.7m above ground over
5MHz, the field strength on the CPICH should be at least 58 dBV/m with an area
probability of 95% (PTS 2001).
The design of the Swedish measurement method is previously described in a number of
papers, e.g. PTS 2004; PTS 2004 II; Beckman et al 2006; Beckman et al 2008.
3. General considerations
To verify coverage one needs to develop a practical test procedure for measuring field
strength. The verification can then easily be performed e.g. in a drive test (PTS 2004; ECC
2007). However, designing such test presents a number of challenges:
- A requirement can be given for a particular field strength measured on the common
pilot channel. However, in the UMTS systems the power to be allocated to the CPICH is
not given by the standard or by the regulator
- There is no given relation between pilot power and service. In the Swedish license
requirements it was assumed that an outdoor signal strength of 58dBV/m on the
CPICH in practise relates to a downlink service indoors of 384 kbps and an uplink
service of 144kbps (PTS 2001). However, building penetration path loss varies in
different environment. Hence, field strength requirement must vary accordingly.
- A license is typically given for area and population coverage while a drive test only
measures along a linear route. In order to convert measurement data from drive testing
to a probability of coverage for a given area with a certain population, one needs a
statistical model based on population density and geography.
4. The primary Common Pilot Channel
The Universal Mobile Telephony System (UMTS) is a 3G systems specified by the Third
Generation Partnership Project organization (3GPP 2002). It has a radio interface based on a
code division multiple access scheme, cdma, and 5MHz wide radio channels. Since the radio
channel is somewhat wider than previous cdma systems it is referred to as: wideband
cdma or WCDMA.
The primary Common Pilot Channel, CPICH, is one of many codes in the WCDMA
common downlink pilot channel (Holma and Toskala 2002). It is a control channel mainly
used for handovers. It does not have a fixed power allocated to it so it is principle not
related to any service in either the up- or down-link.
4.1 Allocating power to the CPICH
In theory it is possible to allocate anything between 0% and 100% of the available power to
the CPICH. In practice the allocated power has a lower bound which can be derived as
follows (PTS 2004)

341

Fig. 1. Illustration of the downlink interference situation at the border between two cells
In order to initiate a soft handover at the border between two cells (Fig. 1.), a cells pilot
must be detected when an adjacent cells pilot is 5 dB stronger. The required E
b
/N
0
on the
primary CPICH on the downlink is approximately 10 dB (3GPP 2002). The processing gain
on the pilot is 10*log(3840/12.2) = 25 dB which means that the minimum output power for
the pilot is approximately: 5 + 10 25= -10dB (10%) compared to the total output power
from the base station. A worse case scenario is of course when the mobile is at the
intersection of 3 cells. The interference level would then of course be doubled. Allocating
between 10% and 20% of the available power in the radio channel is also often suggested in
industry literature (PTS 2004 II). However, it is in the interest of the operators not to increase
the pilot power unnecessarily since raising the pilot power will mean that less power is
available for services.
4.2 Relationship between pilot power and services
As described above, there is no given relation between pilot power and services. Still, the
regulator needs to have measurable criteria:
First of all one needs to consider what measure is most suitable. By tradition regulators uses
prefers to measure the signal strength in e.g. dBV/m. The main reason for this is that this
parameter is easy to measure in a drive test and is independent on frequency and antenna
gain. The relationship between signal strength E (as measured in dBV/m) and signal
power P (as measured in dBm) can we written as:
P = E - 20log10f - 77.219 + G, (1)
where f is the frequency given in MHz and G the antenna gain given in dBi.
Assuming that 10% of the available power is allocated to the primary CPICH and that the
building penetration path loss is known, it is now possible to estimate the pilot power

342
needed to provide the required services for different environments by calculating the link
budgets (Holma and Toskala 2002).
The base station has typically 10-20 W (40-43 dBm) output power available, while the mobile
unit has 0.125 W (21 dBm). The Noise Factor of the base station is typically ~4 dB compared
to ~7 dB for the mobile receiver. Antenna diversity is implemented at the base station for the
uplink and therefore approximately 4-5 dB lower Eb/N
0
than required in the downlink.
Still the downlink has a 10-15 dB path loss advantage over the uplink in a symmetrical
service. In case of asymmetrical load (higher bitrates in the downlink than in the uplink), the
10-15 dB advantage reduces to around 5-10 dB (assuming 384 kbits/s downlink and 144
kbit/s uplink).
Uplink coverage can be improved by introducing Tower Mounted Low Noise Amplifiers,
i.e. an amplifier directly after the antenna. The gain of this is that the feeder losses in the
uplink can be ignored (expect for a short jumper cable between the antenna and the
amplifier), and that the TMA often has a better Noise Factor (NF) than the base station (1.52
dB compared to 4-5 dB). TMA is widely used by the operators to improve coverage in rural
areas.
Mobile a terminals are used in a variety of environments, but to a large extent they are used
indoors. The signal is thus being attenuated as it has to propagate through the walls or
windows of the building where the user is located. Therefore, the link budget needs to
include a margin for the penetration loss in case service is planned for indoor users.
It is evident that a single penetration loss value will not apply to all environments. In rural
areas, people often live in small houses that have thin walls and windows in different
directions, thus giving a lower penetration loss. In Sweden single family house are mainly
constructed out of wood, while multi family and multistory buildings are normally made of
concrete.
In the Swedish example, the following guidelines for building attenuation was suggested:
1. In rural areas, single family houses (11 dB attenuation)
2. In suburban areas, single family houses and semi detached homes (11dB)
3. In urban areas, concrete houses with large separation (16dB)
4. In dense urban areas, concrete houses with small separation (20dB)
The link budget calculations are summarized in Tables 1 and 2. Calculations are done for the
four different scenarios mentioned above, with and without tower mounted low noise
amplifiers, TMA, in rural, and for packet switched uplink and downlink data rates of
144kbps and 384kbps, respectively. The input data is in accordance with the 3GPP UMTS
release 99 standard (3GPP 2001) and Holma and Toskala (2002).
In literature link budgets normally includes a margin for the log normal fading, which can
be described statistically, to arrive at a maximum path loss that can be used for radio
planning purposes. When comparing the above link budget with the license requirements, it
is important to understand that the margin for the statistical variation of the measured
signal in the outdoor environment is already
4.3 Coverage criteria
In Table 3 the main results of the link budget calculations are presented. As can be seen, in
all cases it is the up-link that limits the service performance. However, the CPICH signal
strength required in order to be able to provide the respective services varies in different
environments. The pilot signal strength requirement of 58dBV/m set out in the Swedish
license seems to be ~ 7dB too low in dense urban and ~8dB too strict in rural environments.

343
Environment Dense Urban Suburban Rural
Rural
TMA
Rural
TMA
Service UL kbit/s 144 144 144 144 144 64
Max mobile transmit
power dBm 21 21 21 21 21 21
Mobile Antenna Gain dBi 0 0 0 0 0 0
Body loss dB 0 0 0 0 0 0
EIRP dBm 21 21 21 21 21 21
Thermal Noise dBm/Hz -174 -174 -174 -174 -174 -174
Noise Figure dB 4 4 4 4 2 2
Noise Density dBm/Hz -170,0 -170,0 -170,0 -170,0 -172,0 -172,0
Noise Power dBm -104,2 -104,2 -104,2 -104,2 -106,2 -106,2
Interference Margin dB 3 3 3 1 1 1
Receiver interference
Power dBm -104,2 -104,2 -104,2 -110,0 -112,0 -112,0
noise + interference dBm -101,2 -101,2 -101,2 -103,2 -105,2 -105,2
Processing Gain dB 14,3 14,3 14,3 14,3 14,3 17,8
Required Eb/No dB 1,5 1,5 1,5 2 2 2
Receiver Sensitivity dBm -113,9 -113,9 -113,9 -115,4 -117,4 -120,9
Base station antenna gain dBi 18 18 18 18 18 18
Cable loss dB 4 4 4 4 1 1
Max Path Loss dB 148,9 148,9 148,9 150,4 155,4 158,9
Fast fading margin dB 4 4 4 4 4 4
Max Fading Path Loss dB 144,9 144,9 144,9 146,4 151,4 154,9
Average Penetration Loss dB 20 16 11 11 11 11
Max outdoor UL Path loss dB 124,9 128,9 133,9 135,4 140,4 143,9
Table 1. Uplink link budgets for different services and environment used for the calculation
of the Swedish license requieremnts
What is then a sufficient pilot strength criteria in order to determine whether an area is
covered or not with 3G? What is evident from the above link budget calculations is that the
signal strength requirements needs to be set differently for different environments. In Table
3, the modified Swedish CPICH requirements for different environments are summarized
(PTS 2004 II).

344
Environment Dense Urban Suburban Rural
Rural
TMA
Service DL kbit/s 384 384 384 384 384
Total available
Power dBm 43 43 43 43 43
Cable Loss dB 4 4 4 4 4
Antenna Gain dBi 18 18 18 18 18
Transmitter total
ERP dBm 57 57 57 57 57
Max Service Power
% 25% 25% 25% 50% 50%
Max Service ERP dBm 51,0 51,0 51,0 54,0 54,0
Thermal Noise dBm/Hz -174 -174 -174 -174 -174
NF dB 7 7 7 7 7
Noise Density dBm/Hz -167 -167 -167 -167 -167
Noise Power dBm -101,2 -101,2 -101,2 -101,2 -101,2
Processing Gain dB 10,0 10,0 10,0 10,0 10,0
Required Eb/No dB 6 6 6 6 6
Receiver
Sensitivity dBm -105,2 -105,2 -105,2 -105,2 -105,2
Base station
antenna gain dBi 18 18 18 18 18
Cable loss dB 4 4 4 4 4
Max Path Loss dB 156,1 156,1 156,1 159,1 159,1
Fast fading margin dB 4 4 4 4 4
Max Fading Path
Loss dB 152,1 152,1 152,1 155,1 155,1
Average
Penetration Loss dB 20 16 11 11 11
Max outdoor DL
Path loss dB 132,1 136,1 141,1 144,1 144,1
Table 2. Downlink link budgets for different services and environment used for the
calculation of the Swedish license requieremnts

345
Environment
Limiting
Link
Required CPICH
[dBV/m]
Modified Swedish
Requirements[dBV/m]
Dense Urban UL 65.1 58
Urban UL 61.1 58
Suburban UL 56.1 52
Rural UL 54.6 52
Rural TMA withTMA UL 49.6 50
Table 3. Summary of the Swedish link budget calculations and modified CPICH
requirements
5. Measurement set-up
The method used to verify the operators networks needs for obvious reasons to be accurate
but also well accepted. The traditional way of performing radio coverage measurements is
by conducting drive tests with a vehicle upon the roof of which antennas are mounted. The
signal is then sampled, measured and stored on equipment carried inside.

Fig. 2. Photo of the measurement car including the antenna solution with an extra disc as
ground plane, used by the Swedish regulator, PTS
5.1 Instrumentation
The measurement system needs by necessity be able to simultaneously detect several control
channels from several base station. The reason for that is that when the measurement is
performed in urban environments the receiver will detect several base stations. In sub-urban
or rural areas, it is of importance to be able to carefully measure at least two base staion
control channels during (soft) handowver.
The standard way of doing this is to perform a so called Top N measurement. The
measurement instrument then measures the scrambling codes transmitted on each detected
CPICH. In a Top N measurement, the system scans for all 512 scrambling codes and
returns the N strongest. In the Swedish measurement the top 6 scrambling codes were
detected and measure but N can typically be any number between 1 and 32.

346
Since the receiver is often in a scattering environment, much of the power is not in the direct
path but rather in the echos reflected from surrounding buildings and objects. The test
instrument therefore needs to support aggregate power measurements where the receiver
integrates over all reflected signals stronger than a certain level, (e.g. -17 below the
maximum power, PTS 2004)
In addition to this it is o course of importance to keep track of the position of each sample,
using a GPS, and store the all the acquired data. Table 4 below gives a list of the instrument
chosen for the Swedish test. In Fig. 3, the test receivers and the operator display are shown.

Instrumentation Equipment
Test receivers: 4 * Agilent E6455C (WCDMA)
GPS Trimble
Computer Shuttle XPC, P4 3.06GHz, 1GB ram, 160GB hd
Display 7 VGA
OS: Windows XP
Application software Nitro (Agilent)
Table 4. List of instrumentation used in the Swedish measurements

Fig. 3. Test receivers and the operator display are shown
5.2 Measurement antenna
For measurements of mobile radio coverage, it is important to have an omni-directional
pattern, especially if different networks with different base station locations are to be
compared.

347

Fig. 4. Photograph of the measurement antenna solution with an extra disc added on top of
the car
Whereas a dipole provides an omni-directional pattern with maximum radiation in the
horizontal plane, this is not the case for a monopole placed on a finite ground plane such as
a car roof (Kronberger et al, 1997). To minimize antenna induced errors in the measurement
one may consider an alternative ground plane.

Fig. 5. The monopole antenna during verification measurements at Allgons (today
Powerwave) compact range in Stockholm Sweden. The monopole and the disc ground
plane are mounted on the roof from a chassi of the same car as in the actualSwdish 3G
verification measurements

348
In order to improve the pattern for the Swedish test, Ribbenfjrd et al (2004) presented a
solution with a monopole on a circular dish placed on the car roof. The design was made
using simulations in CST Microwave Studio. A photograph of the antenna is seen in Fig. 4. It
is very difficult to make accurate measurement or simulation of an antenna mounted on a
complete vehicle, and the validation of the design was therefore made on the part of the roof
seen in Fig. 5. The result is shown in Fig. 6 where it is compared to the pattern of a
monopole mounted directly on the car roof.

Fig. 6. Simulated and measured azimuth pattern of a monopole antenna with a ground
plane disc on the roof at 2GHz The co-polar pattern of a monopole antenna directly on the
roof is also shown. The inserted drawing shows the CST-model of the antenna with disc on
the car roof. (from Ribbenfjrd et al 2004)
The results show that a substantial improvement in both azimuth and elevation is possible if
an additional ground plane in the form of a disc or a cone with corrugated edge is added.
The peak-to-peak gain variation in the horizontal plane is reduced from ~5 dB to ~2.5 dB.
6. Compensation for polarization mismatch
For practical reasons, regulators want to measure the vertical field component only, i.e.
using a vertically polarized antenna. However, the operators may very well use polarization
diversity on reception at the base station. In this case, the base station typically transmits on
a 45 degrees slant linear polarization. Therefore, we can expect that the field at the
measurement location will have a non-zero horizontally polarized component.

349

Fig. 7. Illustration of antenna systems for space (left) and polarization (right) diversity
reception. In order to receive uncorrelated signals with a space diversity system the distance
between the antennas (d) must typically be in the order of 10 wavelength
6.1 Cross polar discrimination
An important parameter related to polarization diversity is the cross polar discrimination or
XPD of the radio channel caused by the environment. With this we mean the power ratio of
vertically and horizontally polarized wave components incident on our receive base station
antenna when a mobile is transmitting.

Environment and source
Mobile
orientation
XPD
(dB)
Frequency
Correlation between
vertical and horizontal
field
env

Urban (Kozono et al 1984)
Vertical car
antenna
4-7 920 MHz 0.02 median
Urban (Vaughan 1990)
Sub-urban
30 on large
groundplane
7
12
463 MHz
-0.003
0.019
Urban & Sub-urban
(Turkmani 1995)
0
45
10
4.6-6.3
1790 MHz
<0.7 for 95%
<0.7 for 95%
Urban (Lotse et al 1996)
Sub-urban
70 15 deg in- an
outdoor
1-4
2-7
1821 MHz
<0.2 for 90%
<0.1 for 90%
Urban & Sub-urban
(Eggers et al 1983)
0
45
4-7
0
1848 MHz
<0.5 for 93%
<0.5 for 93%
Urban (Eggers et al, 1998)
Car mounted
monopole
7.62.1 970 MHz 0.090.09
Urban & Sub-urban
(Lempiainen and Laiho-
Steffens, 1998)
Random, in- and
outdoor
<5 1739 MHz -0.25 to 0.24
Sub-urban (Wahlberg et
al, 1997)
Vertical mobile
LOS & NLOS
7 1800 MHz < 0.2 average
Urban & Sub-urban (Joyce
et al, 1999)
Vertical dipole 8-11
900 & 1800
MHz
< 0.7 all values
Urban (Wahlberg et al,
1997) Sub-urban
Vertical
monopole
5
10
1800 MHz -
Table 5. References from measurements of cross polar discrimination in different environments

350
In Table 5, a number of known references from measurements of cross polar discrimination
in different environments are summarized. For example, we find that for an sub-urban
environment we typically have XPD = 6dB, between 2 and 5dB in urban environments, and
about 5 to 10dB in sub-urban.
6.2 Polarization mismatch
For 3G network verification purposes, we are interested in the amount of power lost if we
transmit on slant 45degrees antenna and receive using a vertically polarized antenna only,
i.e. the polarization mismatch loss. By virtue of reciprocity, we can use the results for a
vertically polarized mobile station transmitting to determine this power. A relevant
measurement is then Turkmani et al (1995) where a value of XPD=10dB is reported for a
vertically polarized monopole in a suburban environment.
Since this environment discriminates the horizontal polarization a concern in using a 45
degrees slanted polarized antenna for transmission is the increased path loss compared to a
vertically polarized. Due to reciprocity the transmission analysis is the same as for the
reception, and the maximum average path loss using a slanted 45 degrees antenna vs. a
vertically polarized antenna is limited to 3 dB if XPD is infinite.
What is then the difference between transmitting on slanted 45 degrees ( P
45
D
) antenna
compared to vertical (
vp
P ) in a typical sub-urban environment? This can be calculated if one
knows the XPD as:

vp
P
P ( ) = +
45
1
1
XPD
2
D
(2)
If we use the example mentioned in Turkmani et al (1995) where a value of XPD = 10dB is
reported, we find that the mean power difference between transmitting on 45 degrees slant
compared antenna compared to a vertical is: ~2.60dB
6.3 Compensation for the polarization mismatch
So, how large should this compensation be? The theory above only shows the expected
differences in measured field strength when transmitting on 45 degrees slant polarization
compared to vertical in a radio channel which discriminates the horizontal polarization.
However, it could also be argued that in some environments the propagation of the
horizontal polarization is more favourable than vertical [10]. When that is true then no
compensation should really be applied at all.
In conclusion we then find that the polarization mismatch adds an error to the measurement
spanning from 0 to 3dB depending on environment. Hence, a general compensation factor
of 3dB would be most fair to the operators.
7. Statistical model
The purpose of the test method is to establish whether or not an operator has fulfilled the
coverage requirements set out in the license [3]. The method must ensure that the license
requirement is fulfilled with sufficient statistical significance and that the sampled data is
uncorrelated.
The way to ensure that the data is uncorrelated is to make sure it is received over a large
enough geographical area. This is done by dividing the geographical area to be verified

351
for 3G coverage into test squares. If 95% of the tested squares have a sampled field
strength exceeding the one set out in the license requirement, the area is considered to be
covered.
7.1 Test squares
The size of the test squares is dependent on the environment and population (Table 6 and
Figure 1). The denser the population the smaller the size. In the evaluation only one sample
for each test square is used.

Environment Population (per square) Size (m)
Rural 0 <= x < 20 500
Suburban 20<= x < 80 250
Urban 80 <= x < 200 125
Dense Urban x >= 200 50
Table 6. Size and population of test squares

50
m
125
m
250
m
500
m
Populated
Scarcely - /un -
populated
Scarcely - /un -
populated
Scarcely - /un -
populated
50m
125m

250m

500m

Populated

Scarcely - /un -
populated

Fig. 8. Size relation between test squares
For practical reasons the measurements are conducted in a drive test using a car. Hence the
sampling rate is dependent on the speed of the car and the size of the test squares. Each
tested square is allocated the value of 1 if the license requirement is fulfilled and 0 if
not. At least 500 test squares are measured in order to assume that the number of 1 are
binomially distributed (n; p) where p >= 0.95 in the event that the license requirements are
fulfilled.

352

Fig. 9. Graphic illustration of the population density in the Stockholm area
8. Results from the Swedish measurement campaign
In 2007 all Swedish 3G licensees reported that they had fulfilled the modified (see Table 3)
coverage requirements. In order to verify these claims the Swedish regulator PTS
subsequently conducted some initial and preliminary tests.

Fig. 10. Graphic illustration of coverage in the Fagersta region at the 52dBV/m CPICH
level. Green Squares indicate Test squares passed, yellow are at the boarder line, and red
square are failed

353
8.1 Suburban environment: test case Fagersta
The first test case was conducted in a typical Swedish suburban environment in an area of
and around the city of Fagersta. The field strength requirement was set to 52dBV/m. In
total 535 test squares were measured and in order to pass the test not more than 39 were
allowed to fail for the operator to comply with the license requirement.
As shown in Table IV, the result from the measurements show that the operator passes the
test easily. Even if the CPICH field strength requirement would be increased to 53dBV/m
would the operator still pass the test indicating that the planning is fairly robust against
fading.

Field strength (dBV/m) No. Failed Squares
53 31
52 23
51 19
50 17
49 16
48 9
47 6
Table 7. Test results from Fagersta
8.2 Urban environment: test case Sundbyberg
The second test was conducted in a typical Swedish urban environment in the city of
Sundbyberg some 10km north of Stockholm. In total 602 test squares were measured and in
order to pass the test not more than 43 could fail for the operator to comply with the license
requirement. In this environment the required field strength on the CPICH is 58dBV/m.

Field strength requirement
(dBV/m)
No. Failed Squares
64 11
63 9
62 5
61 3
60 1
59 0
58 0
57 0
Table 8. Test results from Sundbyberg

354
As is evident from Table 8, the coverage planning is even more robust and the field strength
on the CPICH higher in urban areas. Even if the requirement is increased with 6dB the result
for the examined operator is still clearly above the limit of 95% area coverage.

Fig. 11. Graphic illustration of coverage in the Sundyberg region at the 58dBV/m CPICH
level. Green Squares indicate Test squares passed, yellow are at the boarder line, and red
square are failed
9. Conclusions
In the beginning of the 21st century, 3G was introduced and most countries in the western
world allocated spectrum for this technology. In Europe, the prevailing approach was to
allocate spectrum through auctions. However, in Sweden the 3G licenses were awarded
after a beauty contest, in which the winners committed themselves to cover a population of
8.886.000 which at the time corresponded to 99.98% of the countrys population. The
coverage requirements were concrete and measurable and in 2007 it was verified that all
Swedish operators complied with the requirements The development of an accepted test
method was an important part of this succesfull licensing.
10. Acknowledgment
The Author would like to thank the participants of the 3G test method working group who
all contributed in the development of the test. However, I would like to particularly
acknowledge Per Wirdemark of Canayma International AB, who has been the principle
engineer behind the design of the measurement method, Bjrn Lindmark at Laird
Technologies who was the driving force behind the antenna development and, Lars Eklund

355
and Urban Landmark at the Swedish regulator PTS, who initiated the work and contributed
to this book chapter with several of its illustrations and results.
11. References
3GPP (2002), BS radio transmission and reception (FDD) - TS 25.104 V3.10.0 (Release 1999).
http://www.3gpp.org, March 2002.
Beckman C., Lindmark B., Karlsson B., Eklund L., Ribbenfjrd D. and Wirdemark P.
Verifying 3G licence requirements when every dB is worth a bilion, European
Conference on Antennas & Propagation: EuCAP 2006
ECC Report 103 (2007). UMTS Coverage Measurements. Nice May 2007.
http://www.erodocdb.dk/Docs/doc98/official/pdf/ECCRep103.pdf
Eggers P, Kovacs I., and Olsen K. (1998) Penetration effects on XPD with GSM 1800 handset
antennas, relevant for BS polarization diversity for indoor coverage, in Proc. 48th
IEEE Veh. Technol. Conf. Ottawa, Canada, May 1998, pp. 1959-1963.
Eggers P., Toftgaard J. and Oprea A. (1983) Antenna systems for base station diversity in
urban small and micro cells, IEEE J. Select. Areas Commun., vol. 11, pp. 1046-1057.
Holma H. and Toskala A., eds. (2002), WCDMA for UMTS Radio Access for Third
Generation Mobile Communications. Chichester, New York,Weinheim, Brisbane,
Singapore, Toronto: John Wiley & Sons, Ltd, 2 ed., 2002.
Joyce R., Barker D., McCarthy M. And Feeney M., (1999) A study into the use of polarisation
diversity in a dual band 900/1800 MHz GSM network in urban and suburban
environments, IEE National Conference on Antennas and Propagation. Page(s):316
319
Kozono S., Tsuruhara T., and Sakamoto M. (1984) Base station polarization diversity
reception for mobile radio, IEEE Trans. Veh. Technol., vol. 33, pp. 301-306, Nov.
Lempiainen J. and Laiho-Steffens K. (1998) The performance of polarization diversity
schemes at a base station in small/micro cells at 1800 MHz., IEEE Trans. Veh.
Technol., vol. 3, pp. 1087-1092, Aug. 1998.
Lotse F., Berg J.-E., Forssen U., and Idahl P. (1996) Base station polarization diversity
reception in macrocellular systems at 1900 MHz, in Proc. 46th IEEE Veh. Technol.
Conf., Apr. 1996, pp. 1643-1646.
Northstream AB (2002). 3G rollout status. ISSN 1650-9862, PTSER- 2002:22, available at
http://www.pts.se.
PTS (2001) Meddelande av tillstndsvilkor fr ntkapacitet fr mobila teletjnster av
UMTS/IMT-2000 standard enligt 15 telelagen (1993:597), HK 01-7950, The
Swedish National Post and Telecom Agency, PTS March 2001
PTS (2004 II), Coverage Requirements for UMTS, The Swedish National Post and Telecom
Agency, PTS, Report Number PTS-ER-2004:32. September 2004
PTS (2004) Method fr uppfljning av tillstndsvilkoren fr UMTS-nten, The Swedish
National Post and Telecom Agency, PTS, Report Number PTS-ER-2004:23. June
2004.
PTS (2008) Dimensionering och kostnad fr utbyggnad av UMTS, The Swedish National
Post and Telecom Agency, PTS, September 2008.
R. Kronberger, H. Lindenmeier, J. Hopf, and L. Reiter, (1997). Design method for antenna
arrays on cars with electrically short elements under incorporation of the radiation
properties of the car body, in IEEE APS Symposium, Montreal, Canada, pp. 418421.

356
Ribbenfjrd D., Lindmark B., Karlsson B., and Eklund L., (2004) Omnidirectional Vehicle
Antenna for Measurementof Radio Coverage at 2 GHz, IEEE Antennas and Wireless
Propagat. Letter, VOL. 3, 269-272, 2004
Turkmani A., Arowojolu A., Jefford P., and Kellett C. (1995) An experimental evaluation of
the performance of two branch space and polarization diversity schemes at 1800
MHz, IEEE Trans. Veh. Technol., vol. 44, pp. 318-326, May 1995.
Wahlberg U., Widell S., and Beckman C. (1997) Polarization diversity antennas, in Proc.
Antenna, Nordic Antenna Symp. Gteborg, Sweden, May 1997, pp. 59-65.
Vaughan R. (1990) Polarization diversity in mobile communications, IEEE Trans. Veh.
Technol., vol. 39, pp. 177-186, Aug. 1990.
20
Inter-cell Interference Mitigation for Mobile
Communication System
Xiaodong Xu
1
, Hui Zhang
2
and Qiang Wang
1

1
Wireless Technology Innovation Institute; Key Laboratory of Universal Wireless Comm.,
Ministry of Education; Beijing University of Posts and Telecommunications,
2
Nankai University
China
1. Introduction
With the commercialization of 3G mobile communication systems, the ability to provide
diversiform data services, high mobility vehicle communication experiences and
asymmetrical services are enhanced further than 2G systems. But at the same time, users
still have higher requirement for high-rate and high-QoS mobile services. Many
international standardization organizations have launched the research and standardization
of 3G evolution system, such as 3GPP Long Term Evolution (LTE) and LTE Advanced
project. The primary three standards of 3G are all based on Code Division Multiple Access
(CDMA), but with the in-depth research of Orthogonal Frequency Division Multiplexing
(OFDM) techniques, OFDM has been emphasized by the mobile communication industry
and used as the basic multiple access technique in the Enhanced 3G (E3G) systems for its
merit of high spectrum efficiency.
OFDM becomes a key technology in the next cellular mobile communication system. As the
sub-carriers in the intra-cell are orthogonal with each other, the intra-cell interference can be
avoided efficiently. However, the inter-cell interference problems may become serious since
many co-frequency sub-carriers are reused among different cells. Under this background,
how to mitigate inter-cell interference and improve the performance for cellular users for
vehicular environments become more urgent.
In this chapter, the research outcomes about Intel-cell Interference Mitigation technologies
and corresponding performance evaluation results will be provided. The Intel-cell
Interference Mitigation strategies introduced here will include three categories, which are
interference coordination, interference prediction and interference cancellation
respectively.
2. Inter-cell interference coordination
Frequency coordination plays important roles in the Inter-cell Interference Coordination
scheme. For frequency coordination, one frequency reuse based Interference Coordination
scheme will be introduced, called as Soft Fractional Frequency Reuse (SFFR). Its frequency
reuse factor will be derived. Simulation results will be provided to show the throughputs in
cell-edge are efficiently improved compared with soft frequency reuse (SFR) scheme.

358
Especially, for Coordinated Multi-point (CoMP) transmission technology, which is the
promising technique in LTE-Advanced, a novel frequency reuse scheme Coordinated
Frequency Reuse (CFR) will be introduced, which can support coordination transmission in
CoMP system. Simulation results are also provided to show that this scheme enables to
improve the throughputs in cell-edge.
2.1 Soft fractional frequency reuse
In order to improve the performance in cell-edge, the SFFR scheme is introduced, which is
based on soft frequency reuse. As shown in Fig.1, the characteristics of such reuse schemes
are given as follows: the whole cell is divided into two parts, cell-centre and cell-edge. In
cell-centre, the frequency reuse factor (FRF) is set as 1, while in cell-edge, FRF is dynamic
and the frequency allocation is orthogonal with the edge of other cells, which can avoid
partial inter-cell interference in cell-edge.
Specially, users in each cell are divided into two major groups according to their geometry
factors. In cell-edge group, users are interference-limited due to the neighbouring cells,
whereas in cell-centre group users are mainly noise-limited. The available frequency
resources in cell-edge are divided into non-crossing subsets in SFFR.

1
u
2
u
3
u
4
u
5
u
Cell3
Cell2
Cell1
6
u
4
5
6
7
8
9
2
1
3

Fig. 1. Concept of Soft Fractional Frequency Reuse
The set of available frequency resources in the cell is allocated as follows: the whole
frequency band is divided into two disjoint sub-bands, G and F , where Gis allocated to the
cell-centre users and F to the cell-edge users. Considering a cluster of 3 cells, as the one
shown in Fig. 1, let F F F F
1 2 3
= , where
i
F denotes the subset of frequencies allocated to
cell i , i ( 1, 2, 3) = , and the subsets
i
F may be overlapped with each other.
Since the cell-edge users are easily subject to co-frequency interference, the frequency
assignments to the cell-edge users greatly rely on radio link performance and system
throughput. Generally, the cell-edge can be divided into 12 regions, as the ones marked by
1, 4, and 9 in Cell 1 (see Fig. 1). Therefore, in a cluster of 3 adjacent cells, there are 9 parts in
the cell-edge corner, which are in the shaded area. Moreover, we take this SFFR model as an
example to deduce the design of the available frequency band assignment for the fields
marked by 1, 2, .., 9.
Inter-cell Interference Mitigation for Mobile Communication System

359
In SFFR, all the available frequencies in cell-edge are divided into 6 non-overlapping
subsets. Such subsets are respectively u
1
, u
2
, u
3
, u
4
, u
5
and u
6
, while the subset in cell-
centre is u
0
. Firstly, we select frequency from the subsets u
1
, u
2
, u
3
. If its not enough,
choose frequency fromu
4
, u
5
, u
6
. If the inter-cell interference increases, we need to add
frequency into u
4
, u
5
, u
6
, and decrease the cover area in cell-edge. If such interference is
controlled in a low extension, we can decrease the frequency in subsets of u
4
, u
5
, u
6
, and
increase the cover area in cell-edge, which enables to improve the frequency utilization.
Moreover, we assume A u u u
1/3 1 2 3
{ , , } = , A u u u
2/3 4 5 6
{ , , } = and A u
3/3 0
{ } = , where A
1/3

denotes the frequency set with 1/3 reuse, A
2/3
denotes the frequency set with 2/3 reuse
and A
3/3
denotes the frequency set with FRF equals to 1.
According to the definition of FRF in references, the FRF of SFFR scheme can be obtained as
follows:

A A A
A A A
1/3 2/3 3/3
1/3 2/3 3/3
1 2 3
3 3 3
+ +
=
+ +
(1)
where the symbol stands for the cardinality of frequency set. Taking into account
that A A A A
1/3 2/3 3/3
= + + , the following relation is obtained:
A u u u
0 1 4
3 3 = + + (2)
Combining Eq.(1) and Eq.(2), the FRF is computed as:

u u u
A A A
0 1 2
1 2
3 3
3 3
= + + (3)
From Eq.(2), we can get the equation about u
1
as follows:

A u u
u
4 0
1
3
3

= (4)
Following the example of Cell 1, the number of available frequencies in cell-centre is u
0
,
whereas in the cell-edge is u u
1 4
2 + . Assuming that u k u u
0 1 4
( 2 ) = + , where k is a
constant parameter, so u
4
can be got from Eq.(4):

u u A
u
k
0 0
4
3 9 4
= (5)
Finally, taking into account Eq.(4) and Eq.(5), Eq.(3) can be expressed in terms of u
0
:

u
k A
0
1 1 5
12 3 9

= + +

(6)
It can be seen from Eq.(6) that as FRF grows, the available frequency resources in cell-centre
increase, while those in cell-edge decrease. Moreover, the performance of the SFR scheme is
compared with 3GPP LTE simulation parameters and the SFFR scheme.

360
45 50 55 60 65 70 75 80 85
18
20
22
24
26
28
30
32
34
36
number of users per cell
A
v
e
r
a
g
e

d
a
t
a

r
a
t
e

i
n

c
e
l
l
-
e
d
g
e

(
K
b
p
s
)
SFFR
SFR

Fig. 2. Comparison of average data rate in cell-edge
Fig. 2 compares the average data rate in cell-edge for SFR and SFFR, where the FRF is set as
8/9. It can be seen that the average data rate in cell-edge decreases as the number of users
per cell increases. However, the SFFR scheme outperforms the SFR scheme for a given
number of users per cell. Specially, as the increase of users, the improvement by the SFFR
scheme is more than that of the SFR scheme, which shows its more effective when the
number of users is large.
In order to mitigate inter-cell interference, a novel inter-cell interference coordination
scheme called SFFR is introduced in this part, which can effectively improve the data rate in
cell-edge. The numerical results show that compared with the SFR scheme, the SFFR scheme
improves the performance in cell-edge.
2.2 Cooperative frequency reuse
In 3GPP LTE-Advanced systems, Coordinated Multi-Point (CoMP) transmission is proposed
as a key technique to further improve the cell-edge performance in May 2008. CoMP
technique implies dynamic coordination among multiple geographically separated
transmission points, which involves two schemes.
a. Coordinated scheduling and/or beamforming, where data to a single UE is
instantaneously transmitted from one of the transmission points, and scheduling
decisions are coordinated to control.
b. Joint processing/transmission, where data to a single UE is simultaneously transmitted
from multiple transmission points.
With these CoMP schemes, especially for CoMP joint transmission scheme, efficient
frequency reuse schemes need to be designed to support joint radio resource management
among coordinate cells. However, based on the above analysis, most of the existing
frequency reuse schemes can not incorporate well with CoMP system due to not considerate
multi-cell joint transmission scenario in their frequency plan rule.
In order to support CoMP joint transmission, a novel frequency reuse scheme named
cooperative frequency reuse (CFR) will be introduced in this part. The cell-edge areas of
each cell in CFR scheme is divided into two types of zones. Moreover, a frequency plan rule

361
is defined, so as to support CoMP joint transmission among neighbouring cells with the
same frequency resources. Compared with the SFR scheme, the simulation results
demonstrate that the CFR scheme yields higher average throughput in both cell-edge and
cell-average points of view with lower blocking probability.
2.2.1 System model
A typical system model for downlink CoMP joint transmission is described in Fig. 3. In the
system, cell users are divided into two classes, namely cell-centre users (CCUs) and cell-
edge users (CEUs). We assume only CEUs can be configured to work under CoMP mode.
Each CEU has a CoMP Cooperating Set (CCS) formed by the cells that provide data
transmission service to this CEU, and the serving cell of each CUE is always included in its
CCS. The CEU with more than one cell in its CCS is regarded as a CoMP CEU, which can be
served by the cells contained in its CCS simultaneously with the same frequency resources.
It is assumed that each cell is configured with one transmitting antenna with one receiving
antenna for each user.
As shown in Fig. 3, Cell 1, Cell 2 and Cell3 are formed a CCS for user 1. So user 1 is regarded
as a CoMP CEU, and can be served by all these three cells simultaneously with the same
frequency resources. Since user 2 is not work under CoMP mode, it can only communicate
with its serving cell, i.e. Cell 1.

Cell2
Cell3
Cell1
User 2
User 1
Signal from serving cell
Signal from cooperative cell

Fig. 3. System Model for downlink CoMP joint transmission
Let
k
denote the CCS of the
th
k CoMP CEU, denote the overall cells in the system, and
{ }
k
denote the cells in set while not in set
k
. Therefore, the signal to interference
plus noise ratio (SINR) on
th
l physical resource block (PRB) for
th
k active CoMP CEU
connected to
th
i cell is determined as follows:

{ } k
k k
s l s s l
k s
i l k
n l n l n
n
P G h
N x P G
2
, ,
,
0 , ,

=
+
(7)
Where
s l
P
,
is the transmission power from
th
s cell on
th
l PRB. For simplicity,
s l
P
,
is constant
assuming no power control.
k
s
G is the long term gain between
th
s cell and the
th
k CoMP CEU,
consisting of propagation path loss and the shadow fading.
k
s l
h
,
denotes the fast fad gain on

362
th
l PRB for the channel between
th
s cell and
th
k CoMP UE. N
0
is the noise power received
within each PRB. And
n l
x
,
is the allocation indicator of
th
l PRB, which can be given by:

th th
n l
if l PRB is used in n cell
x
otherwise
,
1,
0,
(8)
In 3GPP LTE standards, it was pointed out that interference coordination is handled by the
system once every 100ms. The information reported by the users and used by the system is
the average SINR value. Thus,
k
s l
h
2
,
is replaced by its mean value
( )
k
s l
E h
2
,
1 = , and Eq. (7)
can be expressed as

{ }
k
k
k
s l s
s
k
i l k
n l n l n
n
P G
N x P G
,
,
0 , ,

=
+
(9)
For the users who dont work under CoMP mode, they only communicate with their serving
cells. The average SINR on
th
l PRB for
th
k user of
th
i cell is then given by:

k
i l i k
i l k
n l n l n
n n i
P G
N x P G
,
,
0 , ,
,

=
+

(10)
Finally, according to Shannon theorem, the corresponding capacity to the user average SINR
on
th
l PRB can be expressed as:

k
i l k
i l
C B
,
, 2
log 1

= +

(11)
Where B is the bandwidth of each PRB, and called SINR gap is a constant related to the
target BER, with ( ) BER ln 5 /1.5 = .
2.2.2 Cooperative frequency reuse scheme
The principle of the CFR scheme that can support CoMP joint transmission will be
introduced here. Each three neighbouring cells are formed as a cell cluster and respectively
marked with cell 1, cell 2 and cell 3. The cell-edge area of each cell is then divided into six
cell-edge zones according to the six different neighbouring cells. Given the marker of each
neighbouring cell, the six cell-edge zones in a cell are then categorized into two types.
Hence, there are total six types of cell-edge zones in a cell cluster. As illustrated in Fig.4,
each cell-edge zone is marked with
j
i
A , where i denotes the cell to which the zone belongs,
j is the marker of the dominant interference cell of this zone, note that { } i j , 1, 2, 3 =
and i j . For simplifying expression, we just take the cell-edge zones in cell 1 into count:
Zone A
2
1
: It is the cell-edge zone of the cells marked with cell 1. Moreover, the dominant
interferer of the users in this zone is the nearest neighbouring cell marked with cell 2.
Zone A
3
1
: It belongs to the cells marked with cell 1. And the dominant interferer is the
nearest neighbouring cell marked with cell 3.

363
3
2
A
2
1
A
3
2
A
3
2
A
1
3 A
1
3
A
3
2
A
3
2 A
3
2
A
1
2
A
1
2 A
1
2
A 2
1
A
2
1
A
3
1
A
3
1
A
3
1
A
1
3
A
Cell1
Cell 2
Cell 2
Cell 2
Cell 3 Cell 3
Cell 3

Fig. 4. Cell-edge areas partition for each cell
In order to support multi-cell joint transmission with neighbouring cells, a cooperative
frequency subset is defined for each cell in CFR scheme. Then the resources are allocated to
users in each cell cluster according to the following frequency reuse rule:
Step1. In each cell, the whole resources are divided into two sets, Gand F , where G F = .
Resources in set G are used for CCUs in each cell. While resources in set F are used for
CEUs.
Step2. Set F is further divided into three subsets, marked by F F F
1 2 3
, , , with ( )
i j
F F i j = .
Step3. For each cell cluster,
i
F is assigned for cell i as a cooperative frequency subset, which
is used for providing cooperative data transmission for the CEUs in neighbouring cells.
Step4.
j
F is assigned for the CEUs in cell-edge zones marked with
j
i
A .
Based on the above mentioned frequency reuse rule, the frequency allocation for a cell
cluster is shown in Fig. 5.

3
2
A
1
3
A
1
3
A
3
1
A
3
2
A
2
1
A
2
1
A
2
1
A
1
2 A
3
2 A
1
2
A
3
1
A
1
2
A
3
2
A
3
1
A
3
2
A
1
3 A
3
2
A
1
F
2
F
3
F
Cell 3
Cell 2
Cell1
G

Fig. 5. Frequency assignment for the boundary areas of each cell cluster
On the one hand, orthogonal frequency subsets are allocated to the adjacent cell-edge zones
that belong to different cells. Hence, the ICI can be reduced by using different frequency

364
resources in adjacent areas of neighbouring cells. On the other hand, according to the
frequency reuse rule,
j
F is allocated for cell-edge zone
j
i
A . Besides, it is the cooperative
frequency subset for cell j , which is the dominant interference cell of zone
j
i
A . Hence, for a
CoMP CEU located in zone
j
i
A , cell i and cell j can form a CCS. And then provide CoMP
joint transmission for this CEU simultaneously with the same frequency resources selected
from
j
F .

1
CF
2
CF
3
CF
CEU2
CUE3
CEU1
Cell 3
Cell2
Cell1
Signal to CEU3
Signal to CEU1
Signal to CEU2

Fig. 6. CoMP joint transmission in CFR system
As shown in Fig.6, when CEU 1 in zone A
3
1
is regarded as a CoMP CEU, its dominant
interference cell marked with cell 3 and the serving cell marked with cell 1 can form a CCS.
Then CEU 1 can be served by these two cells with the same frequency resources selected
from set F
3
. Whats more, we can see that the whole frequency resources could be reused in
all cells. Hence, the frequency reuse factor in CFR scheme can achieve to 1. In CCS selection,
we introduce an algorithm for the CCS selection. Let N denote the total number of cells in
the system, M denote the maximum number of cells in a CCS of a CEU. The
th
k CEUs
CCS, denoted as
k
, can then be selected according to the users long term gain
i
k
G as
follows:

Algorithm: CCS Selection
c
k
, count 0 .
d Calculate the long term gain
i
k
G between
th
k CEU and
th
i cell, for i N 0,..., 1 = .
{ }
N
k k k
G G G G
0 1 1
, ,...,

e Find serving cell for
th
k CEU
( )
i i
k k
i G G G argmax ,
s i
f Update
{ }
th
k k
i cell
count count 1 +

365
g If count M < ,
{ }
i
k
G G G
( )
i i
k k
i G G G argmax ,
Else stop.
h If
s i
k k
G G thr , go to f
Else stop.

It has been proved that the maximum size of UE-specific CoMP cooperating set equal to 2 is
enough to achieve CoMP gain for 3GPP case 1 in references. Hence, the value of Mis set to
2 in this paper. CEUs with two cells in their CCS are regarded as CoMP CEUs, whose SINR
can be improved by CoMP joint transmission with the same frequency resources according
to the introduced frequency reuse rule.
2.2.3 Performance analysis
System level simulations are performed to evaluate the performance of the introduced CFR
scheme. As performance metrics, we used the blocking probability and the average
throughput in both the cell-edge and cell-average points of view. The universal frequency
reuse (UFR) where PRBs are randomly assigned to the different users in each cell
irrespective of their category (CEU or CCU) is taken as a reference scheme. Another
reference scheme is SFR scheme, which assigns a fixed non-overlapping cell edge
bandwidth to a cluster of three adjacent cells. For the introduced CFR scheme, two cases are
studied, where Thr is 0 dB and 5 dB respectively.
We focus on an OFDMA-based downlink cellular system. A number of UEs are uniformly
dropped within each cell. The basic resource element considered in the system is the PRB,
which consist of 12 contiguous subcarriers. It is assumed that all the available PRBs are
transmitted with equivalent power. Only one PRB can be assigned to each active UE. The
main simulation parameters listed in Table.1 are based on 3GPP standards.

Parameters Values
Carrier Frequency 2 GHz
Bandwidth 10 MHz
Subcarrier spacing 15 kHz
Number of subcarriers 600
Number of PRBs 50
The number of cells 21
Cell radius 500m
Maximum power in BS 46 dBm
Distance-dependent path loss L=128.1+37.6log10 d (dB), d in km
Shadowing factor variance 8dB
Shadowing correlation
distance
50m
Inter cell shadow correlation 0.5

366
Fig. 7 shows the blocking probability of the introduced CFR scheme and the conventional
SFR scheme as a function of the loading factor. We can see that CFR scheme performs quite
better than the SFR scheme. Specially, the blocking probability reduced by SFR scheme is
50% more than SFR scheme. For example, if it is required that the blocking probability must
not exceed 5%, Fig.7 indicates that the admissible loading factor of the SFR scheme is only
30%, while the admissible loading factor of the introduced CFR scheme is more than 60% of
the total frequency resource. This improvement in the CFR scheme results from the
frequency reuse rule designed for each cell cluster. According to the frequency reuse rule,
the number of available frequency resources for the cell-edge areas of each cell is twice as
great as the conventional SFR scheme.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.05
0.1
0.15
0.2
0.25
Loading factor
B
l
o
c
k
i
n
g

p
r
a
b
a
b
i
l
i
t
y

SFR
CFR thr=5dB
CFR thr=0dB

Fig. 7. Blocking probability as a function of the loading factor
Fig. 8 shows the cell-edge average throughput per user for the three different frequency
reuse schemes considered in this paper. It can be seen that the average throughput per
CEU decreases as the number of users increases in all the three schemes. That is because
the probability of PRBs collision increases as the number of users grows. In other words,
the ICI increases when the average number of users per cell grows. Moreover, compared
with UFR scheme, both CFR scheme and SFR scheme yield a significant improvement in
terms of cell-edge average throughput owing to the frequency reuse plans for cell-edge
areas.
We can also observe that the introduced CFR scheme achieves higher cell-edge average
throughput than SFR scheme. When Thr is 0 dB, no user works under CoMP mode.
Compared with SFR scheme, the cell-edge average throughput is improved by 4 to 8%,
which is achieved mainly owing to the frequency reuse rule designed in CFR scheme. When
Thr is set to 5dB, the throughput raised by the introduced CFR scheme is 30 to 40% more
than the SFR scheme, that is because part of the CEUs are regarded as CoMP users whose
throughput can be further improved by CoMP joint transmission.

367
5 10 15 20 25 30
100
200
300
400
500
600
700
800
900
1000
Number of users per cell
C
e
l
l
-
e
d
g
e

a
v
e
r
a
g
e

t
h
r
o
u
g
h
p
u
t

p
e
r

u
s
e
r

[
K
b
p
s
]

UFR
SFR
CFR Thr=0dB
CFR Thr=5dB

Fig. 8. Cell-edge average throughput per user

5 10 15 20 25 30
5
10
15
20
Number of users per cell
C
e
l
l
-
a
v
e
r
a
g
e

t
h
r
o
u
g
h
p
u
t

[
M
b
p
s
]

UFR
SFR
CFR Thr=0dB
CFR Thr=5dB

Fig. 9. Cell-average throughput as a function of the number of users per cell

368
Fig. 9 shows the cell-average throughput of the introduced CFR scheme and the two
conventional frequency reuse schemes as a function of the number of users per cell. From
the graph, we can see that the cell-average throughput of the introduced CFR scheme
outperforms that of the SFR scheme due to better cell-edge performance and lower blocking
probability. When Thr is set to 0dB, the cell-average throughput is improved by 1 to 3%.
While when Thr is set to 5dB, the cell-average throughput is improved by 5 to 9%.
It is also can be seen that when the number of users per cell is small, i.e. less than 15, CFR
scheme with Thr equals to 5dB achieves the best results among the three schemes under
consideration. When the number of users is large, UFR scheme achieves the better results
than the introduced CFR scheme. However, the payoff for this higher average cell
throughput of UFR scheme is a huge decrease in cell-edge average throughput which can be
observed from Fig. 8.
2.2.4 Summary
In this part, a novel frequency reuse scheme named CFR is introduced to support CoMP
joint transmission and further improve cell-edge performance. First, the method for cell-
edge areas partition is introduced, which divides the cell-edge areas of each cell into two
types of zones. Then, the frequency plan rule is defined for each cell cluster, which assigns a
cooperative frequency subset for each cell and makes CoMP users in cell-edge zones can be
served by multi-cell joint transmission with the same frequency resources. In addition, the
algorithm is given for the CEUs to select cells in their CCS. The simulation results
demonstrate that the introduced CFR scheme significantly outperforms the conventional
SFR scheme in terms of blocking probability, cell-edge average throughput and cell-average
throughputs.
3. Inter-cell interference prediction
In order to mitigate the inter-cell interference in OFDMA systems, three schemes are given
in 3GPP organization, which respectively are interference coordination, interference
cancellation and interference randomization. However, the traditional inter-cell interference
mitigation schemes belong to passive interference suppression measures, and its
effectiveness is still limited. Considering this situation, an active interference mitigation
strategy will be introduced in this part, named as interference prediction. By means of the
immediate interference prediction in cell, it enables to efficiently avoid and eliminate inter-
cell interference, which is a novel type of active interference mitigation strategy.
For interference prediction, this part takes use of the optimal estimation theory. Generally,
the problems about optimal estimation theory can be classified into three categories: The
first is the model parameter estimation problem, such as the least squares method. The
second is time series and optimal filtering estimates problem (the optimal estimation of
signal or state). The third is the optimal information fusion estimation. According to the
actual situation in inter-cell interference prediction, the second problem of optimal
estimation is focused.
The inter-cell interference prediction principle is based on optimal estimation theory,
forecasting the co-frequency interference in the next timeslot by means of the former or
current channel state, and making the mean square error to be the smallest. The optimal
estimation theory includes time-series estimation, optimal filtering estimation method, etc.

369
Especially, the optimal filtering estimation aims to estimate the signal state, including
several filtering estimation algorithms, such as Wiener filter, Kalman filter, and so on.
3.1 Time series
In time series analysis, it aims to establish the time series model, predict and control signal
change state based on such model. Moreover, we define the observation sequence as
{ }
t n
z z z z
1 2
, , , , , the linear mixed coefficient as
{ }
t n
a a a a
1 2
, , , , , then the future
value
{ }
t k
z k 0
+
> can be predicted by means of current and past time series records
{ }
t t t
z z z
1 2
, , ,

. The predicted value is written as
t k t
z
+
, which meets following condition:

i t i t k t
i
z a z
0

+
=
=
(12)
The optimal predicted value
t k t
z
+
should make the mean square error be minimum, which
should obey

( ) t k t k t
Min E z z
2
+ +

(13)
In the above theoretical derivation, the present and past observed records
{ }
t t t
z z z
1 2
, , ,

belong to be infinite series, which is difficult to achieve in practice.
Considering this situation, the finite time series of recursive predictor are introduced, such
as Box-Jenkins method, Astrom method, etc. Besides, the steps of Box-Jenkins method are as
follows:
a. For the observed sequence
{ }
t
z t N 1, 2, , = , calculate its correlation coefficient and
partial autocorrelation coefficient, then test whether the sequence is non-stationary
white noise sequence. If such sequence is white noise series, go to the end. If such
sequence is non-stationary series, take model according to non-stationary time series
principle. Else if the sequence is stationary series, take zero for the mean of such
sequence and then make model by the Box-Jenkins method.
b. Test the type of zero mean stationary series. Illustrately, determine the series
{ }
t
z t N 1, 2, , = belong to which model, such as autoregressive (AR) model, moving
average (MA) model and autoregressive moving average (ARMA) model.
c. After the model is identified, judge the highest level of such model, and make fitted test
from low level to high level. For example, if
{ }
t
z t N 1, 2, , = belong to AR model, make
use of ( ) AR n n , 1 , and then the fitted test.
d. Compare with different models and find the right model. On this basis, respectively
take adaptive test and error test for the initial model, and select the optimal model.
e. Make prediction by the established model.
3.2 Optimal filter estimation
The nature of filtering is the statistical estimation problem. For example, linear minimum
variance estimation methods try to make the variance of estimated value and the actual
value minimum. Moreover, such filter is also known as the optimal filter, such as Wiener
filter and Kalman filter. The interference prediction process by Kalman filter is shown
respectively in Fig. 10 and Fig. 11.

370
Initial ,
Calculate interference in prior
Calculate covariance in prior
Calculate Kalman factor
Calculate the revised interference value
Calculate the revised covariance
1 k k = +
( 1/ 1) X k k ( 1/ 1) P k k

( / 1) ( 1/ 1) X k k X k k =
( / 1) ( 1/ 1)
T
P k k P k k Q = +
1
( ) ( / 1) [ ( / 1) ]
T T
K k P k k H HP k k H R

= +

( / ) ( / 1) ( )[ ( ) ( / 1)] X k k X k k K k Z k HX k k = +
( / ) [ ( ) ] ( / 1) P k k I K k H P k k =
1 k k = +
Estimate the interference
in next time

( 1/ ) ( / ) X k k X k k + =

Fig. 10. Interference prediction by Kalman filter

Fig. 11. Interference prediction results by Kalman filter
3.3 Effectiveness principles
The effectiveness of channel prediction criteria was analyzed and the relationship with the
predicted time delay and the prediction accuracy of SINR are described in the reference.
According to the wireless signal propagation, the mobility rate is closely related with the
signal coherence time. If users mobility rate increases, the coherence time becomes shorter.
0 20 40 60 80 100 120 140 160 180 200
-8
-6
-4
-2
0
2
4
6
8
10
time(s)
s
i
g
n
a
l

s
t
r
e
n
g
t
h
(
d
B
m
)
Estimated Signal (k+1)
Interference Signal

371
Else users mobility rate decreases, the coherence time becomes longer. The relationship of
coherence time and the predicted delay time are divided into three categories,
respectively t , t and t .
Fig.12 shows the prediction results when the coherence time is much greater than the time
delay, which is that t . At this time, user is in a slow moving state, and the channel
state information (CSI) can be easily obtained. From Fig.12, we can see the predicted SINR in
delay time approximates to the actual value.

SINR
Predicted value
delay t

Fig. 12. SINR prediction ( t )
Fig.13 shows the prediction results when the coherence time approaches to the time delay,
which is that t . At this time, users moving speed is in a medium state. In order to
ensure the continuity of information transmission, the SINR should obey the outage criteria
and keep a conservative prediction, which is the threshold SINR value.

SINR
conservative predicted value
delay
t


372
Fig. 14 shows the prediction results when the coherence time is far less than the time delay,
which is that t . At this time, users rate is in a high speed state, the coherence time is
shorter and the CSI is hard to be obtained. In this situation, we only need to predict the
average SINR.

SINR
Predicted value
delay
t

3.4 Summary
In this part, the inter-cell interference prediction is introduced, which is an active
interference mitigation method. The theoretical basis, which is the optimal estimation
theory, is provided with including of two parts: time series and the optimal filter estimation.
Besides, the reliability is also analyzed by means of prediction accuracy, which is based on
the relationship of the coherent time and the time delay. In addition, the trend for the actual
measured radio signals is analyzed with AR model, MA model and ARIMA model. The
analytical results are provided to show time series model can efficiently predict the radio
signals change and then mitigate the interference effectively.
4. Inter-cell interference cancellation
Inter-cell interference cancellation strategy aims at interference suppression at the user
equipments by improving the processing gain. In order to solve this problem, two basic
schemes have been discussed in 3GPP proposals. One is to take spatial suppression at the
UE side by means of multiple antennas; the other is to directly detect and subtract the inter-
cell interference in order to enable inter-cell-interference cancellation. Usually, the inter-cell
interference cancellation strategy is used to get the processing gain through suppress strong
interference. According to the degree of knowledge available about interferers, interference
cancellation methods can be distinguished as three categories, which are blind, semi-blind,
and full-knowledge.
Many inter-cell interference cancellation methods are based on generalized spatial diversity.
Beam forming is introduced in inter-cell interference cancellation in references. By
distinguish different users in space, it effectively reduces interference among users. But on

373
the other hand, it brings with extra interference from main lobe and strong side lobe. A
method of subcarrier-based virtual MIMO in inter-cell interference cancellation was
proposed, which is introduced in OFDM-based systems with a frequency reuse factor equal
to 1. But when UE is located between sectors, inter-sector interference cannot be reduced by
the subcarrier-based virtual MIMO (SV-MIMO) due to loss of channel separability. Inter-cell
interference cancellation by virtual smart antennas was also studied, which proposes a
method for estimating inter-cell symbol timing offsets using multiple signal classification
(MUSIC) algorithm. For the use of MUSIC algorithm, the premise is to know the number of
source. But in practice, the number of source can not be accurately obtained, which may
make MUSIC algorithm not work. Moreover, in most case, many similar algorithm needs to
know current channel state information, but at the same time, the complexity of system may
be increased if acquire it in downlink. As a result, how to mitigate inter-cell interference in
no precise channel is an important problem.
In order to effectively mitigate inter-cell interference in OFDM-based systems, this part
focuses on the inter-cell interference cancellation strategy. A novel inter-cell interference
mitigation method for OFDM-based cellular systems will be introduced. Compared to the
existing methods, the independent component analysis based on blind source separation is
presented in inter-cell interference, and the signal to interference plus noise (SINR) is set up
as the objective function. This scheme can adapt to the no precise channel conditions, and
can mitigate inter-cell interference in a semi-blind state of source signal and channel
information.
4.1 Inter-cell interference model
Considering the downlink in cell-edge, assume this MIMO system with q transmission
antennas in the serving eNodeB, and p receiving antennas in UE. In such scenario, UE not
only receives useful signal from current communicating base station, but also receives noise
and interference from other adjacent base stations. The example is shown in Fig. 15.

BS
eNode B1
eNode B 3
UE
eNode B 2

Fig. 15. Inter-cell interference in cell-edge
For many OFDM-based systems, the original signal is transmitted from OFDM transmitter
and through MIMO antenna array. The process of inter-cell interference mitigation is shown
in Fig.16. Further, we assume the original signal interfered by inter-cell interference and
thermal noise, and the channel information is unknown.

374

.
.
.
MIMO System
OFDM
Transmitter
.
.
.
OFDM
Transmitter
.
.
.
OFDM
Receiver
OFDM
Receiver
.
.
.
Whitening
ICA algorithm
.
.
.
y(t)=Wx(t) P/S
.
.
.
Remove
means
r1(t)
x1(t )
Whitening
Remove
means
rp(t)
xp(t)
.
.
.
Inter-cell Interference Cancellation
Inter-cell interfernce i (t)
thermal noise n(t)

Fig. 16. Inter-cell interference mitigation process
According to the principle in radio signal propagation, the thermal noise can be seen as
independent with transmission signals from eNodeB. Compared to the inter-cell
interference from other cells, we assume the useful signal is statistically independent with
the co-frequency interference from other different cells. So it can be thought that useful
signal, unknown inter-cell interference and thermal noises are statistically independent and
irrelevant with each other. Moreover, some parameters are defined as follows:
k n
s t u t i t i t i t n t
1 2
( ) { ( ), ( ),..., ( ),..., ( ), ( )}
= denotes as the source signal, which is constructed by

the useful signal, inter-cell interference and the thermal noise, also written as
T
n
s s s
1 2
( , ,..., ) .
Specifically, u t ( ) denotes as the useful signal.
k
i t ( ) denotes as the kth unknown additive
inter-cell interference with the same frequency, which is in the range of 1 to n 2 . The
dimension for the number of eNodeB reused with the co-frequency subcarriers is n 2 .
n t ( ) denotes as the additive zero mean thermal noise, also the Gaussian noise.
At the receiving end, we denote r t ( ) as the received signal, mixed with the useful signal,
unknown additive interference and noise.
k
, and are respectively the mixing vectors.
So r t ( ) can be written as the following equation:

n
k k
k
r t u t i t n t
2
1
( ) ( ) ( ) ( )
=
= + +
(14)
Furthermore, Let A denote as linear mixing matrix, which reflects temporal radio signals
transmission process and all interference from other adjacent cells are linear mixture. Then
the inter-cell interference model can be written as:
r t As t ( ) ( ) = (15)
From Eq.(15), it can been seen that the dimension of r t ( ) is the same to s t ( ) , which is equal
to n . In order to separate the useful signal from inter-cell interference and thermal noise, we
take interference cancellation by independent component analysis (ICA), which is an
important method in blind source separation (BSS). As shown in Fig.16, we deal with the
received signal by remove means and whitening, and such process is set as the transition
matrix V . So we can get:
x t Vr t ( ) ( ) = (16)

375
On the other side, we set up a separation matrix W , and make WVA I = in theory.
Moreover, assume y t ( ) is the separated signal after removing interference and noise,
so y t ( ) can be got by following equation:
y t Wx t WVAs t Gs t ( ) ( ) ( ) ( ) = = = (17)
But in fact, because exist with errors and uncertainties, we need to get an optimal
approximate solution that can make such a separation matrix W approach to the
condition G I = . By means of ICA methods, an objective function is established, which takes
W as a variable function. When W takes some value, the objective function can achieve to
minimum or maximum. At this moment, the variable W is the optimal approximate
solution.
4.2 Independent component analysis
In many fields, it needs to separate all the source signals from the mixed signals with no
precise knowledge of the source signals and the channel information, whose processes are
usually called as blind source separation (BSS). In order to solve such problem, many
schemes have been researched. When the source signals are not independent with each other
in BSS, some separating schemes are introduced, such as sparse component analysis (SCA),
smooth component analysis (SMOCA), non-negative matrix factorization (NMF), and so on.
However, the complexity of such algorithms is still high and hard to realize in application.
On the other hand, when the source signals are independent with each other in BSS, the
independent component analysis (ICA) schemes are proposed. By principle of
independence, the separating complexity is reduced and the results are also improved.
Specially, some methods exist in ICA, such as Informax, Fast ICA, generalized eigenvalue
decomposition, etc. Informax algorithm is proposed by Bell, whose characteristic is
searching for the maximum mutual information between the received signal and the output
signal, but its convergence is always slowly. Fast ICA is a fast and fixed-point algorithm
proposed by A. Hyvrinen, whose characteristic is computing the maximum kurtosis by
iterations. Although its convergence is improved compared with Informax, the effects of
thermal noise are always not included in iterations. The ICA based on generalized
eigenvalue decomposition is proposed by L. Parra, whose characteristic is decomposing
generalized eigenvalue for the received signal. However, this method is limited by the type
of source signals.
The critical step in ICA process is to make the estimated independent component gradually
approach to the source signal by means of establishing objective function and finding its
optimal solution.
According to the classical formula dealing with ICA problems, some requirements must be
made in the known conditions in order to get definite solution, as follows:
a. The source signals are all real random signals, and the respective mean is zero.
Moreover, these signals are statistically independent with each other.
b. There is at most one source signal whose probability density characteristic is the
Gaussian distribution, while the other source signals obey non-Gaussian probability
distribution.
c. For the source signals, the approximate probability distribution functions (PDF) need to
be acquired.

376
d. The number of source signals and received signals is equal, and the mixing matrix A is
a square matrix.
But in inter-cell interference cancellation, we only need to separate the useful transmission
signals from the inter-cell interference and noise, which means we dont need to separate all
source signals from the mixed signals. Specially, we can take use of some identity of the
useful signals in this situation, such as training sequence.
Based on the above analysis, and in order to verify the inter-cell interference cancellation
scheme by ICA methods, some conditions need to be illustrated as follows:
a. Considering the offset of carrier frequency and phase, it may bring with some intra-cell
interference actually. In analysis, we neglect such interference, but only consider the
inter-cell interference and thermal noise. Moreover, the inter-cell inferences from
different cells are seen as statistically independent with each other, and all these signals
are linear mixed.
b. By the principle of BSS, the means of the received signal from different eNodeB can be
respectively conversion into zero by the process of removing means, and such signals
can also be changed into real random signals by whitening, as shown in Fig.16.
c. According to the experienced formulas, the voice signals are usually with Super-
Gaussian distribution, for the kurtosis of such signals is positive. While the image
signals are usually with Sub-Gaussian distribution, for the kurtosis of such signals is
negative. Among the mixed signals, only thermal noises are with Gaussian distribution,
whose kurtosis is zero.
d. Specially, if the strength of inter-cell interference from adjacent eNodeB is much
stronger than the useful signal in cell-edge, it may reach the handover threshold and the
inter-cell handover process may be triggered. Moreover, we set the handover threshold
as -10dB in this paper, without considering ping-pong effects.
e. Because only need to separate partial signals from the mixed signals, not all the signals,
we dont require all signals received by UE are included in the mixed signals, which
means the number of the source signals and the received signals may be unequal. By
generalized matrix theory, generalized eigenvalue and generalized inverse matrix can
be computed, so it doesnt restrict that A is a square matrix.
Based on the compared analysis, ICA method is taken into inter-cell interference
cancellation for OFDM-based systems. Different from traditional ICA methods in BSS, we
only need to separate partial useful signals from the mixed signals, not all the signals.
Moreover, the training sequence can be used to identify the useful signal in order to avoid
the ambiguity of separated signal, and this process may be in a semi-blind state.
By means of the existing ICA methods, a Max-SINR ICA algorithm is constructed, which
considers the effects of both interference and thermal noise, raises the speed of convergence
and improves the output SNR.
4.2.1 Establish objective function
Considering the signal to interference plus noise ratio (SINR) is one of important measured
parameters in communication systems, we set up SINR as the objective function.
Assume the following system parameters are given when the system is set up: the source
signal s t ( ) , the received signal x t ( ) , the separated signal y t ( ) , the interference and
noise e t y t s t ( ) ( ) ( ) = , and the weighted coefficient in the ith antenna is defined as
i
, so the
SINR in user equipment can be defined as follows:

377

T
T
T
T
s t s t
F y SINR
e t e t
s t s t
y t s t y t s t
( ) ( )
( ) 10lg
( ) ( )
( ) ( )
10lg
[ ( ) ( )] [ ( ) ( )]
= =
=

(18)
Assume the separated signal in the ith antenna is
i
y t ( ) , the number of receiving antenna is
p . For s t ( ) is unknown in the conditions, we take the weighted arithmetic mean y t ( ) in place
of s t ( ) as an unbiased estimation, that is

p
i i
i
s t y t y t
p
1
1
( ) ( ) ( )
=
= =

(19)

T
T
y t y t
F y
y t y t y t y t
( ) ( )
( ) 10lg
[ ( ) ( )] [ ( ) ( )]
=

(20)
Also as y t W x t ( ) ( ) = , y t W x t ( ) ( ) = , where W is the separation matrix, so Eq. (20) can be
equal to

T T
T T
Wx t x t W
F W
W x t x t x t x t W
( ) ( )
( ) 10lg
[ ( ) ( )] [ ( ) ( )]
=

(21)
Then, take expectations onF W ( ) , which is
G W E F W ( ) { ( )} = (22)
For x t ( ) , let
T
C E x t x t { ( ) ( )} = represent the variance matrix, and let
T
C E x t x t x t x t {[ ( ) ( )] [ ( ) ( )] } =
represent the covariance matrix, so we can get

T
T
WCW
G W
WCW
( ) 10lg =

(23)
4.2.2 Optimize initial separation matrix
In order to get the maximum of G W ( ) about W , take partial derivative in Eq. (10), and let it
equate to zero, that is:

G W
W
( )
0
(24)
Through Eq. (24), we can get an equation about W . By solve this equation, the solution
W can be got. According to the proof in Theorem 1, W is the eigenvector of generalized
matrix C C
1
.
Theorem 1: The separation matrix W is the eigenvector of generalized matrixC C
1
, where
C
1
is the general inverse matrix of covariance matrixC
, and C is the variance matrix.

Proof:
By simplify Eq. (24), we can get

378

T T
CW CW
WCW WCW
2 2
=

(25)
Then combined with the known conditions
T
s
C AC A = ,
T
s
C AC A =

and WA I = , Eq. (25) can
be simplified as:
T T T
s s
WCW WAC A W C = =
T T T
s s
WCW WAC A W C = =

Put the above expressions into Eq. (25), so

s s
CW CW
C C
=

(26)

s
s
C
CW CW
C
( ) =

(27)
For the source signals
T
n
s t s s s
1 2
( ) ( , ,..., ) = , for i n 1,..., = , and j n 1,..., = , when i j ,
because
i
s and
j
s are independent and irrelevant, so
T
i j
E s s { } 0 = ; only when i j = , its nonzero,
and let
T
i i i
b E s s { } = , which means the variance. Further, set the covariance
as
T
i i i i i
b E s s s s {[ ] [ ] } =
. On this basis, the matrix constructed by variance and covariance

is respectively as follows:
T
s i n
C E s t s t diag b b b
1
{ ( ) ( )} [ ,..., ,..., ] = =
T
s i n
C E s t s t s t s t diag b b b
1
{[ ( ) ( )] [ ( ) ( )] } [ ,..., ,..., ] = =

Assume
i
i
i
b
b
=

, by matrix calculating formula, we can get
s
i n
s
C
diag
C
1
[ ,..., ,..., ] =
. Then
simplify Eq. (27), as following:

i n
C C W diag W
1
1
[ ,..., ,..., ]
(28)
As is shown in Eq. (28), W is the eigenvector of generalized matrixC C
1
. So we can get the

separation matrix W by compute the eigenvector of C C
1
. The proof is the end.

On the other side, the sizes of eigenvalues reflect different signal strength. By Eigenvalue
separation, the information about the useful signal can be found among the mixed signals,
which enables to optimize the separating process. On this basis, we need to separate partial
useful signals from the mixed signals, while remove the inter-cell interference and thermal
noise represented by some other eigenvalues.
4.2.3 Max-SINR ICA algorithm
In order to mitigate the inter-cell interference by cancellation, we take SINR as the objective
function, making it to the maximum. Based on Fast ICA algorithm in BSS, the steps of this
algorithm are introduced as follows:

379
a. Step1: Remove the mean of the original signals r t ( ) by filtering.
b. Step2: Perform the data whitening through the equation
T
x t D E r t
1/2
( ) ( )
= , where D is
the eigenvalue matrix and E is the eigenvector matrix. Moreover, D and E can be got
by the generalized eigenvalue decomposition of the matrix
T
R E r t r t { ( ) ( )} = .
c. Step3: According to the mixed signal x t ( ) , construct the variance matrix and the
covariance matrix as:
T
C E x t x t { ( ) ( )} = ,
T
C E x t x t x t x t {[ ( ) ( )] [ ( ) ( )] } =

d. Step4: Calculate the generalized eigenvalues and generalized eigenvectors by Eq. (29),
and find the eigenvalue and its eigenvector W
, which is the initial separation matrix.

s
s
C
CW CW
C
( ) =

(29)
e. Step5: Set the initial value k 0 = , and normalize W
by the equation W W W
0

/ = .
f. Step6: Judge whether
T
k k
W W 1 , in which is arbitrary small. If set up, output
k
W ;
else go to following iterations:
while k k
max

while
T
k k
W W 1 >
do
T
k k k
W E W x x W
3
1 1
{( ) } 3

=

k
T
k k k i i
i
W W W W W
1
0
( )
=
=

k k k
W W W / =
end
k k 1 = +
end
g. Step7: Separate the useful transmitting signals from the mixed signals.
y t Wx t ( ) ( ) = (30)
4.2.4 Select useful signals
Considering the inter-cell interference from different cells must be with different training
sequences, so the useful signals can be selected according to following equation:

( )
s
N
T
i i
i
d a t a t
2
0
1
( ) ( )
=
=
(31)
Where
s
N is the length of training sequence constructed by the Gold sequence,
i
a t
( ) is the
estimated value of training sequence symbol, and a t
0
( ) is the known value of training
sequence symbol.
On the other hand, the estimated training sequence of useful signal and the known training
sequence must be with the minimum distance, so it can get
{ }
n
d Min d d d
1 2
, ,..., = (32)
By means of Eq.(32), we can select the useful signal u t ( ) from the separated signals.

380
4.3 Performance evaluation
In order to verify the results of such inter-cell interference cancellation method introduced
in this paper, the static simulation is performed in the OFDM-based environment. The
simulation parameters are described in Table 2, and the process of inter-cell interference
cancellation by Max-SINR ICA is shown in Fig.16. In static simulation, we assume the
position of each UE is fixed, and neglect the Doppler frequency effect.

Parameters Values
Channel environment Rayleigh fading
Carrier Frequency 2 GHz
Bandwidth 10 MHz
Distance of sub-carrier 15 kHz
FFT size 1024
No. of subcarriers 512
No.of cells 3
Cell radius 1 km
Modulation BPSK
Convolutional codes rate 1/2
Symbols per frame 200
Length of symbol 1/20 ms
In performance evaluation, some parameters are taken as measurement, such as input SINR,
output SNR, average number of iterations, etc. Moreover, we assume the number of signals
in the received signal subspace equal to the source signal, and there are only four source
signals mixed in linear that respectively from three different eNodeB and the thermal noise.
One is the useful transmission signal, while the others are seen as the interference.
By the analysis, we compare the Max-SINR ICA algorithm with the classical Fast ICA
algorithm, and the results illustrate that the advantage of the introduced algorithm is
obvious. Specifically, the convergence of such two ICA algorithms is compared. Two
situations are considered in simulation, which respectively the length of processing frame is
fixed and the strength of thermal noise is fixed.
4.3.1 Convergence comparison
In order to compare the convergence of Max-SINR ICA algorithm with other classical ICA
algorithms, we take the Fast ICA algorithm as an example, and simulation is performed. The
result is shown in Fig. 17, where the average number of iterations is as a function of the
input SINR (
in
SINR ). The length of processing frame is set as 50 symbols, while the input
SNR (
in
SNR ) is set as 40 dB.
From Fig. 17, it can be seen that the convergence speed for the introduced Max-SINR ICA is
faster than Fast ICA. With the same value of
in
SINR , the average number of iterations for
Max-SINR ICA is less than Fast ICA. Generally, the iteration number is determined by the
search speed of the separation matrix W , which should satisfy the condition
that
T
W W 1 . Moreover, is a positive real variable and arbitrarily small.

381
For Fast ICA algorithm, the Newton iteration is taken into search for the optimized solution,
with a superlinear convergence speed. On the other side, it takes random vectors of unity
length for the initial separation matrix W
, which averagely needs more iteration to search

separation matrix compared with Max-SINR ICA.

0 5 10 15 20 25 30
2
4
6
8
10
12
14
16
18
20
22
Input signal-to-intererence-noise ratio (dB)
A
v
e
r
a
g
e

n
u
m
b
e
r

o
f

i
t
e
r
a
t
i
o
n
s
Fast ICA
Max-SINR ICA

Fig. 17. Iterations of ICA algorithms
But for Max-SINR ICA algorithm, the analytical method is taken in searching for the initial
separation matrix, which directly acquires a closed form solution by generalized eigenvalue
decomposition. Then the Newton iteration is taken, which is based on the optimized initial
separation matrix. Because there isnt an iterative process that search for the optimized
initial separation matrix, its iterative steps are less than Fast ICA algorithm. Moreover, we
can know from Theorem 1 that the initial W
is an eigenvector matrix when SINR takes the

maximum value in the objective function.
4.3.2 Fix the length of processing frame
In order to compare the performance of the introduced algorithm under different thermal
noise, we fix the length of processing frame N , and let N 50 = symbols. In such situation, we
only need to consider the effects of thermal noise in the mixed signals.
Furthermore, we give a definition about the output SNR (
out
SNR ) and the length of
processing frame N , which is as the following equation.

N
out
k
u k
SNR
N u k u k
2
2
1
1 ( )
10log
[ ( ) ( )]
=

=

(33)
In Eq. (33), u k ( ) is the kth sample of useful signal u t ( ) , while u k
( ) is the estimated value

of u k ( ) . On this basis, simulation is taken. The relationship of
out
SNR and
in
SINR is
illustrated in Fig. 17, where
out
SNR is as a function of
in
SINR with different strength of
thermal noise. Moreover, we select
in
SNR to reflect the fluctuation of such strength. From
Fig.17, it can be seen that
out
SNR with both Max-SINR ICA algorithm and Fast ICA
algorithm is robust with the increase of
in
SINR .

382
As shown in Fig.18, we make
in
SINR vary from -10dB to 30dB, and
out
SNR grows slowly with
the increase of
in
SINR . One reason is that the output SNR by ICA algorithm is affected by
the mutual information among the source signals and the probability distribution of each
signal. For such characteristics are determined, the limited change of
in
SINR plays a little
effect in
out
SNR .When
in
SNR is equal to 40dB,
out
SNR is around from 18dB to 22dB. But when
in
SNR is equal to 10dB,
out
SNR is around from 9dB to 14dB. Based on this analysis, it can be
found that by means of ICA algorithm, the higher
in
SNR is, the higher
out
SNR is.

-10 -5 0 5 10 15 20 25 30
0
5
10
15
20
25
30
O
u
t
p
u
t

s
i
g
n
a
l
-
t
o
-
n
o
i
s
e

r
a
t
i
o

(
d
B
)
Max-SINR ICA,(SNR)in=40dB
Fast ICA,(SNR)in=40dB
Region 1: performance improvement
Region 2: performance degradation
threshold

Fig. 18.
out
SNR and
in
SINR (fix the length of processing frame)
On the other side, with the same condition, such as
in
SNR is equal, the Max-SINR ICA
algorithm shows a better performance than the Fast ICA algorithm. Especially,
out
SNR improved by the Max-SINR ICA algorithm is a little more than
out
SNR improved by
the Fast ICA algorithm.
However, with the increase of
in
SINR , the increase of
out
SNR is still limited, whose growth
rate is slower than
in
SINR . As a result, when
in
SINR increases into some value, it reaches to
balance:
out in
SNR SINR = . Moreover, this state is shown as the slash through the origin in
Fig.18, which divides the graph into two regions: Region 1 and Region 2.
In Region 1,
out in
SNR SINR > , which means that the interference mitigation by ICA algorithm
is effective. But in Region 2,
out in
SNR SINR < , which means that the interference mitigation
by ICA algorithm is not only ineffective, but also degrades the performance worse as the
growth of
in
SINR .
Compared with the Fast ICA algorithm, the Max-SINR ICA algorithm raises the threshold
in
SINR of Region 1 and Region 2. It can be seen in Fig.18 that the threshold
in
SINR for the
Max-SINR algorithm is a little larger, which means if
in
SINR is in this area, the performance
is improved by the Max-SINR ICA algorithm, but degraded by the Fast ICA algorithm.
Fig. 19 shows the processing gain for such two ICA algorithms, when the length of
processing frame is fixed. It can be found that the processing gain decreases with the
increase of
in
SINR . Besides, as
in
SINR continuously increases, we can set the area with the
positive processing gain as Region 1, while the area with the negative processing gain as
Region 2. Among Region 1 and Region 2 is the threshold line.

383
-10 -5 0 5 10 15 20 25 30
-20
-15
-10
-5
0
5
10
15
20
25
30

P
r
o
c
e
s
s
i
n
g

g
a
i
n

(
d
B
)
threshold

Fig. 19. Processing gain (fix the length of processing frame)
Specially, when
in
SINR is lower than the threshold, the processing gain is positive, which
enables to improve the performance. Whats important, the lower the
in
SINR is, the higher
the processing gain is, which is useful to the users in cell-edge. But when
in
SINR is higher
than the threshold, the processing gain is negative, which degrades the performance.
Compared with the performance brought by such two algorithms, the processing gain
brought by the Max-SINR ICA algorithm is larger with the same
in
SNR . Moreover, the
introduced algorithm also raises the threshold
in
SINR . When
in
SINR is among this area, the
processing gain can be improved by the Max-SINR ICA algorithm, but degraded by the Fast
ICA algorithm.
4.3.3 Fix the strength of thermal noise
In order to measure the effects brought by the length of processing frame, we fix the
strength of thermal noise in the mixed signals, which is in a form of fixed signal to noise
ratio,
in
SNR dB 40 = . Moreover, the simulation result is shown in Fig.20, and
out
SNR is also
set as a function of
in
SINR with different lengths of the processing frame.
In static simulation, we respectively take the length of the processing frame as 50 and 100,
and the performance brought by such two ICA algorithms is compared. Further, it can be
seen that the performance can be divided into two regions:
In Region 1, the performance is improved, where
out in
SNR SINR > . With the increase
of
in
SINR , it shows that for the same ICA algorithm, the longer the length of the processing
frame is, the higher the
out
SNR is. The reason is that the independence among source signals
is easier to be established with longer processing frames. But in Region 2, the performance is
degraded, where
out in
SNR SINR < , and it is degraded worse as
in
SINR increases gradually.
Moreover, when the length of the processing frame is longer, the threshold
in
SINR between
Region 1 and Region 2 also becomes a little higher.
The reason why Region 1 and Region 2 exist in Fig. 18 and Fig. 20 is that: The output SNR by
ICA algorithm is mainly affected by the mutual information among the source signals and
the probability distribution of each signal. Once such characteristics are determined in the

384
mixed signals, the limited change of
in
SINR plays a little effect in
out
SNR . At this time, as the
growth of
in
SINR ,
out
SNR increases slowly, such curve may gradually reach to the threshold.
Before this threshold, its Region 1. Else, its Region 2.

-10 -5 0 5 10 15 20 25 30
0
5
10
15
20
25
30
O
u
t
p
u
t

s
i
g
n
a
l
-
t
o
-
n
o
i
s
e

r
a
t
i
o

(
d
B
)
Max-SINR ICA,N=100
Fast ICA,N=100
Max-SINR ICA,N=50
Fast ICA,N=50
threshold

Fig. 20.
out
SNR and
in
SINR (fix the strength of thermal noise)
Compared with the Fast ICA algorithm, both
out
SNR and the threshold
in
SINR are raised by
the Max-SINR ICA algorithm, with the same processing frame. From Fig. 20, it can be seen
that in the threshold area, the performance is improved by the Max-SINR ICA algorithm,
but degraded by the Fast ICA algorithm.

-10 -5 0 5 10 15 20 25 30
-10
-5
0
5
10
15
20
25
30
35

P
r
o
c
e
s
s
i
n
g

g
a
i
n

(
d
B
)
Max-SINR ICA,N=100
Fast ICA,N=100
Max-SINR ICA,N=50
Fast ICA,N=50
threshold

Fig. 21. Processing gain (fix the strength of thermal noise)
Fig.21 shows the processing gain for such two ICA algorithms, when the strength of thermal
noise is fixed. It can be found that the processing gain decreases with the increase of
in
SINR .
In Region 1, the processing gain is positive, and enables to improve the performance. While

385
in Region 2, the processing gain is negative, and degrades the performance. Similar to
Fig.19, it also can be found from Fig.21 that the longer the length of the processing frame is,
the higher the processing gain is.
Compared with the Fast ICA algorithm, both the processing gain and the threshold are
raised by the Max-SINR ICA algorithm with the same processing frame. The conventional
Fast ICA has forced the interference to zero, not considering the effect of the additive
thermal noise. Meanwhile, the introduced algorithm minimizes both the interference and
noise in order to maximize SINR. Thus the effect of the noise enhancement can be
suppressed by the introduced algorithm, which gives the performance improvement.
Based on the above analysis, its proper to use ICA algorithm under lower
in
SINR ,
higher
in
SNR and with longer lengths of the processing frame, which enables to mitigate the
inter-cell interference, and improve the performance. Specially, it had better employ such
inter-cell interference algorithm in practical application when the range of
in
SINR is below
10dB, but
in
SNR is above 10dB.
On the other side, it is worth noting that the effect of user mobility isnt considered because
of static simulation. Actually, when the length of processing frame is too large, such
mobility cant be tracked for the Doppler frequency effect and time varying channel. In
practice, the length of processing frame should be limited by the maximum speed of UE,
which need to be researched by dynamic simulation in the future.
4.4 Summary
In order to cancel inter-cell interference, one inter-cell interference mitigation method is
introduced, which is based on ICA algorithm. Compared to finding the maximum kurtosis
in classical ICA algorithms, such as Fast ICA, Max-SINR ICA algorithm is introduced, which
sets SINR as the objective function in this algorithm. As an important measured factor in
interference mitigation, it need try to make such function get the maximum value. By
optimize the initial separation matrix in iterations, the convergence speed of this introduced
algorithm is faster than Fast ICA algorithm. Furthermore, two situations are divided in
simulation, which respectively fix the length of processing frame and fix the strength of
thermal noise.
By means of ICA algorithm, the output SNR increases as the growth of the input SINR, but
the processing gain gradually decreases as the growth of the input SINR. Moreover, the
lower the SINR is, the higher the output SNR and the processing gain are.
On the other side, as the growth of the input SINR, there are two regions for the
performance. When the input SINR is lower than the threshold, the performance is
improved. But when the input SINR is higher than the threshold, the performance is
degraded.
Besides, the effects brought by the thermal noise and the length of the processing frame are
considered. When the input SNR is higher in the mixed signals, the output SNR is higher.
When the length of the processing frame is longer, the output SNR is also higher. Whats
more, compared with the Fast ICA algorithm, the Max-SINR algorithm raises the output
SNR and the processing gain in the same conditions.
According to the above comparison, it can be found that this inter-cell interference
cancellation method is performed well with lower SINR. So its good to improve the quality
of service for users in cell-edge where is always in the state of lower SINR. Another
advantage is that this algorithm can be performed in a semi-blind state, with no precise
knowledge of source signal and channel information. Moreover, it may not bring with extra

386
interference, which is much better than many existing inter-cell interference cancellation
algorithms.
5. Conclusion
In this chapter, the inter-cell interference mitigation for mobile communication system is
analyzed and three kinds of solutions with inter-cell interference coordination, inter-cell
interference prediction and inter-cell interference cancellation are introduced with system
models, theoretical analyses and simulation results.
For interference coordination, Soft Fractional Frequency Reuse and Coordination Frequency
Reuse schemes are introduced. Their frequency reuse factors are derived. Simulation results
are provided to show the throughputs in cell-edge are efficiently improved compared with
soft frequency reuse scheme.
The inter-cell interference prediction is an active interference mitigation method. The
theoretical basis, which is the optimal estimation theory, is provided with including of two
parts: time series and the optimal filter estimation. Besides, the steps of Box-Jenkins method
are introduced in addition. The reliability is also analyzed by means of prediction accuracy,
which is based on the relationship of the coherent time and the time delay.
For inter-cell interference cancellation, two major technologies are described in this chapter,
which are space interference suppression and interference reconstruction/subtraction
respectively. Based on the independent component analysis (ICA) technology in blind
source separation, a semi-blind interference cancellation algorithm is introduced, named as
Max-SINR ICA, which aims to improve the output SNR and optimize the initial iterative
separation matrix. Simulation results show that the iterative convergence speed for Max-
SINR ICA algorithm is faster than the traditional Fast-ICA algorithm. By the Max-SINR ICA
algorithm, the inter-cell interference can be efficiently cancelled in a semi-blind state,
especially with lower input SINR, higher input SNR and longer processing frame.
6. References
3GPP. (2005). R1-050507, Soft frequency reuse scheme for UTRAN LTE, Huawei. 3GPP TSG
RAN WG1 Meeting #41, Athens, Greece.
3GPP. (2005). R1-051396. Comparison of bit repetition and symbol repetition for inter-cell
interference mitigation.
3GPP. (2006). R1-060416. Combining inter-cell-interference co-ordination/avoidance with
cancellation in downlink and TP.
3GPP. (2006). R1-060518. TP for combining beam-forming with other inter-cell interference
mitigation approaches.
3GPP. (2006). TR 25.814 v7.1.0, Physical layer aspects for evolved UTRA (Release 7).
3GPP. (2006). TR 25.913, Requirements for Evolved UTRA (E-UTRA) and Evolved UTRAN
(E-UTRAN).
3GPP. (2007). TR 25.912. Feasibility Study for Evolved UTRA and UTRAN.
3GPP. (2008). R1-082024, A discussion on some technology components for LTE-Advanced,
Ericsson. 3GPP TSGRAN WG1 #53, Kansas City, MO, USA.
3GPP. (2008). R1-083569, Further discussion on Inter-Cell Interference Mitigation through
Limited Coordination, Samsung. 3GPP TSGRAN WG1 #54bits, Prague, Czech
Republic.

387
3GPP. (2009). R1-091688, Potential gain of DL CoMP with joint transmission, NEC Group.
3GPP TSGRAN WG1 #57, San Francisco, USA.
3GPP. (2009).TR 36.814 v1.0.1, Further Advancements for E-UTRA Physical Layer Aspects
(Release 9).
A. Hyvarienen, J. Karhunen, E. Oja. (2001). Independent Component Analysis, John Wiley
and Sons.
Haipeng Le, Lei Zhang, Xin Zhang, and Dacheng Yang. (2007).A Novel Multi-Cell OFDMA
System Structure using Fractional Frequency Reuse, In Proc. of IEEE PIMRC 2007,
pp.1-5.
Hanbyul Seo and Byeong Gi Lee. (2004). A proportional-fair power allocation scheme for
fair and efficient multiuser OFDM systems, In Proc. of IEEE Globecom 04 ,vol.6,
pp.3737-3741.
H.L. Bertoni. (2000). Radio propagation for modern wireless systems. Prentice Hall, Inc.
Huiling Jia, Zhaoyang Zhang, Guanding Yu, Peng Cheng, and Shiju Li. (2007).On the
Performance of IEEE 802.16 OFDMA System under Different Frequency Reuse and
Subcarrier Permutation Patterns, In Proc. of IEEE International conference on
communications, ICC 07, pp.5720-5725.
Hui Zhang, Xiaodong Xu, Xiaofeng Tao, Ping Zhang. (2009). An Inter-Cell Interference
Mitigation Method for OFDM-Based Cellular Systems Using Independent
Component Analysis. IEICE Transactions on Communications, Vol.E92-B, No.10.
Hui Zhang, Xiaodong Xu, Jingya Li, Xiaofeng Tao. (2009). Multicell Power Allocation
Method based on Game Theory for Inter-Cell Interference Coordination. Science in
China, Series F: Information Sciences, Vol.52, No.12, pp: 2378-2384.
Hui Zhang, Jingya Li, Xiaodong Xu, Shuang Wang, Ping Zhang. (2009). Multi-cell Subcarrier
Allocation based on Interference Forecast by Kalman Filter. Journal of Beijing
University of Posts and Telecommunications, Vol.32, No.3, pp.86-90.
Hui Zhang, Xiaodong Xu, Jingya Li, and Xiaofeng Tao.(2010) Subcarrier Resource
Optimization for Cooperated Multipoint Transmission. International Journal of
Distributed Sensor Networks, vol. 2010.
Hui Zhang, Jingya Li, Xiaodong Xu, Tommy Svensson. (2009). Channel Allocation based on
Kalman Filter Prediction in Downlink OFDMA Systems. IEEE VTC 2009-Fall.
Hyvarinen, A. (1999). Fast and robust fixed-point algorithms for independent component
analysis. IEEE Trans. on Neural Networks, vol.10, no. 3, pp. 626 - 634.
I. Kostanic, W. Mikhael. (2002).Rejection of the co-channel interference using non-coherent
independent component analysis based receiver, Proc. IEEE Midwest Symposium on
Circuits and Systems, vol. 2, Aug.2002.
I.Kostanic, W.Mikhael. (2004).Blind source separation technique for reduction of co-channel
interference. IEE Electronics Letters, Vol. 38, No. 20, pp.1210 1211.
J.G.Andrews. (2005). Interference cancellation for cellular systems: a contemporary
overview. IEEE Wireless Commun. Magazine, vol.12, no. 2, pp. 19 - 29.
J. Tomcik. (2006). Qualcomm, MBFDD and MB TDD wideband mode, IEEE 802.20-05/68r1.
K.N. Lau, K. Yu, K. Ricky.(2006). Channel adaptive technologies and cross layer designs for
wireless systems with multiple antennas theory and applications. Canada: John
Wiley & Sons, Inc., Publication, pp. 1-503.
K.I. Lee, Y.H. Ko. (2006). An inter-cell interference cancellation method for OFDM cellular
systems using a subcarrier-based virtual MIMO. Proc. IEEE VTC, pp. 1-5.

388
Ki Tae Kim, Seong Keun Oh. (2007).A Universal Frequency Reuse System in a Mobile
Cellular Environment, In Proc. of IEEE VTC 2007-Spring, pp.2855-2859.
Ki Tae Kim, Seong Keun Oh. (2008).An Incremental Frequency Reuse Scheme for an
OFDMA Cellular System and Its Performance, In Proc. of IEEE VTC 2008-Spring,
pp.1504-1508.
K.W. Park, K.I. Lee, Y.S. Cho. (2006). An inter-cell interference cancellation method for
OFDM-Based cellular systems using a virtual smart antenna. IEICE Trans. on
Commun., vol. E89-B, no.1, pp. 217-219.
L. Prarra, P. Sajda.(2003). Blind source separation via generalized eigenvalue decomposition.
Journal of Machine Learning Research, no.4, pp. 1261-1269.
M. Barkat. (2005). Signal Detection and Estimation. Artech House Publishers.
Q.H. Spencer, C.B. Peel. (2004). An introduction to the multi-user MIMO downlink. IEEE
Commun. Magazine, vol.42, pp. 60-67.
S.E. Elayoubi, O. B. Haddada, and B. Fourestie. (2008). Performance Evaluation of
Frequency Planning Schemes in OFDMA-based Networks, IEEE Trans. Wireless
Commun., vol. 7, no.5, pp. 16231633.
S.R. Curnew, J. How. (2007). Blind signal separation in MIMO OFDM systems using ICA
and fractional sampling. Proc. International Symposium on Signals, Systems and
Electronics, pp. 67-70.
T. Ristaniemi, J. Joutsensalo. (1999). Nonlinear algorithm for blind interference cancellation,
Proc. IEEE Signal Processing Workshop on Higher-Order Statistics, pp.43-47.
T. Yang. (2004). Diversity wireless receivers with efficient co-channel interference
suppression. Proc. IEEE Advances in Wired and Wireless Communication, pp.145-147.
Xu Fangmin, Tao Xiaofeng, Zhang Ping. (2009). A Frequency Reuse Scheme for OFDMA
Systems, Journal of Electronics&Information Technology,vol. 3, no.4, pp.903-906.
Xu Xiaodong, Zhang Hui, Li Jingya, Tao Xiaofeng, Zhang Ping. (2009). An Improved
Exponential Distributed Power Control Algorithm for MIMO Cellular Systems.
IEEE WiCOM 2009.
1. Introduction
The latest advancements of the 3
rd
generation (3G) universal mobile telecommunications
system(UMTS) have led to the long termevolution (LTE) standard release (referred to as 3.9G)
within the 3
rd
generation partnership project (3GPP). LTE does not meet the requirements
for the fourth generation (4G) systems dened by the international telecommunication union
(ITU). Therefore, work on LTE-Advanced within 3GPP has recently started. LTE-Advanced
can be seen as the continuous evolution of wireless service provision beyond voice calls
towards a true ubiquitous air-interface capable of supporting multimedia services (Sesia et al.,
2009).
LTE-Advanced systems face a number of essential requirements and challenges which include
coping with limited radio resources, increased user demand for higher data rates, asymmetric
trafc, interference-limited transmission, while at the same time the the energy consumption
of wireless systems should be reduced. Driven by the ever-growing demand for higher data
rates to effectively use the mobile Internet, future applications are expected to generate a
signicant amount of both downlink (DL) and uplink (UL) trafc which requires continuous
connectivity with quite diverse quality of service requirements. Given limited radio resources
and various propagation environments, voice over IP applications, such as Skype, and
self-generated multimedia content platforms, such as YouTube, and Facebook, are popular
examples that impose a major challenge on the design of LTE-Advanced wireless systems.
One of the latest studies from ABI Research, a market intelligence company specializing
in global connectivity and emerging technology, shows that in 2008 the mobile data trafc
around the world reached 1.3 Exabytes (10
18
). By 2014, the study expected the amount to reach
19.2 Exabytes. Furthermore, it has been shown that video streaming is one of the dominating
application areas which will grow signicantly (Gallen, 2009).
In order to meet such diverse requirements, especially, the ever-growing demand for mobile
data, a number of different technologies have been adopted within the LTE-Advanced
framework. These include smart antenna (SA)-based (also known as directional antennas
or antenna arrays) multiple-input multiple-output (MIMO) systems (Bauch & Dietl, 2008a;b;
Foschini & Gans, 1998; Kusume et al., 2007) and efcient multiuser transmission techniques
such as multiuser MIMO using precoding to achieve, for example, space division multiple
access (SDMA) (Fuchs, et al., 2007), and networked MIMO, i.e. coordinated multipoint (CoMP)
systems. Therefore, there is a broad agreement recently among LTE standardization groups
Rami Abu-alhiga
1
and Harald Haas
2
The University of Edinburgh
United Kingdom
Novel Co-Channel Interference Signalling
for User Scheduling in Cellular
SDMA-TDD Networks
21
that MIMO will be the key to achieve the promised data rates of 1 Gbps and more (Seidel,
2008).
It is well known that co-channel interference (CCI), caused by frequency reuse, is considered
as one of the major impairments that limits the performance of current and 4G wireless
systems (Haas & McLaughlin, 2008). To outmaneuver such obstacle, various techniques
such as joint detection, interference cancelation, and interference management have been
proposed. One of the most promising technology is to utilize the adaptability of SAs. Spatial
signal pre-processing along with SAs can provide much more efcient reuse of the available
spectrum and, hence, an improvement in the overall system capacity. This gain is achievable
by adaptively utilizing directional transmission and reception at the base station (BS) in order
to enhance coverage and mitigate CCI. One of the key challenges to overcome, however, is the
signalling overhead which increases drastically in MIMO systems.
Unlike the traditional resource allocation in single-input single-output (SISO) fading channels,
which is performed in time and frequency domains, the resources in MIMO systems are
usually allocated among the antennas (the spatial domain). From closed-loop MIMO point
of view, channel aware adaptive resource allocation has been shown to maintain higher
system capacity compared to xed resource allocation (Ali et al., 2007; Gesbert et al., 2007;
Koutsimanis & Fodor, 2008). In particular, adaptive resource allocation is becoming more
critical with scarce resources and ever-increased demand for high data rates.
It is shown that for closed-loop MIMO the optimal power allocation among multiple
transmit antennas is achieved through the water-lling algorithm (Telatar, 1999). However,
to enable optimal power allocation, perfect channel state information (CSI) at the transmitter
is required. Some other work focused on transmit beamforming and precoding with limited
feedback (Love, et al., 2005; 2003; Mukkavilli et al., 2002; 2003; Zhou et al., 2005), where the
transmitter uses a quantized CSI feedback to adjust the power and phases of the transmitted
signals. To further reduce the amount of feedback and complexity, different strategies such
as per-antenna rate (an adaptive modulation and coding approach that controls each antenna
separately) and power control algorithms have been proposed (Catreux et al., 2002; Chung
et al., 2001a;b; Zhou & Vucetic, 2004; Zhuang et al., 2003). By adapting the rate and power
for each antenna separately, the performance (error probability (Gorokhov et al., 2003) or
throughput (Gore et al., 2002; Gore & Paulraj, 2002; Molisch et al., 2001; Zhou et al., 2004))
can be improved greatly at the cost of slightly increased complexity. Additionally, antenna
selection is proposed to reduce the number of the spatial streams and the receiver complexity
as well. Various criteria for receive antenna selection or transmit antenna selection are
presented, aiming at minimizing the error probability (Bahceci et al., 2003; Ghrayeb &Duman,
2002; Gore et al., 2002; Gore & Paulraj, 2002; Heath & Paulraj, 2001; Molisch et al., 2003) or
maximizing the capacity bounds (Molisch et al., 2003; Zhou & Vucetic, 2004). It is shown that
only a small performance loss is experienced when the transmitter/receiver selects a good
subset of the available antennas based on the instantaneous CSI (Zhou et al., 2004). However,
it is found that in spatially correlated scenarios, proper transmit antenna selection cannot
just be used to decrease the number of spatial streams, but can also be used as an effective
means to achieve multiple antenna diversity (Heath & Paulraj, 2001). When the channel links
exhibit spatial correlation (due to the lack of spacing between antennas or the existence of
small angular spread), the degrees of freedom (DoF) of the channel are usually less than the
number of transmit antennas. Therefore, using transmit antenna selection, the resources are
allocated only to the uncorrelated spatial streams so that an enhanced capacity gain can be
achieved.
Most of the above work focused on the point-to-point (P2P) link in single user scenario. In a
multiuser MIMO (MU-MIMO) context, MIMO communication can offer signicant capacity
growth by exploiting spatial multiplexing and multiuser scheduling. Therefore, opportunistic
approaches have recently attracted considerable attention (Choi et al., 2006; Viswanath et al.,
2002). So far, opportunistic resource allocation in a MU-MIMO scenario is still an open
issue. Wong et al. and Dai et al. (Dai et al., 2004; Wong et al., 2003) consider a multiuser
MIMOsystemand focused on multiuser precoding and turbo space-time multiuser detection,
respectively. More recent work has addressed the issue of cross-layer resource allocation in DL
MU-MIMO systems (Wang & Murch, 2005). In broadcast MU-MIMO channels, dirty-paper
coding (DPC) (Costa, 1983) can achieve the maximum throughput (Goldsmith et al., 2003;
Vishwanath et al., 2003; Weingarten et al., 2004). In particular, DPC can accomplish this
by using successive interference precancelation through employing complex encoding and
decoding. Unfortunately, DPC is classied as a nonlinear technique that has very high
complexity and is impractical. Due to the fact that DPC is computationally expensive for
practical implementations, its contribution is primarily to determine the achievable capacity
region of MU-MIMO channel under a per-cell equal power constraint. Therefore, many
alternative practical precoding approaches are proposed to offer a trade-off complexity for
performance (Airy et al., 2006; Chae et al., 2006; Hochwald et al., 2005; Pan et al., 2004; Shen
et al., 2005; Windpassinger et al., 2004). These alternatives considered different criteria and
methods such as minimum mean squared error (MMSE) (Schubert & Boche, 2004; Shi et al.,
2008), channel decomposition, and zero forcing (ZF) (Chen et al., 2007; Choi & Murch, 2004;
Spencer et al., 2004; Wong et al., 2003).
One of the most attractive approaches is the block diagonalization (BD) algorithm which
supports orthogonal multiple spatial stream transmission. In BD algorithm, the precoding
matrix of each user is designed to lie in the null space of all remaining channels of other
in-cell users, and hence the intracell multiuser CCI is pre-eliminated (Chen et al., 2007; Shen
et al., 2005; Spencer et al., 2004). In particular, SA-based SDMA, implementing BD algorithm,
can multiplex users in the same radio frequency spectrum (i.e. same time-frequency resource)
within a cell by allocating the channel to spatially separable users. This can be done while
maintaining tolerable, almost negligible, intracell CCI enabled by BD signal pre-processing
capabilities. Moreover, channel aware adaptive SDMA scheme can be achieved through joint
exploitation of the spatial DoF represented by the excess number of SAs at the BS along with
multiuser diversity. Generally, the radio channel encountered by an array of antenna elements
is referred to as beam. In other words, SA technology along with BD algorithm can enable
the BS to adaptively steer multiple orthogonal beams to a group of spatially dispersed mobile
stations (MSs) (Choi et al., 2006), as depicted in Fig. 1.
The joint beam selection and user scheduling for orthogonal SDMA-TDD (time division
duplex) system is a key problem addressed in this chapter. From precoding point of view,
the availability of CSI of all in-cell users at the BS is crucial in multiuser (MU)-MIMO
communication scenario to optimally incorporate different precoding techniques such as BD,
adaptive beamforming, or antenna selection, in order to increase the overall system spectral
efciency. Basically, there are two methods for providing a BS with CSI of all associated MSs,
namely limited (quantized) feedback and analog feedback. Limited feedback (also know as
direct feedback) involves the MS to measure the DL channel and to transmit a feedback
messages of quantized CSI reports to the BS during the UL transmission. Alternatively,
the second method, referred to as UL channel sounding according to LTE terminology,
involves the BS to estimate the DL channel based on channel response estimates obtained
391 Novel Co-Channel Interference Signalling for User Scheduling in Cellular SDMA-TDD Networks
MS1
Users
data
streams
UL channel sounding
MS2
MS3
MS4
BD
beam-
forming
Joint beam
and user
selection
Fig. 1. A block diagram of SA-based MU-MIMO transmission implementing BD
beamforming
from reference signals (pilots) received from the MS during UL transmission. Channel
sounding offers advantages in terms of overhead, complexity, estimation reliability, and delay.
Closed-loop SDMA-TDD networks can benet from these advantages to avoid outdated
feedback scenarios, enhance the network throughput, and reduce the computational cost at
the user side. Clearly, TDD systems offer a straightforward way for the BS to acquire the
CSI enabled through channel reciprocity (Love, et al., 2004). The advantages of UL channel
sounding are discussed later in this chapter and a more detailed treatment can be found, for
instance, in the technical documents of the evolved universal terrestrial radio access (EUTRA)
study item launched in the LTE concept (Sesia et al., 2009).
In summary, UL channel sounding method is considered as one of the most promising
feedback methods for SA-based SDMA-TDD systems due to its bandwidth and delay
efciency. In particular, UL channel sounding avoids the usage of dedicated feedback physical
channels which results in utilizing the available bandwidth for data transmission much more
effectively. In addition, UL channel sounding requires a shorter duration of time to convey the
feedback information to the BS compared to the direct feedback method. This feature reduces
the probability of having outdated feedback especially in fast varying channel conditions.
In interference-limited scenarios and according to Shannon capacity formula, the system
performance is limited by the CCI from adjacent cells. Meanwhile, conventional channel
sounding (CCS) only conveys the channel state information (CSI) of each active user to the
BS. Therefore, CSI is only a suboptimal metric for multiuser spatial multiplexing optimization
in interference-limited scenarios.
In light of the above, the benchmark system considered in this chapter for the system
level analysis of the feedback methodology is a closed-loop SDMA TDD system. In the
benchmark system, a BD technique is utilized to optimize the MU-MIMO spatial resources
allocation problem based on perfect instantaneous CSI of each in-cell active user obtained
from UL channel sounding pilots. The main goal of the benchmark system is to adaptively
communicate with a group of users over disjoint spatial streams while optimizing the gains
of the MU-MIMO channels. The optimization aims at enhancing the overall system capacity
using xed and uniformly distributed transmit power.
Most of the ZF-based precoding algorithms (e.g. BD) have been designed to only mitigate
intracell CCI from different users in the same cell without considering CCI coming from
transmitters in neighbouring cells. In a cellular environment, especially when full frequency
reuse is considered, intercell (also known as other-cell) CCI becomes a key challenge which
cannot be eliminated by BD-like algorithms. Moreover, it is shown that intercell CCI can
signicantly degrade the performance of SDMAsystems (Blum, 2003). More specically, if the
BS schedules a group of users only based on the available CSI, the scheduling decision may be
optimum for a noise limited system, but high intercell CCI at the respective MSs might render
the scheduling decision greatly suboptimum. Therefore, the signal-to-interference-plus-noise
ratio (SINR) would be a more appropriate metric in multicell interference limited scenarios,
but this metric cannot directly be obtained from CCS. Thus, the key challenge here is to
provide knowledge of intercell CCI observed by each user to the BS in addition to CSI. If,
furthermore, intercell CCI observed by each SA at the BS itself is taken into account, the beam
selection and user scheduling process can be jointly improved for both DL and UL (Abualhiga
& Haas, 2008).
2. Contributions and assumptions
The contribution associated with the feedback-based interference management for SDMA
presented in this chapter can be split into three main parts:
A novel interference feedback mechanism is developed. Specically, it is proposed to
weight the UL channel sounding pilots by the level of the received intercell CCI at
each MS. The weighted uplink channel sounding pilots act as a bandwidth-efcient and
delay-efcient means for providing the BS with both CSI and intercell CCI experienced
at each active user. Such modication will compensate for the missing interference
knowledge at the BS when traditional UL channel sounding is used. In addition, through
exploitation of channel reciprocity the technique will act as implicit inter-cell interference
coordination (ICIC) avoiding any additional signalling between cells.
A novel procedure is developed to make the interference-weighted channel sounding
(IWCS) pilots usable for the scheduler to optimize the spatial resource allocation during the
UL slot. It is proposed to divide the metric obtained from the IWCS pilots by the intercell
CCI experienced at the BS. The resulting new metric, which is implicitly dependent on DL
and UL intercell CCI, provides link-protection awareness and it is used to jointly improve
spectral efciency in UL as well in DL.
Finally, in order to facilitate a practical implementation, a heuristic algorithm (HA) is
proposed to reduce the computational complexity to solve user scheduling problem.
The key assumptions for the system level analysis of the IWCS pilots performance can be
summarized as follows:
The considered closed-loop SDMAsystemenjoys perfect knowledge of the MIMOchannel
coefcients of each active user. Hence, this channel knowledge at both BS and MS is
exploited to decompose the channel matrix into a collection of uncoupled parallel SISO
channels.
The considered problem of jointly adapting the MU-MIMO link parameters for a set of at
fading co-channel interfering MIMO links exploits two DoF: transmit antenna selection,
and user selection. Since these two DoF are associated with two different layers (the
physical (PHY) layer and medium access control (MAC) layer) the problem is considered
to be optimized in a cross-layer fashion.
The time and frequency DoF (e.g. frequency channel dependent scheduling and dynamic
frequency resource allocation) are not considered in this study.
This chapter assumes that appropriate methods are in place that completely eliminate
or avoid intracell CCI. Therefore, the system is only limited by intercell CCI. However,
the level of intercell CCI usually outweighs thermal noise and the system is, therefore,
interference limited.
According to the denition of the UL channel sounding mechanism adopted by LTE,
channel sounding pilots are different from the demodulation pilots dedicated to the
process of coherent data detection. This implies that modifying the UL channel sounding
pilots does not hamper the channel estimation processes required for coherent data
detection. In particular, the only purpose of the proposed modication on the UL channel
sounding pilots is to add interference awareness to the channel sounding technique. In
addition, according to the LTE technical documents related to the UL channel sounding
pilots, the predetermined sounding waveforms are transmitted using orthogonal signals
among all active users in all cells using the same frequency band. The sounding pilot
sequences are chosen to be orthogonal in frequency domain among all of the users
antennas (Sesia et al., 2009). In summary, the above properties enable the BS to estimate the
UL wideband channel for each antenna of each active user without any intracell or intercell
CCI between the channel sounding pilots. Moreover, errors in the channel estimation due
to the presence of noise is beyond the scope of this chapter. As a consequence, perfect
channel estimation is considered as outlined in the rst assumption.
3. Overview of feedback methods
Basically, there are two methods for providing a BS with the CSI of all MSs, namely direct
feedback and UL channel sounding.
1. Direct feedback: According to LTE terminology this feedback method is termed
codebook-based feedback (Abe & Bauch, 2007). The MS determines the best entry in
a predened codebook of precoding (beamforming) vectors/matrices and transmits a
feedback indication to the BS conveying the index value. In codebook feedback, the MS
uses downlink channel estimates to determine the best codebook weight or weights for
the BS to use as a precoding vector/matrix. The MS creates a feedback indication that
includes the codebook index and then sends the feedback indication to BS. This method can
be considered as a candidate option for frequency division duplex (FDD) systems which
require an explicit transfer of the DL CSI during the UL transmission due to the absence of
channel reciprocity.
It is worth mentioning that the physical feedback channel needs to have some reference
signals to facilitate the coherent detection of the feedback information at the BS.
2. UL channel sounding: The MS transmits a sounding waveform on the UL and the BS
estimates the UL channel of the MS from the received sounding waveform. The sounding
pilot sequences are chosen to be orthogonal between all of the users antennas and also are
designed to have a low peak to average power ratio (PAPR) in the time domain, (Fragouli,
et al., 2003; Popovic, 1992). The details of UL channel sounding are given in 3GPP technical
documents (Sesia et al., 2009). However, a brief treatment of the uplink channel sounding
signal model is given below.
According to LTE technical documents related to uplink channel sounding, the BS instructs
the MS where and how to sound (i.e., send a known pilot sequence) on the uplink. The
information obtained from uplink channel sounding at the BS is used to determine DL
beamforming weights for MIMO channel dependent scheduling on the uplink, as well as
for MIMO channel dependent scheduling on the DL. According to the structure discussed
in 3GPP technical documents, channel sounding pilots enable the BS to estimate the UL
wideband channel for each antenna of each active user without any intracell or intercell
CCI between the channel sounding pilots.
Any codebook-based feedback scheme must account for the number of SAs at the BS. In a
codebook-based feedback scheme, the MS must be able to estimate the DL channel no matter
how many SAs are available at the BS. Thus, the computational complexity at the MS, and
the information that is required to be fed back increase with the number of antennas at BS.
In contrast, channel sounding schemes are independent of the number of BS antennas. In
other words, the problem of channel estimation is much more difcult in a codebook-based
feedback scheme than in a channel sounding scheme. More specically, in codebook-based
feedback, the air-interface must enable the MS to estimate the channel between its antennas
and a relatively larger number of SAs at the BS. Such estimation imposes a heavy processing
load on a MS in a direct feedback scheme, while in a channel sounding scheme the estimation
process takes place at the BS side (Hassibi & Hochwald, 2003).
For instant, consider the case where the BS has eight transmit antennas and the MS has a two
receive antennas. In a channel sounding scheme, the BS must estimate the channel between
its eight antennas and the two transmit antennas. In contrast, in a codebook-based feedback
scheme, the air interface must enable the MS to estimate the channel between its two antennas
and the eight transmit antennas (an eight-source channel estimation problem, which is much
more difcult).
In TDDsystems codebook-based feedback schemes tend to have much higher latency between
the time of the channel estimation and the time of the subsequent DL transmission. The
resulting outdated CSI can have detrimental effects on the performance of closed-loop
transmission schemes, especially in fast fading channels. In contrast, channel sounding
reference signals can be transmitted at the end of the UL slot. They can directly be exploited
for the subsequent DL transmission. For these reasons, in a TDDsystem, UL channel sounding
is preferred over codebook-based feedback.
4. SDMA with block diagonalization adaptive beamforming
4.1 Overview of SA technology
Originally, SA pre-processing techniques were proposed for military communications. Due to
the signicant technological advancements over the past two decades, SA-based technologies
have become a cost-efcient solution for commercial communication systems to overcome
some of the major challenges such as multipath fading, CCI, and capacity limitations
especially for the cell-edge users. By exploiting the spatial diversity and the spatial processing
capabilities of SA, an efcient utilization of available bandwidth and, hence, an increased
system spectral efciency is facilitated.
This section highlights the major features of SA-based SDMA systems relevant for the
main contributions in this chapter. More specically, the review is aimed at the benchmark
SDMA system considered in this chapter. Also, this section briey describes the generalized
BD beamforming method for multiuser SDMA system (Pan et al., 2004), where the BS
transmits multiple spatially multiplexed independent data streams to a group of users
selected according to a scheduling criterion. Due to physical size constraints at the user
side, the MSs are assumed to be equipped with limited number of multiple omnidirectional
antennas (OAs) (two throughout this chapter). This assumption is also convenient in order to
maintain affordable cost and reduced complexity at the mobile. As depicted in Fig. 2, each SA
Smart antenna
Fig. 2. A schematic illustration of a SA-based beamforming SDMA system
consists of an array of antenna elements and is dedicated to directionally transmit/recieve a
single independent spatial stream. Such spatial stream is referred to as an effective antenna
according to LTE terminology.
4.2 Multiuser MIMO BD system model
Consider a single-cell downlink MU-MIMO system where the BS is equipped with N
T
transmit SAs, and communicates over multiple MIMO channels with K users. It is assumed
that all users are equipped with the same number of receiving OAs denoted as N
R
. To simplify
the following analysis, it is assumed that N
R
< N
T
and N
T
is an integer multiple of N
R
in
order to serve all users. Fig. 3 illustrates an example of the considered SDMA system where
N
R
= 2.
MIMO
encoder
) (
1
t x
) (t x
MIMO
decoder
) (
1
1
t y
1 , 1
1
h
Data
sink
) (
1
t y
Data
Source
MIMO
decoder
Data
sink
) (t
K
y
2 , 1
1
h
1 , 1
K
h
) (
2
1
t y
) (
2
t x
) (t x
T
N
2 , 1
K
h
) (
1
t y
K
) (
2
t y
K
Fig. 3. SDMA system model with multiple MIMO users each equipped with 2 OAs
The at fading MIMO channel matrix for user k is denoted as H
k
where h
(j,i)
k
is the fading
coefcient between the j
th
transmit antenna and the i
th
receive antenna of user k. For analytical
simplicity, the rank r
k
of H
k
is assumed to be equal to min(N
R
, N
T
) for all users. Again,
channel estimation errors caused by various reasons such as feedback delay, and feedback
quantization error, etc. are beyond the scope of this chapter. Hence, it is assumed that the BS
has perfect CSI for all users. By assuming that the number of data streams s
k
to user k is equal
to N
R
, the transmitted data streams to user k can be denoted as a N
R
-dimensional vector x
k
where
K
k=1
s
k
N
T
. Since CSI is available at both sides of the MIMO link, it is assumed
that the MIMO transmission includes linear pre-processing and post-processing performed at
both BS and MS, respectively. Prior to transmission, the data vector of user k, x
k
, is multiplied
by a N
T
N
R
precoding matrix T
k
. In this chapter, it is assumed that T
k
is generated using
the BD beamforming algorithm which is a member of the zero-forcing (ZF) type multiuser
precoding algorithms (Spencer et al., 2004). At the BS, the data vectors of the K users are
linearly superimposed and propagated over the channel fromall N
T
antennas simultaneously.
It is also assumed that the elements of x
k
are independent and identically distributed (i.i.d.)
with zero mean and unit variance. The signal vector received at user k is
y
k
= H
k
K
=1
T
+n
k
(1)
Equation (1) can be rewritten in the form of summation of the desired signal, the interference
signal, and the noise signal as follows
y
k
= H
k
T
k
x
k
+H
k
K
=k
T
+n
k
(2)
where the rst termon the right-hand-side of (2) is the desired signal, the second termdenotes
the intracell CCI experienced by user k, and the last term n
k
is the additive white Gaussian
noise vector received by user k. According to key principles of BD algorithms, T
is designed
such that H
k
T
= 0 for = k and , hence, the intracell CCI is completely eliminated (nulled).

The block matrices H
S
and T
S
can be dened as the system channel matrix and the system
precoding matrix, respectively, as follows:
H
S
=
_
H
H
1
H
H
2
. . . H
H
K
_
H
(3)
T
S
= [T
1
T
2
. . . T
K
] (4)
Under the constraint of zero intracell CCI among users within the same cell, the optimal
solution can be obtained by diagonalizing the product of (3) and (4) H
S
T
S
. Now,

H
S
can
be dened as the block matrix of the MIMO links interfering with user k as follows
H
S
=
_

H
H
1
. . .

H
H
k1

H
H
k+1
. . .

H
H
K
_
H
(5)
In light of the above, the intracell CCI free constraint requires that T
k
is selected to lie in
the null space of

H
S
. The details of designing the BD precoding matrices and the associated
constraints can be found in (Chen et al., 2007; Pan et al., 2004; Spencer et al., 2004).
5. Problem statement
To fully exploit multiuser diversity the following questions have to be addressed. In a spatial
multiplexing opportunistic SDMAsystemwith an excess number of SAs at the BS, howshould
the optimal set of spatially separable users be chosen? What is the appropriate allocation
of the transmit/receive antennas (spatial beams) targeting the selected users? Since CCS
pilots only provide a sub-optimal metric (i.e. CSI), how can a better metric be provided (i.e.
instantaneous SINR) for such optimization problem while maintaining the same inherent
feedback bandwidth and delay efciency? For practical reasons such as cost and physical
size, the number of SAs at the BS is greater than the number of OAs at the MS, as is the case
in EUTRA. Algorithms that achieve spatial multiplexing gains such as V-BLAST receivers
(Foschini, 1996), however, require that the number of antennas at the receiver is greater than
or equal to the number of transmit antennas. Consequently, the number of DL spatial streams
is limited by the number of OAs at the MS side. This situation results in a spatial DoF for the
selection of the subset of transmit SA for per user DL transmission.
Note that according to the fourth assumption in section 2, it is assumed that the BD algorithm
can only eliminate intracell CCI. However, the system is still limited by intercell CCI.
Therefore, throughout this chapter, whenever the term interference is mentioned, it refers to
CCI originated from transmitters in the adjacent cells.
6. Interference-aware metric for downlink optimization
In the following we make use of channel reciprocity which is best available in the TDD mode.
The newinterference-weighted feedback concept proposed in this chapter is applied to the UL
CCS pilots. The rst major contribution of this chapter is the use of modied pilots, termed UL
IWCS pilots, which are proposed to replace the UL CCS pilots. In particular, the CCS pilots are
modied to become IWCS pilots by weighting (dividing) the UL CCS pilots by the magnitude
of the intercell CCI received at each MS. The UL IWCS pilots are then used at the BS to extract
the CSI plus the level of the intercell CCI experienced by the respective MS. Thereby, the SINR
at each active MS is conveyed to the respective BS without any additional signalling overhead.
The key idea of applying the interference-weighting concept to the CCS pilots is depicted in
Fig. 4 for a SISO case.
M
a
g
n
i
t
u
d
e

o
f

p
i
l
o
t

s
i
g
n
a
l
UL CCS
P
UL IWCS
Other-cell CCI
at MS
DL
I
P
DL
I
Fig. 4. Interference-weighting concept applied to the pilot signal in a SISO case
Consider an UL CCS transmission of a SDMA cellular interference-limited scenario with a
BS equipped with N
BS
SAs and K active users, each equipped with N
MS
OAs. A narrowband
at fading channel is assumed, i.e., a frequency subcarrier if orthogonal frequency division
multiple access (OFDMA) or single carrier frequency division multiple access is used. We
assume that both the BS and MSs experience sufcient local scattering. Therefore, both real
and imaginary parts of the entries of H
k
are samples of a zero-mean Gaussian distribution.
Hence, the conventional channel sounding MS-to-BS pilots transmission can be modeled as
follows:
y
u
(t) = H
k
(t)z
k
(t) +n
k
(t) (6)
where t is the time index. The predetermined pilot signal z
k
(t) is a N
MS
-dimensional vector;
the received signal y
k
(t) is a N
BS
-dimensional vector. Conventional channel sounding pilots
can be used to estimate two metrics: distance-dependent link gain (link budget), and the
multipath fading channel coefcients (small-scale fading). By weighting (6) by the intercell
CCI experienced by user k in the downlink, the IWCS transmission can be written as follows:
y
k
(t) =
H
k
(t)z
k
(t)
I
DL
k
(t 1)
+n(t) (7)
where I
DL
k
(t 1) is the amplitude of the intercell CCI experienced by k
th
MS at (t 1)
th
time interval. Consequently, the interference-weighting concept enables the sounding pilots
to be used by the BS to obtain the DL interference-aware-metric O
DLIAM
k
(t) C
N
BS
N
MS
which can be formulated as follows
O
DLIAM
k
(t) =
H
k
(t)
I
DL
k
(t 1)
(8)
From an information theoretic point of view, this metric (which can be considered as the
square root of the instantaneous MIMO DL signal-to-interference ratio (SIR)) provides better
feedback information compared to the CCS case. Consequently, the DL joint user scheduling
and transmit SA selection can be improved since the DL interference-aware-metric can be
used to obtain the SINR at the user side. Alternatively, quantized SINR can be fed back via the
direct feedback method, but this requires transmission resources and longer time (potentially
resulting in outdated feedback).
6.1 Interference estimation
From a practical point of view, the CCI experienced by a MS or a BS can be estimated by
allowing the respective entity to sense the channel when no intended transmission is taking
place. For instance, the CCI sensing period can be set for an active MS to be between the
end of the DL period and the beginning of the switching time (DL to UL). Then, the MS can
quantize the sensed CCI during the switching time period. Afterwards, a pilot weighted by
the quantized magnitude of CCI can be transmitted during the subsequent UL period. This
means that strong CCI will cause an articial attenuation of the pilot signal. This will pretend
a bad channel at the BS, i.e the probability that the BS will schedule this particular resource
block for this MS is reduced.
6.2 Optimization methodology
The general approach used in this chapter to maximize the sum capacity is a brute-force
search. As in the example illustrated in Fig. 5, each BS, equipped with 4 SAs, will select 2
MSs each equipped with 2 OAs to access its spatial resources. Clearly, each MS experiences
independent levels of CCI, and is subject to independent channel conditions on the desired
link. For instance, the BS in cell 1 can schedule 2 users out of 3 possible candidates. For
any active user, a BS can assign 2 out of 4 SAs to establish communication links. Using
combinatorial basics, the BS has (
3
2
) (
4
2
) = 18 options to select antennas and user pairs in the
given example.
The procedure followed to extract the optimization metrics provided by IWCS pilots is
illustrated in the example depicted in Fig. 6 which is based on the example of Fig. 5. Basically,
each way in which the BS can distribute the 4 SAs among 2 out of 3 users forms a possible
solution. FromFig. 6, two arbitrary solutions are highlighted by shading them with squares of
different colors and different styles for the borderline. By considering the solution shaded by
blue squares of solid borderline, it can be seen that MS1 is allocated the rst two SAs, while
Cell2 Cell1
Interference link
Desired link
MS1
MS3 MS2
MS4
MS5
MS6
I
1
I
2
I
3
Fig. 5. Interference-limited multiuser MIMO system where each BS is equipped with 4
antennas, each MS is equipped with 2 antennas.
3
) 2 , 4 (
3
3
) 1 , 4 (
3
2
) 2 , 4 (
2
2
) 1 , 4 (
2
1
) 2 , 4 (
1
1
) 1 , 4 (
1
3
) 2 , 3 (
3
3
) 1 , 3 (
3
2
) 2 , 3 (
2
2
) 1 , 3 (
2
1
) 2 , 3 (
1
1
) 1 , 3 (
1
3
) 2 , 2 (
3
3
) 1 , 2 (
3
2
) 2 , 2 (
2
2
) 1 , 2 (
2
1
) 2 , 2 (
1
1
) 1 , 2 (
1
3
) 2 , 1 (
3
3
) 1 , 1 (
3
2
) 2 , 1 (
2
2
) 1 , 1 (
2
1
) 2 , 1 (
1
1
) 1 , 1 (
1
#4 beam BS
#3 beam BS
#2 beam BS
#1 beam BS
#2 ant.
MS3
#1 ant.
MS3
#2 ant.
MS2
#1 ant.
MS2
#2 ant.
MS1
#1 ant.
MS1
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
I
H
SVD
MS1
Eigen
values
?
MS3
Eigen
values
MS3
Metric
coefficient
MS1
Metric
coefficient
Fig. 6. Interference-limited virtual MU-MIMO matrix for the example illustrated in Fig. 5.
MS3 is allocated the last two antennas. Clearly, it can be seen that each SA allocation forms a
(2 2) square matrix containing the coefcients of the optimization metric.
The next step is to obtain the eigenvalues of each square matrix via singular value
decomposition (SVD). Afterwards, the eigenvalues of each group of selected users are
summed. Finally, the summation of eigenvalues will be used to nd the optimum solution
among all possible solutions according to the different scheduling criteria. To examine all
possible solutions, two approaches are considered: exhaustive search (ES) and HA. The details
of the two approaches are discussed in section 8.
7. Link-protection-aware metric for uplink and downlink optimization
The main purpose of this section is to propose a method to accommodate the IWCS pilots to
suit UL optimization. Since the amplitude of CCI experienced at the BS I
UL
j
(t 1) (referred
to as UL interference) where (j 1, ..., N
BS
) is different from the CCI experienced at the MS
I
DL
k
(t 1), the DL interference-aware-metric in (8) is not suitable for UL optimization. In
order to use the IWCS pilots for UL optimization, the BS weights each row of the metric
dened in (8) by the received interference at the associated SA I
UL
j
(t 1) at the BS, which
is assumed to be known at the BS side. Thus, the new optimization metric, referred to as
link-protection-aware-metric, O
LPAM
k
(t) C
1N
MS
can be formulated as follows:
O
LPAM
k
(t) =
H
(j)
k
(t)
I
DL
k
(t 1)I
UL
j
(t 1)
(9)
where H
(j)
k
(t) is the j
th
row in H
k
(t).
Basically, this metric allows the BS to decide which subset of receive antennas it should use for
user k, and which users should be grouped together. Since the link-protection-aware-metric is
inversely proportional to I
DL
k
(t 1), a MS experiencing low interference level has higher
chances to be selected. Due to channel reciprocity, a user that receives little interference
from a set of users (in Tx mode) in a particular time slot, this user only causes little
interference to the same set of users (now in Rx mode) in a different time slot. Therefore,
the link-protection-aware-metric decreases the probability of scheduling users that are
potential strong interferers. Consequently, this forms an inherent link-protection for the
already established links in the neighboring cells. Similarly, this metric can be used for DL
optimization to jointly select a subset of users experiencing favorable channel conditions and
low intercell CCI to receive from a subset of SAs causing low intercell CCI to the neighboring
cells. Note that the link-protection-aware-metric is simultaneously used to improve both UL
and DL performance. Hence, the cross-layer scheduling for UL has to be the same as for DL,
which reduces the scheduling time.
According to the MIMO literature (Foschini & Gans, 1998; Telatar, 1999), if the channel matrix
H
k
is known at the BS, then the instantaneous DL MIMO channel capacity of user k using
xed transmit power
P
t
N
BS
per antenna can be expressed as the sum of capacities of r SISO
channels each weighted with power gain
k
i
where (i 1, ..., r), r is the rank of the channel,
and
k
i
are the eigenvalues of H
k
H
H
k
. P
t
is the total transmit power at the BS. Assuming an
interference-limited system, the instantaneous MIMO capacity of user k can be expressed as
follows:
C
k
=
r
i=1
log
2
(1 +
P
t
N
BS
I
DL
k

2

k
i
) (10)
By using the system model introduced above and (10), the primary objective is to nd the
optimum way, according to the scheduling criterion in use, in which the BS distribute N
BS
antennas among
N
BS
N
MS
spatially separable users out of user population of size K, where a
selected user communicates with exactly N
MS
SAs at the BS. For instance, in the case of a
maximum sum rate scheduling criterion, the optimum solution maximizes the capacity of the
multiuser MIMO channel at the expense of fairness.
The sample-space population (SP) (the size of the pool containing all possible solutions) of
such problem is formulated later in this chapter. The resulting optimization problem can be
written as follows:
C
max
= arg max
vSP
N
BS
N
MS
k=1
C
v
k
(11)
where SP are all possible choices of allocating the beams to the members of v, while v
represents a possible choice of grouping the scheduled users.
7.1 Summary of the optimization metrics
In the DL, the following four metrics are used in conjunction with the different scheduling
criteria (section IV summarizes them):
1 The scheduler only uses perfect instantaneous measurements of the MIMO multipath
fading channel coefcients (ignoring the distance-dependent link-gain). This metric is fair
by nature due to the uniform distribution of the users and the i.i.d. distributed fading, and
it is referred to as DL blind-metric O
DLBM
k
(t).
2 The scheduler uses instantaneous measurements of the distance-dependent link-gain. In
contrast, this metric is greedy by nature because it tends, in most of the cases, to select
those users which are closer to the BS, and it is referred to as DL link-gain-aware-metric
O
DLLGAM
k
(t).
Note, the rst two metrics above are supported by the conventional channel sounding
pilots and they do not provide the BS with information about interference at the mobile,
I
DL
k
(t 1).
3 The DL interference-aware-metric O
DLIAM
k
(t) dened in (8) is used. This metric is
supported by the IWCS pilots or by means of explicit signalling which, however, requires
additional resources for signalling trafc.
4 To foster link-protection awareness the metric developed in (9), O
LPAM
k
(t), is used.
Two cases are examined for the UL:
1 The scheduler uses instantaneous measurements of the CSI of each spatial stream (each
row of H
k
) divided by the interference level experienced by the associated SA at the
BS. This metric, referred as UL interference-aware-metric, O
ULIAM
k
(t) C
1N
MS
can be
expressed as follows:
O
ULIAM
k
(t) =
H
(j)
k
(t)
I
UL
j
(t 1)
(12)
It is important to mention that this metric can be obtained using UL CCS pilots since UL
interference I
UL
j
(t 1) is available at the BS without feedback.
2 This is similar to the fourth case for the DL metrics which is dened as the
link-protection-aware-metric in (9). This metric jointly considers UL and DL performance.
In summary, the results shown in section 10 are associated with four different metrics for DL
transmission, and two different metrics for UL transmission.
7.2 Numerical example
To show the link-protection feature of the new metric dened in (9), a simple example is
presented in Fig. 7. In this example, the arbitrary numbers quantifying the gain of each link
and the interference experienced by each entity are used to estimate the achievable capacity
using (10). It is assumed that cell 2 has an established DL transmission with MS3 and the
argument for the achievable DL capacity is
H
MS3
I
MS3
=
9
3
. The neighboring BS, BS1, has got
assigned the same time slot for UL transmission and it attempts to select between MS1 and
MS2. If BS1 uses the UL interference-aware-metric, MS1 is scheduled for UL;
H
MS1
I
BS1
=
6
2
>
H
MS2
I
BS1
=
5
2
. Hence, the argument for the achievable UL capacity is
H
MS1
I
BS1
=
6
2
. As a result, it
is assumed that the interference at MS3 increases to I
MS3
= 4.5 due to the low shadowing
MS1
MS2
BS1
MS3
BS2
MS4
Cell1
UL
Cell2
DL
I
BS1
=2
H
MS2
=5
H
MS1
=6
I
MS2
=0.5
I
MS1
=3
I
BS2
=2.5
I
MS3
=3
I
MS4
=3
H
MS3
=9
H
MS4
=6
High shadowing
H
i
g
h
s
h
a
d
o
w
i
n
g
Fig. 7. Example of interference-limited 2-cell scenario with which the basic working principle
of the link-protection-aware-metric is illustrated.
conditions between MS3 and MS1. This reduces the argument of the current achievable DL
capacity in cell 2 to
9
4.5
. Thus, the arguments of the achievable system capacity are
6
2
and
9
4.5
.
In contrast, if BS1 uses the link-protection-aware-metric, MS2 is scheduled;
H
MS1
I
MS1
I
BS1
=
6
32
<
H
MS2
I
MS2
I
BS1
=
5
0.52
. Hence, the argument of the achievable UL capacity is
H
MS1
I
BS1
=
5
2
.
Consequently, the interference at MS3 does not change (due to the high shadowing between
MS2 and MS3), and therefore, the arguments of the achievable cell capacity are
5
2
and
9
3
for
cell 1 and cell 2, respectively. In summary, in the rst case the sum of the arguments of the
cell capacity is 5 whereas in the second case the sum is 5.5. Clearly, it can be seen from this
example that the link-protection-aware-metric used in the second case improves the overall
spectral efciency.
8. Heuristic algorithm for reduced computational complexity
8.1 Introduction
Generally the ES is not practical due to its computational complexity. Therefore, in this section
a heuristic algorithm is proposed to reduce the involved complexity.
8.2 Exhaustive search mathematical model
The complexity of exhaustive search approach for scheduling optimization of SDMA system
depends on the total number of users K, the number of SAs at the BS N
BS
, and the number
of antennas at each MS N
MS
. The search burden for the scheduler is equivalent to the SP size.
Using combinatorial fundamentals, (13) and (14) can be obtained. By comparing (13) with (14),
it can be noticed that the multiuser diversity plays a signicant role when K >
N
BS
N
MS
; due the
fact that not all users can be scheduled.
If K
N
BS
N
MS
SP =
N
BS
!
_
(N
MS
!)
K
(N
BS
K N
MS
)!
_
. .
SDMA DoF
; (13)
if K >
N
BS
N
MS
SP =
K!
__
N
BS
N
MS
_
!
_
K
N
BS
N
MS
_
!
_
. .
Multiuser DoF
N
BS
!
(N
MS
!)
_
N
BS
N
MS
_
. .
SDMA DoF
. (14)
It is important to mention that (13) and (14) are applicable only for scenarios in which N
BS
is
an integer multiple of N
MS
. To gain insight into the inuence of each dimension of the DoF
on the SP, Fig. 8, Fig. 9(a), and Fig. 9(b) are obtained assuming N
MS
to be 2. By comparing
5
10
15
20
25
30
4
8
12
16
20
10
0
10
5
10
10
10
15
10
20
10
25
Number of users
Multiuser MIMO downlink scheduling optimization
Number of antennas at the BS
S
a
m
p
l
e
s
p
a
c
e

p
o
p
u
l
a
t
i
o
n
2 antennas at each MS
4 antennas at each MS
Fig. 8. The search complexity versus the number of SAs at the BS and the number of users,
assuming 2 and 4 antennas at each MS
Fig. 9(a) with Fig. 9(b), it can be seen that SP grows polynomially with the number of users,
and exponentially with the number of antennas at the BS (Learned et al., 1997). Clearly, the ES
approach is computationally expensive for large number of users and antennas.
5 10 15 20 25 30
10
0
10
5
10
10
10
15
10
20
10
25
Number of users
S
e
a
r
c
h

s
p
a
c
e

s
i
z
e
4 antennas at the BS
(a) Search space size versus the number of users
4 6 8 10 12 14 16 18 20
10
0
10
5
10
10
10
15
10
20
10
25
Number of antennas at the BS
S
e
a
r
c
h

s
p
a
c
e

s
i
z
e
1 user
5 users
10 users
15 users
20 users
25 users
30 users
(b) Search size versus the number of SAs at the BS
Fig. 9. Search space size assuming 2 antennas at each MS
8.3 Heuristic to reduce complexity
A HA is developed to signicantly reduce the computational complexity of nding a close
to optimum solution. The algorithm exploits two basic principles. The rst one is based
on the elimination of users that experience excessive interference while the second is based
on an angle of arrival (AoA) sectorization approach. The former aims to reduce the SP
by suppressing those users exposed to interference levels greater than a predetermined
interference threshold. The latter converts the cell of interest (CoI) optimization problem into
smaller optimization problems, which can be solved simultaneously. This can be done by
sectorizing the cell based on the AoA of the uplink channel sounding signal. Subsequently,
the antennas at the BS are distributed among these sectors according to the user density of
each sector. As a consequence, the original SP of the CoI will be reduced signicantly as the
complexity is split into smaller parallel search spaces.
8.4 Numerical examples
To demonstrate the working mechanism of the HA, let us consider a cell with 28 users, each
equipped with 4 antennas, and the BS equipped with 16 antennas. According to (14), the
scheduler has to search through approximately 3.027 10
9
possibilities to nd the optimum
solution. By applying the interference based user elimination approach, let us assume that
8 users have been found to be exposed to severe interference levels. Similarly, assume that
the cell can be equally divided into four sectors with the same user density. According to the
proposed HA, the original optimization problem becomes four identical sector optimization
problems. Hence, each sector has 4 associated antennas at the BS, and
288
4
= 5 admissible
users, each equipped with 4 antennas. Since only one user can be supported per sector, the
scheduler has to search through 5 possibilities per sector, which can be done in parallel for all
sectors.
To showthe effect of the sectorization and the number of antennas at the MS, another example
is used. In this example, a cell with 6 users, each equipped with 2 antennas, and the BS
equipped with 12 antennas, is considered. Following a similar procedure as in the previous
example, assume that all users are admissible, and that the cell can be equally sectorized into
only 2 sectors with the same user density. Thus, each sector has 6 associated antennas at the
BS, and 3 users, each equipped with 2 antennas. According to (13), the optimization problem
with a search space of 3.992 10
7
is converted into 2 identical optimization problems, each
with a search space of 120. It is worth noting in this example that multiuser diversity has not
been exploited, and a simple sectorization technique is applied. However, the scheduler search
burden is signicantly reduced, and thus the practicability of the proposed interference-aware
antenna selection and scheduling algorithm is greatly enhanced.
9. Scheduling criteria
Scheduling criteria can be generally classied into two groups: greedy and fair. In this chapter,
the considered optimization metrics are tested using three scheduling criteria. Specically,
a greedy criterion referred to as maximum capacity (MC) (Borst & Whiting, 2003; Gesbert
et al., 2007), and two fair criteria referred to as proportional-fair (PF) (Chaponniere et al., 2002;
Viswanath et al., 2002) and score-based (SB) (Bonald, 2004) are considered.
9.1 Maximum capacity scheduling criterion
MC criteria always assigns the antennas to those users with the highest eigenvalue sum,
hence, it is efcient in terms of overall throughput but may look oppressive for low SINR
users, typically located far from the BS (cell-edge users).
The achievable capacity according to MC scheduling criterion can be formulated as follows:
C
MC
= arg max
vSP
{C
v
}, (15)
where C
MC
is the achievable capacity using MC criterion, C
v
is the instantaneous capacity of
the v
th
option.
9.2 Proportional fair scheduling criterion
The PF criterion seeks to assign the antennas to those users with the highest eigenvalue
sum relative to their mean eigenvalue sum. This causes the scheduler to realize a reasonable
trade-off between throughput efciency and fairness.
The achievable capacity according to PF scheduling criterion can be formulated as follows:
C
PF
= arg max
vSP
_
C
v
C
v
_
, (16)
where C
PF
is the achievable capacity using PF criterion, C
v
and C
v
are the instantaneous and
average capacity of the v
th
option, respectively.
9.3 Score-based scheduling criterion
The SB scheduler assigns the antennas to those users with the best scores. The score
corresponds to the rank of the current eigenvalue sum among the past values observed over
a time window of a particular length, W (Bonald, 2004).
The achievable capacity according to SB scheduling criterion can be formulated as follows:
C
SB
= arg min
vSP
{S
v
}, (17)
where C
SB
is the achievable capacity using the SB criterion, and S
v
is the score of the
instantaneous capacity of the v
th
option among the past values observed over the time
window W. The throughput-fairness trade-off is customizable by changing the window size
W. More specically, setting the windowsize equal to one, this will result in a similar behavior
as observed with the MC criterion. A large window size converges to the behavior of the PF
scheduler.
Since the score, S
, is an integer number several scheduling scenarios could result in the same

best rank (the rst position). In this case, the solution which achieves the highest throughput
is selected. Therefore, in this chapter the SB criterion effectively results in a hybrid version of
both the SB scheduler and the MC scheduler.
10. Results and discussion
In this section, the performance of the different optimization metrics is assessed using the
cumulative distribution function (cdf) of per-user capacity and cell capacity for both UL and
DL. The per-user capacity in this work is the 2 2 spatial multiplexing MIMO capacity which
can be calculated using (10). The cell capacity is the sum capacity of the group of scheduled
users. In order to shed some light on fairness of the considered optimization metrics, the cdf
of the scheduled (served) user distance to the BS is obtained for all considered scheduling
criteria.
Atwo-tier cellular platformconsisting of hexagonal cells is used in the simulation. In each cell,
the BS and MSs are equipped with (6 or 8) and 2 antennas, respectively. Using the parameters
given in Table 1, the 19-cell cellular TDD system with uniform user distribution, as shown in
Fig. 10, is simulated using the Monte Carlo method. The channel matrix of k
th
user H
k
is a
zero-mean i.i.d. complex Gaussian random matrix. The system level performance evaluation
is based on both central cell and wrap-around techniques. The DL and UL performances are
Fig. 10. Interference-limited multiuser MIMO system implementing orthogonal SDMA
analyzed for asynchronous UL/DL transmissions, where the two possible link-directions (UL
or DL) occur with equal probability and independently from each other. All entities (BSs and
MSs) in the system use xed transmit power of 30 dBm.
In the wrap-around technique, the multicell layout is folded such that cells on the right side of
the network are connected with cells on the left side and, similarly, cells in the upper part
of the network get connected to cells in the lower part. The created area may be seen as
borderless, but with a nite surface, and it may be visualized as a torus. The wrap-around
technique is suitable for both downlink and uplink simulations. One of the main advantages of
wrap-around technique is that the decision taken by a scheduler in a particular cell inuences
the scheduling behavior in the adjacent cells. Such an observation cannot be collected using
the central cell technique. Another advantage compared to the central cell technique is that
simulation data can be collected from all cells, which may reduce the required simulation
time to collect sufcient statistics.
10.1 Downlink performance with interference awareness
In order to avoid a huge SP size, the BS is assumed to be equipped with 6 SAs. Therefore,
the same time-frequency resource is simultaneously used, at maximum, three times within
the same cell for three different users. The results in Fig. 11(a) show that the cell capacity
can be signicantly enhanced if knowledge of interference is taken into account in the
antenna selection and MIMO user scheduling process. For example it can be seen from Fig.
11(a), for the MC criterion (unmarked curves), that the median of the DL capacity for the
Channel propagation
environment Suburban micro cell
BS-to-BS distance 300-500 m
Number of interfering tiers 2
Pilot Tx power 30 dBm
Number of users 300-1000
Carrier frequency 2 GHz
Minimum distance between
MS and BS 35 m
Number of antennas at BS 6-8
Number of antennas at MS 2
log-normal shadowing
standard deviation,
S
8 dB
Correlation distance
of shadowing 50 m (as in UMTS 30.03)
Path loss model (dB) 31.5 +35 log10(d)
Table 1. Simulation parameters
0 5 10 15 20 25 30 35 40
0
0.2
0.4
0.6
0.8
1
Downlink capacity in CoI [bits/s/Hz]
C
u
m
u
l
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MCblind
MClinkgainaware
MCinterferenceaware
SBblind
SBlinkgainaware
SBinterferenceaware
PFblind
PFlinkgainaware
PFinterferenceaware
Proportional fair (PF)
Max capacity (MC)
Scorebased (SB)
(a) Cell performance
0 5 10 15
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Peruser capacity in CoI [bits/s/Hz]
C
u
m
u
l
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MCblind
MClinkgainaware
MCinterferenceaware
SBblind
SBlinkgainaware
SBinterferenceaware
PFblind
PFlinkgainaware
PFinterferenceaware
Max capacity (MC)
Scorebased (SB)
(b) User performance
Fig. 11. Downlink performance comparison among the DL interference-aware and both blind
and link-gain-aware metrics using ES approach
DL interference-aware-metric (using the novel concept of interference-weighted pilots) is 10
bps/Hz while it is 7.5 and 4 bps/Hz for the link-gain-aware-metric and the blind-metric,
respectively. As expected from the greedy characteristics of the link-gain-aware-metric, it can
be seen that it outperforms the blind-metric. The greediness of these metrics will be revisited
later when the issue of fairness is discussed. Similarly, the per-user capacity in the case of the
DL interference-aware-metric is substantially improved for all scheduling criteria. From Fig.
11(b), for the SB criterion (circle-marked curves), it can be seen that the median of the per-user
capacity resulting from the DL interference-aware-metric is 2 bps/Hz while it is 1 and 0.5
bps/Hz for link-gain-aware-metric and the blind-metric, respectively.
The results in Fig. 12 are obtained for the case of a cell radius of 300 m. It is shown
that both DL interference-aware-metric and link-gain-aware-metric prioritize the users
closer to the BS when the MC criterion is used. For instance, Fig. 12, assuming the MC
criterion (unmarked curves), shows that the median of the served user distance to the
BS of the DL interference-aware-metric and the link-gain-aware-metric are 160 m and
140 m, respectively, while it is 190 m for the blind-metric case. Alternatively, when the PF
criterion (diamond-marked curves) is used the DL interference-aware-metric shows high
level of fairness (median is 220 m) compared to both blind-metric (median is 175 m) and
link-gain-aware-metric (median is 210 m).
0 50 100 150 200 250 300
0
0.2
0.4
0.6
0.8
1
Link distance to BS [m]
C
u
m
u
l
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MCblind
MClinkgainaware
MCinterferenceaware
SBblind
SBlinkgainaware
SBinterferenceaware
PFblind
PFlinkgainaware
PFinterferenceaware
Scorebased (SB)
Max capacity (MC)
Fig. 12. Fairness comparison among the three considered scheduling criteria in terms of the
distance between the BS and the scheduled user.
It is worth noting that the distance distribution of the served users associated with the
blind-metric is only slightly affected by using the SB and the PF criteria. This is expected due to
the inherent fairness of this metric which is a consequence of the i.i.d. distributed small-scale
fading and the random distribution of the users.
10.2 Performance of heuristic algorithm
In order to study the efciency of the HA, a DL scenario is simulated under the following
assumptions: 8 SAs at the BS, 2 antennas at each MS, xed sectorization is used to equally
divide the cell into two sectors, only the best 8 users in terms of the experienced intercell CCI
are considered in each cell, and the MC criterion is used. It has been found using simulations
that the search space has been reduced from 176400 to 80 on average (depending on the
user density per sector). This is achieved at the following cost: The median of the achievable
throughut using DL interference-aware-metric in combination with the HA approach only
reaches approximately 90% of the median of the optimum capacity using the exhaustive
search approach. This can be seen in Fig. 13.
Since the ES approach is computationally too expensive, simulation results for larger SP sizes
cannot be obtained within reasonable time. Therefore, all the subsequent results are obtained
using the HA approach due to its low complexity and close to optimum performance as
demonstrated in Fig. 13. Furthermore, in order to increase the spatial SDMA DoF, the BS is
assumed to be equipped with 8 SAs, hence 4 users can be scheduled.
10.3 Uplink performance with link-protection awareness
FromFig. 14(a), for the MCcriterion (square-marked curves), it can be seen that the UL median
cell capacity using the link-protection-aware-metric, (9), is 27 bps/Hz which corresponds to an
enhancement of 12.5%of the UL median cell capacity using the UL interference-aware-metric,
(12), which is 24 bps/Hz.
0 5 10 15 20 25 30 35 40
0
0.2
0.4
0.6
0.8
1
Downlink capacity in COI [bits/s/Hz]
C
u
m
u
l
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MCblindES
MClinkgainawareES
MCinterferenceawareES
MCblindHA
MClinkgainawareHA
MCinterferenceawareHA
Fig. 13. Throughput comparison for the metrics considered in the previous section for the
MC criterion using both ES and HA approaches
0 10 20 30 40 50 60
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SDMA wrapped~around 19 CeIIs scenario, each of radius = 300 m, (2x2 spatiaI muItipIexing)
UpIink ceII capacity [bits/s/Hz]
C
u
m
u
I
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MC~UL~nterference~awaremetric
MC~UL~Link~protection~aware~metric
PF~UL~nterference~awaremetric
PF~UL~Link~protection~aware~metric
SB~UL~nterference~awaremetric
SB~UL~Link~protection~aware~metric
0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
UpIink user capacity [bits/s/Hz]
C
u
m
u
I
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MC~UL~nterference~awaremetric
MC~UL~Link~protection~aware~metric
PF~UL~nterference~awaremetric
PF~UL~Link~protection~aware~metric
SB~UL~nterference~awaremetric
SB~UL~Link~protection~aware~metric
Fig. 14. Uplink performance comparison between the UL interference-aware-metric and the
link-protection-aware-metric
Notice that the link-protection-aware-metric, (9), jointly considers UL and DL and it implicitly
provides the BS with CCI interference observed at the MSs. Since channel reciprocity
is assumed, a user that observes high interference in the DL will in turn cause high
interference to the corresponding users in the other cell when they are receiving data from
their BS. Therefore, not scheduling a user that observes high interference is benecial in
two ways: (a) other users might observe less interference at these particular transmission
resources, and scheduling these users would result in higher user capacity or reduced
transmission power; and (b) interference caused to other users in the neighboring cells is
automatically kept a low level. This effectively results in interference-aware radio resource
management. These observations are afrmed by noting the results of the per-user capacity
of the link-protection-aware-metric, which outperforms the UL interference-aware-metric
for all scheduling criteria considered. For instance, from Fig. 14(b), for the SB criterion
(circle-marked curves), it can be seen that the median of the per-user capacity resulting
from the link-protection-aware-metric is 2.9 bps/Hz while it is 2.5 bps/Hz for the UL
interference-aware-metric. However, the PF criterion (triangle-marked curves) does not
provide a noticeable improvement due to the inherent fairness. The fairness constraint
prioritizes a fair allocation of the spatial resources among the users over preventing strong
interferers from being scheduled.
10.4 Downlink performance with link-protection awareness
The same analysis carried out in the previous subsection for the UL is now applied to the DL.
From Fig. 15(a), for the MC criterion (square-marked curves), it can be seen that the median
0 10 20 30 40 50 60
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DownIink ceII capacity [bits/s/Hz]
C
u
m
u
I
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n
MC~DL~nterference~aware~metric
MC~DL~Link~protection~aware~metric
PF~DL~nterference~aware~metric
PF~DL~Link~protection~aware~metric
SB~DL~nterference~aware~metric
SB~DL~Link~protection~aware~metric
0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
DownIink user capacity [bits/s/Hz]
C
u
m
u
I
a
t
i
v
e

d
i
s
t
r
i
b
u
t
i
o
n

f
u
n
c
t
i
o
n MC~DL~nterference~aware~metric
MC~DL~Link~protection~aware~metric
PF~DL~nterference~aware~metric
PF~DL~Link~protection~aware~metric
SB~DL~nterference~aware~metric
SB~DL~Link~protection~aware~metric
Fig. 15. Downlink performance comparison between the DL interference-aware-metric and
the link-protection-aware-metric.
of the DL capacity using the link-protection-aware-metric, (9), is 19 bps/Hz while it is 15
bps/Hz for the DL interference-aware-metric, (8). Similarly, the per-user capacity in the case
of the link-protection-aware-metric is improved for all scheduling criteria. From Fig. 15(b), for
the SB criterion (circle-marked curves), it can be seen that the median of the per-user capacity
resulting from the link-protection-aware-metric is 3.5 bps/Hz while it is 1.9 bps/Hz for the
DL interference-aware-metric.
Irrespective of the nature of the employed scheduling criterion, the results show that the
gain in the UL is lower compared to the DL. This primarily is due to the fact that BS-to-BS
interference degrades the link-protection selectivity in UL. Given that all antenna groups
at the BS site are in close proximity, the large-scale fading (log-normal shadowing) on the
interference links can, thus, be considered the same for this case. Therefore, the variations
among the interference links depend only on the small-scale fading.
On the other hand, the tendency of greediness of the SB criterion, explained in section 9, allows
for the exploitation of multiuser diversity in the DL given the high user density. Furthermore,
due to the relatively high variance of the statistics of DL interference, a marginal improvement
is achieved using the PF criterion, as it can be seen in Fig. 15. The gains are only materialized at
the high interference percentiles due to the emphasis on fairness when using the PF criterion
(triangle-marked curves).
11. Conclusions
Weighting the UL channel sounding pilots by the instantaneous CCI observed locally (referred
to as interference-weighted channel sounding pilots) enables the users to implicitly convey
the level of CCI to its BS along with channel state information. In addition, ICIC is achieved
without any explicit inter-cell signalling as follows: A user that experiences high CCI in the
DL would equally cause high interference to the same users that caused this interference when
the transmission directions are reversed (due to channel reciprocity). The pilot-weighting
mechanism "articially" deteriorates the channel in this case so that this particular user will
most probably be scheduled on a different channel which is benecial for this user as well
as for the interfering users in the neighboring cells (resulting in interference coordination). If
additionally the modied pilots are weighted by the UL interference (observed at the BS), this
provides combined knowledge of the interference at both ends. It has been shown that this
additional information, for example, enables interference-aware user scheduling to improve
the capacity compared to systems which only utilize the conventional channel sounding
pilots.
It has been found that compared to both blind-metric and link-gain-aware-metric, a
capacity gain of 150% and 35%, respectively, at the 10
th
percentile can be achieved
when the novel downlink interference-aware-metric is used assuming the maximum
capacity criterion. By considering the score-based policy, simulations show that using the
link-protection-aware-metric results in a capacity gain of 230% and 15% at the 10
th
percentile
compared to both downlink and uplink interference-aware-metric, respectively. Marginal
capacity gains have been obtained for the PF policy which ensures fairness at the expense
of capacity efciency. However, please notice that for the sake of conciseness only a single
channel has bee assumed in this study. Higher gains are envisaged for the PF policy if a
broadband OFDMA system with multiple resource blocks would have been considered.
Utilizing the proposed heuristic algorithmsignicantly reduces the computational complexity
to approximately 0.05% of the complexity of the exhaustive search approach. This reduction
in complexity is achieved at the cost of 8% loss at the 10
th
percentile cell capacity.
12. References
Abe, T. & Bauch, G. (2007). Differential codebook mimo precoding technique, Global
Telecommunications Conference, 2007. GLOBECOM 07. IEEE, pp. 3963 3968.
Abualhiga, R. & Haas, H. (2008). Implicit Pilot-Borne Interference Feedback for Multiuser
MIMO TDD Systems, Proc. of the International Symposium on Spread Spectrum
Techniques and Applications (ISSSTA), IEEE, Bologna, Italy, pp. 334338.
Airy, M., Bhadra, S., Heath, R. & Shakkottai, S. (2006). Transmit Precoding for the Multiple
Antenna Broadcast Channel, Proc. of the 63rd Vehicular Technology Conference (VTC 06),
Vol. 3, IEEE, Melbourne, Australia, pp. 13961400.
Ali, S. H., Lee, K.-D. & Leung, V. C. M. (2007). Dynamic Resource Allocation in OFDMA
Wireless Metropolitan Area Networks, IEEE Wireless Communications 14(1): 613.
Bahceci, I., Duman, T. & Altunbasak, Y. (2003). Antenna Selection for Multiple-Antenna
Transmission Systems: Performance Analysis and Code Construction, IEEE
Transactions on Information Theory 49(10): 26692681.
Bauch, G. & Dietl, G. (2008a). Enhanced mimo for imt-advanced wireless systems, 2008 IET
Seminar on Wideband and Ultrawideband Systems and Technologies: Evaluating current
Research and Development, pp. 1 21.
Bauch, G. & Dietl, G. (2008b). Multi-user mimo for achieving imt-advanced requirements,
International Conference on Telecommunications (ICT 2008), pp. 1 7.
Blum, R. (2003). MIMO Capacity with Interference, IEEE Journal on Selected Areas in
Communications 21(5): 793801.
Bonald, T. (2004). A Score-Based Opportunistic Scheduler for Fading Radio Channels, Proc. of
the European Wireless Conference (EWC), Barcelona, Spain.
Borst, S. & Whiting, P. (2003). Dynamic channel-sensitive scheduling algorithms for
wireless data throughput optimization, IEEE Transactions on Vehicular Technology
52(3): 569586.
Catreux, S., Driessen, P. & Greenstein, L. (2002). Data Throughputs Using Multiple-Input
Multiple-Output (MIMO) Techniques in a Noise-Limited Cellular Environment, IEEE
Transactions on Wireless Communications 1(2): 226235.
Chae, C.-B., Mazzarese, D. & Heath, R. W. (2006). Coordinated Beamforming for Multiuser
MIMO Systems with Limited Feedforward, Fortieth Asilomar Conference on Signals,
Systems and Computers (ACSSC) pp. 15111515.
Chaponniere, E. F. Black, P. J., Holtzman, J. M. & Tse, D. N. C. (2002). Transmitter Directed
Code Division Multiple Access System Using Path Diversity to Equitably Maximize
Throughput, US Patent 6449490 .
Chen, R., Heath, R. W. & Andrews, J. G. (2007). Transmit Selection Diversity for Unitary
Precoded Multiuser Spatial Multiplexing Systems With Linear Receivers, IEEE
Transactions on Signal Processing 55(3): 11591171.
Choi, L.-U. & Murch, R. (2004). A Transmit Preprocessing Technique for Multiuser
MIMO Systems Using a Decomposition Approach, IEEE Transactions on Wireless
Choi, W., Forenza, A., Andrews, J. G. &Heath Jr., R. W. (2006). Capacity of Opportunistic Space
Division Multiple Access with Beam Selection, Proc. of the Global Telecommunications
Conference (GLOBECOM 06), IEEE, San Francisco, USA, pp. 15.
Chung, S. T., Lozano, A. & Huang, H. (2001a). Approaching Eigenmode BLAST Channel
Capacity Using V-BLAST With Rate and Power Feedback, Proc. of the 54th Vehicular
Technology Conference (VTC 01), Vol. 2, Atlantic City, New Jersey, pp. 915919.
Chung, S. T., Lozano, A. &Huang, H. (2001b). LowComplexity Algorithmfor Rate and Power
Quantization in Extended V-BLAST, Proc. of the 2001 IEEE 53rd Vehicular Technology
Conference, Vol. 2, Atlantic City, New Jersey, pp. 910914.
Costa, M. (1983). Writing on Dirty Paper, IEEE Transactions on Information Theory 29(3): 439441.
Dai, H., Molisch, A. & Poor, V. H. (2004). Downlink Capacity of Interference-Limited
MIMO Systems with Joint Detection, IEEE Transactions on Wireless Communications
3(2): 442453.
Foschini, G. J. (1996). Layered Space-Time Architecture for Wireless Communication in a
Fading Environment when Using Multi-Element Antennas, Bell Labs Technical Journal
1(2): 4159.
Foschini, G. J. & Gans, M. J. (1998). On Limits of Wireless Communications in a Fading
Environment when Using Multiple Antennas, Wireless Personal Communications
6(6): 311335.
Fragouli, C., Al-Dhahir, N. & Turin, W. (2003). Training-Based Channel Estimation
for Multiple-Antenna Broadband Transmissions, IEEE Transactions on Wireless
Fuchs, M. & Del Galdo, G. & Haardt, M. (2007). Low-Complexity SpaceTimeFrequency
Scheduling for MIMOSystems With SDMA, IEEE Transactions on Vehicular Technology
56(5): 27752784.
Gallen, C. (2009). In 2014 Monthly Mobile Data Trafc Will Exceed 2008 Total, ABI Research,
Retrieved January 3, 2011, from www.abiresearch.com/press/
Gesbert, D., Kiani, S. G., Gjendemsj, A. & ien, G. E. (2007). Adaptation, Coordination,
and Distributed Resource Allocation in Interference-Limited Wireless Networks,
Proc. of the 7th IEEE International Symposium on Wireless Communication Systems
95(12): 23932409.
Ghrayeb, A. & Duman, T. (2002). Performance analysis of MIMO Systems with Antenna
Selection Over Quasi-Static Fading Channels, pp. 333337.
Goldsmith, A., Jafar, S., Jindal, N. & Vishwanath, S. (2003). Capacity Limits of MIMO
Channels, IEEE Journal on Selected Areas in Communication 21(5): 684702.
Gore, D., Heath, R. & Paulraj, A. (2002). Statistical Antenna Selection for Spatial Multiplexing
Systems, Proc. of the International Conference on Communications (ICC 02), Vol. 1, New
York, USA, pp. 450454.
Gore, D. & Paulraj, A. (2002). MIMO Antenna Subset Selection with Space-Time Coding, IEEE
Transactions on Signal Processing 50(10): 25802588.
Gorokhov, A., Gore, D. & Paulraj, A. (2003). Receive Antenna Selection for MIMO
Flat-Fading Channels: Theory and Algorithms, IEEE Transactions on Information
Theory 49(10): 26872696.
Haas, H. & McLaughlin, S. (eds) (2008). Next Generation Mobile Access Technologies:
Implementing TDD, Cambridge University Press, ISBN: 13:9780521826228.
Hassibi, B. & Hochwald, B. M. (2003). How Much Training is Needed in Multiple-Antenna
Wireless Links?, IEEE Transactions on Information Theory 49: 951963.
Heath, R. & Paulraj, A. (2001). Antenna Selection for Spatial Multiplexing Systems Based on
Minimum Error Rate, Proc. of the International Conference on Communications (ICC 01),
Vol. 7, Helsinki, Finland, pp. 22762280.
Hochwald, B., Peel, C. & Swindlehurst, A. (2005). A Vector-Perturbation Technique for
Near-Capacity Multiantenna Multiuser Communication. part II: Perturbation, IEEE
Transactions on Communications 53(3): 537544.
Koutsimanis, C. & Fodor, G. (2008). A Dynamic Resource Allocation Scheme for Guaranteed
Bit Rate Services in OFDMA Networks, Proc. of the IEEE International Conference on
Communications (ICC 08), pp. 2524 2530.
Kusume, K., Joham, M., Utschick, W. & Bauch, G. (2007). Cholesky factorization with
symmetric permutation applied to detecting and precoding spatially multiplexed
data streams, IEEE Transactions on Signal Processing 55(6): 3089 3103.
Learned, R., Willsky, A. & Boroson, D. (1997). Low complexity optimal joint detection for
oversaturated multiple access communications, IEEE Transactions on Signal Processing
45(1): 113123.
Love, D. J. & Heath, R. (2005). Limited Feedback Unitary Precoding for Spatial Multiplexing
Systems, IEEE Transactions on Information Theory 51(8): 29672976.
Love, D. J., Heath, R. & Strohmer, T. (2003). Grassmannian Beamforming for
Multiple-Input Multiple-Output Wireless Systems, Proc. of the International Conference
on Communications (ICC 03), Vol. 4, IEEE, pp. 26182622.
Love, D. J., Heath, R., Santipach, W. & Honig, M. L. (2004). What is the Value of Limited
Feedback for MIMO Channels, IEEE Communications Magazine .
Molisch, A., Win, M. & Winters, J. (2001). Capacity of MIMO Systems with Antenna Selection,
Proc. of the International Conference on Communications (ICC 01), Vol. 2, pp. 570574.
Molisch, A., Win, M. & Winters, J. (2003). Reduced-Complexity Transmit/Receive-Diversity
Systems, IEEE Transactions on Signal Processing 51(11): 27292738.
Mukkavilli, K., Sabharwal, A., Aazhang, B. & Erkip, E. (2002). Performance Limits on
Beamforming with Finite Rate Feedback for Multiple Antenna Systems, Proc. of the
36th Asilomar Conference on Signals, Systems and Computers, Vol. 1, pp. 536540.
Mukkavilli, K., Sabharwal, A., Erkip, E. & Aazhang, B. (2003). On Beamforming with Finite
Rate Feedback in Multiple-Antenna Systems, IEEE Transactions on Information Theory
49(10): 25622579.
Pan, Z., Wong, K.-K. & Ng, T.-S. (2004). Generalized Multiuser Orthogonal Space-Division
Multiplexing, IEEE Transactions on Wireless Communications 3(6): 19691973.
Popovic, B. (1992). Generalized chirp-like polyphase sequences with optimal correlation
properties, IEEE Transactions on Information Theory 38: 14061409.
Schubert, M. & Boche, H. (2004). Solution of the Multiuser Downlink Beamforming
Problem with Individual SINR Constraints, IEEE Transactions on Vehicular Technology
53(1): 1828.
Seidel, E. (2008). Progress on "LTE Advanced the New 4G Standard, Newsletter, NOMOR,
Munich, Germany.
Sesia, S., Touk, I. & Baker, M. (2009). LTE - The UMTS Long Term Evolution: From Theory to
Practice, Wiley.
Shen, Z., Chen, R., Andrews, J., Heath, R. & Evans, B. (2005). Low Complexity User Selection
Algorithms for Multiuser MIMOSystems with Block Diagonalization, Proc. of the 39th
Asilomar Conference on Signals, Systems and Computers., pp. 628632.
Shi, S., Schubert, M. & Boche, H. (2008). Downlink MMSE Transceiver Optimization for
Multiuser MIMO Systems: MMSE Balancing, IEEE Transactions on Signal Processing
56(8): 37023712.
Spencer, Q., Swindlehurst, A. & Haardt, M. (2004). Zero-Forcing Methods for Downlink
Spatial Multiplexing in Multiuser MIMO Channels, IEEE Transactions on Signal
Processing 52(2): 461471.
Telatar, E. (1999). Capacity of Multi-Antenna Gaussian Channels, European Transaction on
Telecommunication 10(6): 585595.
Vishwanath, S., Jindal, N. & Goldsmith, A. (2003). Duality, Achievable Rates, and Sum-Rate
Capacity of Gaussian MIMO Broadcast Channels, IEEE Transactions on Information
Theory 49(10): 26582668.
Viswanath, P., Tse, D. &Laroia, R. (2002). Opportunistic Beamforming Using Dumb Antennas,
Proc. of the International Symposium on Information Theory, IEEE, p. 449.
Wang, C. & Murch, R. (2005). Adaptive Cross-Layer Resource Allocation for Downlink
Multi-User MIMO Wireless System, Proc. of the 61st Vehicular Technology Conference
(VTC 05), Vol. 3, pp. 16281632.
Weingarten, H., Steinberg, Y. &Shamai, S. (2004). The Capacity Region of the Gaussian MIMO
Broadcast Channel, Proc. of the International Symposium on Information Theory (ISIT 04),
Chicago, USA, pp. 174182.
Windpassinger, C., Fischer, R., Vencel, T. & Huber, J. (2004). Precoding in Multiantenna
and Multiuser Communications, IEEE Transactions on Wireless Communications
3(4): 13051316.
Wong, K.-K., Murch, R. & Letaief, K. (2003). A Joint-Channel Diagonalization for Multiuser
MIMO Antenna Systems, IEEE Transactions on Wireless Communications 2(4): 773786.
Zhou, S., Wang, Z. & Giannakis, G. (2005). Quantifying the Power Loss When
Transmit Beamforming Relies on Finite-Rate Feedback, IEEE Transactions on Wireless
Zhou, Z., Dong, Y., Zhang, X., Wang, W. & Zhang, Y. (2004). A Novel Antenna Selection
Scheme in MIMO Systems, International Conference on Communications, Circuits and
Systems (ICCCAS 04), Vol. 1, pp. 190194.
Zhou, Z. & Vucetic, B. (2004). MIMO Systems with Adaptive Modulation, Proc. of the 59th
Vehicular Technology Conference (VTC 04), Vol. 2, pp. 765769.
Zhuang, H., Dai, L., Zhou, S. & Yao, Y. (2003). Low Complexity Per-Antenna Rate and Power
Control Approach for Closed-Loop V-BLAST, IEEE Transactions on Communications
51(11): 17831787.
Xiaolin Hou and Hidetoshi Kayama
DOCOMO Beijing Communications Laboratories Co., Ltd.
China
1. Introduction
The merits of 3GPP long term evolution (LTE), such as high spectral efciency, very low
latency, support of variable bandwidth, simple architecture, etc, make it the most competitive
candidate for the next generation mobile communications standard. In the rst release of LTE,
only single transmit antenna is supported in the uplink due to its simplicity and acceptable
performance. However, in order to keep its current leading position, LTE needs further
evolvement (known as LTE-Advanced (LTE-A)) to provide better performance, including a
higher uplink spectrum efciency. Therefore, multiple transmit antennas must be supported
in the LTE-A uplink and one important issue is the uplink demodulation reference signal
(DMRS) design, which will inuence uplink channel estimation accuracy and eventually
determine uplink reliability and throughput.
In this study we rst briey review the current status of DMRS in LTE uplink and then
different DMRS enhancement schemes are investigated for LTE-A uplink multiple-input
multiple-output (MIMO) transmission. Also, two-dimensional channel estimation algorithms
are provided to realize accurate uplink channel estimation. With computer simulations,
the performances of several candidate LTE-A uplink DMRS design schemes are evaluated
and compared. Finally some basic conclusions are provided together with the latest
standardization progress.
2. DMRS in LTE uplink
LTE uplink is based on single-carrier frequency division multiple access (SC-FDMA) due
to its low peak-to-average power ratio (PAPR). There are two types of reference signal in
LTE uplink: DMRS used for data reception and sounding reference signal (SRS) used for
scheduling and link adaptation. In this study we only focus on DMRS design and related
channel estimation for the physical uplink shared channel (PUSCH).
Take frame structure type 1 for example, each LTE radio frame is 10ms long and consists of
20 slots of length 0.5ms. A subframe is dened as two consecutive slots. For the normal cyclic
prex (CP) case, each slot contains 7 symbols. The two-dimensional time-frequency resource
can be partitioned as resource blocks (RBs) and each RB corresponds to one slot in the time
domain and 180 kHz in the frequency domain. In LTE uplink, the DMRS for PUSCH in the
frequency domain will be mapped to the same set of physical resource blocks (PRBs) used
for the corresponding PUSCH transmission with the same length expressed by the number of

Demodulation Reference Signal Design and
Channel Estimation for LTE-Advanced Uplink
22
subcarriers, while in the time domain DMRS will occupy the 4th SC-FDMA symbol in each
slot for the normal CP case, as shown in Fig. 1.
Fig. 1. DMRS in LTE uplink
In order to support a large number of user equipments (UEs) in multiple cells, a large number
of different DMRS sequences are needed. A DMRS sequence r
()
u,v
(n) is dened by a cyclic shift
(CS) of a base sequence r
u,v
(n) according to
r
()
u,v
(n) = e
jn
r
u,v
(n), 0 n < M
RS
sc
(1)
where M
RS
sc
= mN
RB
sc
is the length of DMRS sequence, m is the RB number and N
RB
sc
is the
subcarrier number within each RB. When the subcarrier bandwidth is set as 15kHz, each RB
will contain 12 subcarriers, i.e., N
RB
sc
= 12. Multiple DMRS sequences can be derived from a
single base sequence through different values of .
The denition of the base sequence depends on the sequence length. For M
RS
sc
3N
RB
sc
, the
base sequence is dened as the cyclic extension of the Zadoff-Chu sequence (Chu, 1972)
r
u,v
(n) = x
q
(nmodN
RS
ZC
), 0 n < M
RS
sc
(2)
x
q
(m) = e
j
qm(m+1)
N
RS
ZC
, 0 m < N
RS
ZC
1 (3)
where x
q
(m) is the q
th
root Zadoff-Chu sequence and N
RS
ZC
is the length of Zadoff-Chu
sequence that is given by the largest prime number such that N
RS
ZC
< M
RS
sc
. For M
RS
sc
<
3N
RB
sc
, the base sequence is dened as the computer generated constant amplitude zero
autocorrelation (CG-CAZAC) sequence
r
u,v
(n) = e
j (n)/4
, 0 n < M
RS
sc
(4)
where the values of (n) are given in (3GPP, TS 36.211).
Base sequences r
u,v
(n) are divided into 30 groups with u {0, 1, ..., 29}. Each group contains
one base sequence (v = 0) with 1 m 5 and two base sequences (v = 0, 1) with 6
m N
max,UL
RB
, where N
max,UL
RB
is the maximum RB number in the uplink. In order to reduce
inter-cell interference (ICI), neighboring cells should select DMRS sequences from different
base sequence groups. Furthermore, there are 3 kinds of hopping dened for the DMRS in LTE
uplink, i.e., group hopping, sequence hopping and CS hopping, where CS hopping should
always be enabled in each slot.
The CS value in a slot is given by = 2n
cs
/12 with
n
cs
= (n
(1)
DMRS
+ n
(2)
DMRS
+ n
PRS
)/12 (5)
where n
(1)
DMRS
is a broadcast value, n
(2)
DMRS
is included in the uplink scheduling assignment
and n
PRS
is given by a cell-specic pseudo-random sequence. Obviously, there are 12 usable
CS values in total for DMRS in LTE uplink.
3. DMRS design and channel estimation for LTE-A uplink
3.1 DMRS enhancement
Current LTE uplink DMRS only considers UE with single transmit antenna. However, in order
to boost the uplink spectrum efciency, multiple transmit antennas must be supported in
LTE-A uplink. Therefore, the uplink DMRS must be enhanced for MIMO transmission and
each UE nowmay have multiple DMRS sequences, depending on its transmit antenna number
(without precoding) or spatial layer number (with precoding).
There are several possible solutions, including CS extension, orthogonal cover code (OCC),
interleaved frequency division multiplexing (IFDM) and their combinations. Considering the
backwards compatibility with LTE and the low PAPR requirement for uplink transmission,
IFDM should be excluded rst. Then CS, OCC and their combinations are promising
candidates for DMRS enhancement and will be discussed in more details in the following
text.
3.1.1 Baseline: CS extension
Considering the backwards compatibility, it is agreedthat cyclic shift separation is the baseline
for the LTE-A uplink DMRS enhancement (3GPP, TR 36.814). Without loss of generality,
uplink precoding is not considered in the following text, therefore, transmit antenna
and spatial layer are equivalent and interchangeable. For single-user MIMO (SU-MIMO)
transmission with n
T
2 spatial layers, it is natural to assign multiple CS values to separate
the multiple spatial layers. Then the questions remained to be answered are how to assign
different CS values to different spatial layers and how to ensure the backwards compatibility
to LTE.
If we assign multiple CS values with the following constraint
n
cs,i
= (n
cs,0
+
C
n
T
i)mod(C), i = 0, 1, ..., n
T
1 (6)
where n
cs,i
corresponds to the CS value of DMRS for the ith spatial layer and C is the constant
value 12 for PUSCH. Then the CS value of DMRS for the rst spatial layer
0
= 2n
cs,0
/12 is
exactly the same as that for the single transmit antenna case in LTE. Therefore, all the original
CS signaling and hopping designs for the single transmit antenna UE in LTE can be kept
unchanged for the multiple transmit antennas UE in LTE-A, once the constraint in Eq. (6) is
satised.
Because this DMRS design can be viewed as binding together the CS values of DMRS as well
as the channel impulse response (CIR) positions of different spatial layers with the maximum
distance constraint, as illustrated in Fig. 2 (Note that the relationship between
i
and
0
will
keep unchanged during CS hopping), we simply call it maximum distance binding (MDB).Its
benets include:
First, the distance between CIRs of different spatial layers in the time domain can be
always maximized, thus the interference between DMRS of different spatial layers can be
minimized;
Second, no additional signaling is required for CS notication and hopping when support
uplink MIMO transmission, therefore, it is completely backwards compatible to LTE;
Third, it can support time-domain inter-slot interpolation that is necessary for moderate to
high mobility cases.
419 Demodulation Reference Signal Design and Channel Estimation for LTE-Advanced Uplink
Fig. 2. CS extension with MDB
Actually, the same DMRS design principle can also be applied to the uplink multi-user MIMO
(MU-MIMO) transmission with single transmit antenna UEs. Now it only requires some
constraint in the uplink scheduling assignment for the CS values of multiple DMRS (because
n
(1)
DMRS
and n
PRS
are the same for all the UEs in the same cell, respectively) as follows
n
(2)
DMRS,i
= (n
(2)
DMRS,0
+
C
n
T
i)mod(C), i = 0, 1, ..., n
T
1 (7)
where n
(2)
DMRS,0
is the scheduled value for the rst UE.
In order to support the above CS scheduling constraint for MU-MIMO transmission, we have
two possible options:
Option 1: No signaling modication
Because the current LTE specication only supports 8 possible values for n
(2)
DMRS
(3GPP,
TS 36.211), a limited number of combinations can be chosen in the uplink scheduling
with the MDB constraint (7) satised. Therefore, for the 2-user case, n
(2)
DMRS,i

{(0, 6), (2, 8), (3, 9), (4, 10)}; while for the 4-user case, n
(2)
DMRS,i
{(0, 3, 6, 9)}.
Option 2: Slight signaling modication
If the specic eld in downlink control information (DCI) format 0 for the CS of DMRS can
be increased from 3 bits to 4 bits, all the possible combinations in the CS scheduling for
MU-MIMO transmission can be supported with the MDB constraint (7) satised.
3.1.2 Further enhancement: CS + OCC
For high-order SU-MIMO, MU-MIMO and coordinated multi-point (CoMP) reception that
will be supported in the further evolvement of LTE, the number of superposed spatial layers
will increase to four or even eight. In order to reduce the interference between multiple spatial
layers, OCC, such as [+1, +1] and [+1, 1], can be further introduced across the two DMRS
symbols within the same subframe.
For MU-MIMO and CoMP reception, CS + OCC can provide some special advantage
compared to CS only scheme, such as capability to multiplex UEs with different transmit
bandwidths and robustness to timing difference of multiple UEs. For SU-MIMO, CS + OCC
may also be attractive for high-order MIMOtransmission and/or high-order modulation. The
combination of CS and OCC could have two variations, i.e., CS + OCC with identical CS and
CS + OCCwith offset CS (TI, 2009), as illustrated in Fig. 3 (a) and Fig. 3 (b), respectively, taking
four spatial layers for example.
(a) CS + OCC
(b) CS + OCC (offset)
Fig. 3. Combination of CS and OCC
However, OCC will lose its effectiveness in some cases, such as when the mobility
increases from low to moderate or PUSCH hopping happens within one subframe. In the
aforementioned situations, CS + OCC with identical CS, abbr. as CS + OCC, cannot work
at all; while CS + OCC with offset CS , abbr. as CS + OCC (offset), still can work, but in
essence only CS takes effect now. Obviously, CS + OCC (offset) occupies twice CS resources
compared to CS + OCC. Meanwhile, to introduce OCC into LTE-Auplink DMRS design, some
additional control signaling may be needed. Otherwise, the linkage between OCC and CS
must be dened to avoid increasing control signaling, i.e., the notication of OCC could be
realized in an explicit way.
3.2 Two-dimensional channel estimation
In order to obtain the time-frequency two-dimensional channel state information (CSI) in
the SC-FDMA uplink, two-dimensional channel estimation is needed for each subframe.
Without loss of generality, assume that the inter-symbol interference (ISI) and the inter-carrier
interference (ICI ) are small and neglectable. Therefore, for PDSCH and corresponding DMRS
within one subframe, the received signal at the k-th subcarrier in the l-th SC-FDMA symbol
can be written as
Y(k, l) = H(k, l) X(k, l) + N(k, l), k
0
k < k
0
+12 N
UL
RB
1, 0 l < 14 (8)
where X(k, l), H(k, l) and N(k, l) denote the transmitted signal, the channel frequency
response (CFR) and the additive white Gaussian noise (AWGN) with zero mean and variance
2
for the k-th subcarrier in the l-th SC-FDMAsymbol, respectively. k
0
is the frequency starting
position of PUSCH and N
UL
RB
is the uplink RB number for PUSCH. For the multipath wireless
channel within one SC-FDMA symbol, the CFR can be related to the CIR as
H(k, l) =
G1
g=0
h(g, l) e
j2kg/K
(9)
where h(g, l) is the g-th multipath component and G is the sample number corresponding to
the maximum multipath delay.
The rst step of two-dimensional channel estimation is to obtain the initial estimated
superposed channels within the two DMRS symbols, i.e.,

H(k, 3) and

H(k, 10), as follows
H(k, l) = Y(k, l) conj(r

(
0
)
u,v
(k)), l = 3, 10 (10)
where conj() represents the complex conjugate and without loss of generality, PUSCH and
DMRS hopping are not considered.
To facilitate the following description, dene the nal estimated channel as

H(k, l). Then the
target of two-dimensional channel estimation is to derive each data symbols

H(k, l) from
H(k, 3) and

H(k, 10). Taking the implementation complexity into account, two concatenated
one-dimensional channel estimation, i.e., frequency-dimensional channel estimation and
time-dimensional channel estimation, will be considered in this study.
3.2.1 Frequency-dimensional channel estimation
For frequency-dimensional channel estimation, discrete-time Fourier transform (DFT) based
channel estimation (Edfors et al., 2000) could be utilized. However, because the RB allocation
to a given UE is generally only a small portion of the overall uplink bandwidth, the CIR energy
leakage will be observed in practice, as shown in Fig. 4, where the rst and the second rows
are for two-antenna and four-antenna cases, while the left and the right columns are for RB#
= 1 and RB# = 10 cases, respectively. Its obvious that the smaller the RB number, the more
severe the CIR energy leakage. This phenomenon will make the CIRs from different transmit
antennas superposed together and difcult to be separated with each other, especially when
the transmit antenna number becomes larger. Furthermore, the frequency domain Gibbs
phenomenon (Oppenheimet al., 1999) will appear at the edges of assignedconsecutive RBs for
a given UE due to the signal discontinuities. Therefore, the estimation accuracy of traditional
DFT-based channel estimation will deteriorate signicantly in practice.
In order to mitigate the aforementioned problems, an improvedDFT-basedchannel estimation
was proposed for LTE(-A) uplink (Hou et al., 2009), which is illustrated in Fig. 5 for each
receive antenna of eNB. After serial-to-parallel (S/P) conversion and K-point fast Fourier
transform (FFT), the received signal is transformed into the frequency domain. Because each
UE (except for UEs in the same MU-MIMOtransmission) occupies different RBs in the uplink,
we can rst separate different UEs by way of frequency division multiplexing (FDM). Then
taking channel estimation for UE1 for example, multiply the separated received DMRS by
Fig. 4. CIR energy leakage
the complex conjugate of the DMRS sequence assigned for the 1st spatial layer and perform
K-point inverse FFT (IFFT) to get the superposed CIRs in the time domain, i.e.,
h(g, l) = i f f t(

H(k, l)), 0 k < K, l = 3, 10 (11)
After the operation of dynamic CIR reservation (DCIR
2
), we can seperate the CIRs for different
spatial layers by way of different CS values. As for the operation of DCIR
2
, the dynamically
reserved CIR for each spatial layer consists of 2 parts with respect to the timing positions, i.e.,
(
C
n
T
i) K, i = 0, 1, ..., n
T
1:
Right part
There are CP samples preserved with the following right boundary
(
C
n
T
i) K + CP 1, i = 0, 1, ..., n
T
1 (12)
where CP is the cyclic prex length of the SC-FDMA symbol and is an adjustable
parameter (0 < 1) that can be optimized in practical implementations.
Left part
There are samples preserved with the following left boundary
[(
C
n
T
i) K + K]mod(K), i = 0, 1, ..., n
T
1 (13)
where is the main lobe width of CIR energy leakage ( =
K
12RB#
) and is an adjustable
parameter (0 <
K/n
T
CP
) that can be optimized in practical implementations. In

order to simply the adjustment, we can dene

=
K
12
and =

RB#
, therefore,

becomes
a constant and only should be adjusted.
The proper choices of and are mainly determined by the noise level, the multipath delay
prole and the assigned RB number for a given UE. And afer DCIR
2
, we can obtain the CIR
for the i-th spatial layer as

h
i
(g, l).
Finally, the frequency-dimensional channel estimation result of DMRS symbols for the i-th
spatial layer can be achieved by K-point FFT and provided to the following time-dimentional
channel estimation block.
H
i
(k, l) = i f f t(
h
i
(g, l)), 0 g < K, l = 3, 10 (14)
Another point should be emphasized is the operation of frequency domain
windowing/dewindowing . Due to the frequency domain Gibbs phenomenon caused
by the discontinuities at the edges of assigned consecutive RBs for a given UE, the overall
channel estimation accuracy will be degraded, especially at the edges of assigned consecutive
RBs. Therefore, some frequency domain window, such as Hanning window, Hamming
window, Blackman window, etc. (Oppenheim et al., 1999), can be further added (see the
dashed-line blocks in Fig. 5) to improve the channel estimation accuracy with some additional
complexity. For example, Blackman window will be adopted in our following computer
simulations.
w(n) = 0.42 0.5cos(2n/M) +0.08cos(4n/M) (15)
where M is the window length and 0 n M. In order not to eliminate the useful signals
within the assigned RBs, the window length should be larger than the assigned bandwidth
(12 RB#) for the corresponding UE.
Note that the improved DFT-based channel estimation can be applied to not only LTE-A
MIMO uplink, but also LTE single-input single-output (SISO) or single-input multiple-output
(SIMO) uplink.
3.2.2 Time-dimensional channel estimation
After frequency-dimensional channel estimation, we only obtain channel estimation results
for two DMRS symbols within each subframe. In order to further acquire channel estimation
result for each data symbol, time-dimensional channel estimation is needed, i.e., inter-slot
interpolation via two DMRS symbols within each subframe. Two practical schemes are
time-dimensional linear interpolation (TD-LI) and time-dimensional average or despreading
(TD-Average/Despreading), i.e.,
H(k, l) = c
l

H(k, 3) + (1 c
l
)

H(k, 10), 0 l < 14 (16)
c
l
= (10 l)/7 , TD LI
c
l
= 1/2 , TD Average
(17)
Fig. 5. The improved DFT-based channel estimation
It should be noted that for the case of CS + OCC with identical CS, TD-Despreading must be
carried out before frequency-dimensional channel estimation.
4. Performance evaluation
Computer simulation results, including both block error rate (BLER) and throughput
performances, will be provided in this section to compare different DMRS design schemes,
i.e., CS only, CS + OCC and CS + OCC (offset).
The simulation parameters are listed in Table 1. Notice that the FFT size is larger than
the usable subcarrier number because of the existence of guard band. There are totally
50 RBs in the uplink and we consider two RB# allocation cases with RB# = 4, 10,
respectively. Furthermore, 2 typical MIMO congurations, i.e., 2 2 and 4 4, are both
simulated. The MIMOtransmission scheme is spatial multiplexing without precoding and the
MIMO detection scheme is minimum mean square error (MMSE) detection. Without loss of
generality, synchronization error and PUSCH hopping are not considered. For the improved
DFT-based frequency-dimensional channel estimation, we simply chose = 0.5 and = 0.2
and the frequency domain window length is set to be M = 1.1 RB# 12. The channel model
is selected as typical urban (TU) with mobility of 3km/h or 30km/h.
First, BLER performances of different DMRS design schemes will be compared. The
same frequency-dimensional channel estimation is utilized for different DMRS design
schemes and time-dimensional channel estimation could be different, i.e., CS + OCC
can only use TD-Despreading, while CS only and CS + OCC (offset) can use TD-LI or
TD-Average/Despreading. Furthermore, the curve with perfect CSI is also provided in
each gure for comparison. The BLER performances are evaluated with two representative
congurations, i.e., 10RB with 16QAM and 4RB with 64QAM (the coding rate is 1/2), for
2 2 and 4 4 MIMO, respectively.
When the mobile speed is as lowas 3km/h, Fig. 6 shows that for the 2 2 MIMOcase different
DMRS design schemes have almost the same BLER performance. The only difference comes
from time-directional channel estimation, i.e., TD-Average/Despreading can achieve slightly
better performance than TD-LI due to the noise averaging effect in low mobility cases. And
fromFig. 7, it can be observed that for the 4 4 MIMOcase the introduction of OCC is helpful
to improve the BLER performance in low mobility cases, especially when the RB number is
small and/or the modulation order is high.
Parameters Values
Carrier frequency 2GHz
Bandwidth 10MHz
FFT size 1024
Usable subcarrier # 600
Cyclic prex 72
Assigned RB # 4, 10
MIMO conguration 2 2
(Spatial multiplexing) 4 4
MIMO detection MMSE
Modulation QPSK,16QAM,64QAM
Channel coding
Turbo
(Coding rate 1/2, 2/3, 3/4)
Synchronization Perfect
PUSCH hopping Disabled
DMRS design
CS Only
CS + OCC
CS + OCC (offset)
Frequency-dimensional Improved DFT-based
channel estimation
Time-dimensional TD-LI
channel estimation TD-Average/TD-Despreading
Channel model Typical Urban (TU)
Mobile speed 3km/h, 30km/h
Table 1. Simulation parameters
However, if the mobile speed increases fromlowto moderate, i.e., from3km/h to 30km/h, the
aforementioned conclusion has to be revised. As shown in Fig. 8, now CS only with TD-LI can
achieve the best performance and OCC will lose its effectiveness. The reason behind is that
when the mobile speed is as high as 30km/h, the wireless channels between two consecutive
slots are relatively fast time-varying, which makes TD-Average/Despreading cannot work
well. On the other hand, TD-LI can still track the time-varying channel effectively. Therefore,
from the mobility point of view, OCC has its apparent limitation, i.e., OCC will mainly work
in the low mobility cases. However, considering the major application scenario of MIMO is
the low mobility environment, OCC is still attractive for DMRS enhancement. And in the
following simulations only 3km/h is considered.
In order to provide a more comprehensive comparison between different DMRS design
schemes, the throughput performances with different modulation level and coding rate are
provided in Fig. 9 and Fig. 10 for 2 2 and 4 4 MIMO, respectively. Three different
modulation schemes (QPSK, 16QAM, 64QAM) and three different coding rates (1/2, 2/3, 3/4)
are simulated, so in total there are nine combinations of modulation and coding. Considering
the mobile speed is low, TD-Average/Despreading will be adopted instead of TD-LI. Also
under this situation, because CS + OCC and CS + OCC (Offset) have neglectable performance
difference, only CS + OCC is simulated, together with CS only and perfect CSI. Therefore,
in each gure there are 27 curves, shown by different line styles and markers. For each
DMRS design scheme, only the envelop of nine curves (each with one specic combination
of modulation and coding) is highlighted to show the highest achievable throughput, which
5 10 15 20 25
10
3
10
2
10
1
10
0
E
s
/N
0
(dB)
B
L
E
R
Perfect CSI
CS Only(TDAverage)
CS Only(TDLI)
CS+OCC
CS+OCC(ofset)(TDDespreading)
CS+OCC(ofset)(TDLI)
(a) 10RB, 16QAM
5 10 15 20 25
10
3
10
2
10
1
10
0
E
s
/N
0
(dB)
B
L
E
R
Perfect CSI
CS Only(TDAverage)
CS Only(TDLI)
CS+OCC
CS+OCC(ofset)(TDLI)
(b) 4RB, 64QAM
Fig. 6. BLER performance (2 2 MIMO, TU, 3km/h)
5 10 15 20 25
10
3
10
2
10
1
10
0
E
s
/N
0
(dB)
B
L
E
R
Perfect CSI
CS Only(TDAverage)
CS Only(TDLI)
CS+OCC
CS+OCC(ofset)(TDLI)
(a) 10RB, 16QAM
15 20 25 30 35
10
3
10
2
10
1
10
0
E
s
/N
0
(dB)
B
L
E
R
Perfect CSI
CS Only(TDAverage)
CS Only(TDLI)
CS+OCC
CS+OCC(ofset)(TDLI)
(b) 4RB, 64QAM
Fig. 7. BLER performance (4 4 MIMO, TU, 3km/h)
5 10 15 20 25
10
3
10
2
10
1
10
0
E
s
/N
0
(dB)
B
L
E
R
Perfect CSI
CS Only(TDAverage)
CS Only(TDLI)
CS+OCC
CS+OCC(ofset)(TDLI)
Fig. 8. BLER performance (4 4 MIMO, 10RB, 16QAM, TU, 30km/h)
5 0 5 10 15 20 25 30
0
1
2
3
4
5
E
s
/N
0
(dB)
T
h
r
o
u
g
h
p
u
t
(
M
b
p
s
)
Perfect CSI
CS Only
CS+OCC
(a) 4RB
5 0 5 10 15 20 25 30
0
2
4
6
8
10
12
14
E
s
/N
0
(dB)
T
h
r
o
u
g
h
p
u
t
(
M
b
p
s
)
CS Only
CS+OCC
Perfect CSI
(b) 10RB
Fig. 9. Throughput performance (2 2 MIMO, TU, 3km/h)
5 0 5 10 15 20 25 30
0
1
2
3
4
5
6
7
8
9
10
E
s
/N
0
(dB)
T
h
r
o
u
g
h
p
u
t
(
M
b
p
s
)
CS Only
Perfect CSI
CS+OCC
(a) 4RB
5 0 5 10 15 20 25 30
0
5
10
15
20
25
E
s
/N
0
(dB)
T
h
r
o
u
g
h
p
u
t
(
M
b
p
s
)
Perfect CSI
CS Only
CS+OCC
(b) 10RB
Fig. 10. Throughput performance (4 4 MIMO, TU, 3km/h)
could be viewed as the throughput performance with the ideal adaptive modulation and
coding. From Fig. 9 and Fig. 10, we can observe that the introduction of OCC can improve
throughput to some extent when some of the following situations are satised, i.e., the antenna
number is large, the RB number is small, and the sigal-to-noise ratio is high.
5. Some basic conclusions and standardization progress
DMRS enhancement is a key step to support MIMO transmission, including SU-MIMO,
MU-MIMO and CoMP, for LTE-A uplink. In this study different DMRS design schemes as
well as channel estimation are investigated. In addition to the baseline of CS extension, the
further combination of CS with OCC is also discussed. In addition to the special advantage
for MU-MIMO and CoMP, CS + OCC is also attractive for high-order SU-MIMO to further
suppress the interferences among the increasing multiple spatial layers. With the enhanced
DMRS design and improved channel estimation, a higher spectrum efciency can be realized
in LTE-A uplink. Meanwhile, considering the backwards compatibility, as less as possible
modication to the current LTE specication is preferred.
In the recent 3GPP RAN1 meetings, it was agreed that (3GPP, R1-102601)
Introduce the OCC in Rel-10 without increasing uplink grant signaling overhead
OCC can be used for both SU-MIMO and MU-MIMO
More design details about DMRS enhancement, such as CS and OCC linkage, DMRS hopping,
etc., are still under discussions and hopefully the uplink DMRS enhancement for LTE-A will
be nalized by the end of 2010.
6. References
3GPP. R1-102601: Final report of 3GPP TSG RAN WG1 #60bis, 3GPP TSG RAN WG1 Meeting
#61, Montreal, Canada, May 10-14, 2010.
3GPP TR 36.814 http://www.3gpp.org/ftp/specs/html-info/36814.htm
3GPP TS 36.211 http://www.3gpp.org/ftp/specs/html-info/36211.htm
Chu, D. C. (1972). Polyphase codes with good periodic correlation properties. IEEE Trans. Info.
Theory, Vol. 18, No. 4, July 1972, pp. 531-532, ISSN 0018-9448
Edfors, O.; Sandell, M.; van de Beek, J. J.; Wilson, S. K. & Borjesson, P. O. (2000). Analysis of
DFT-based channel estimators for OFDM. Wireless Pers. Commun., Vol. 12, No. 1, Jan.
2000, pp. 55-70, ISSN 0929-6212
Hou, X.; Zhang, Z. & Kayama, H. (2009). DMRS design and channel estimation for
LTE-Advanced MIMO uplink, Proc. IEEE VTC09-Fall, Alaska, USA, Sept. 20-23, 2009.
Oppenheim, A. V.; Schafer, R. W. & Buck, J. R. (1999). Discrete-Time Signal Processing, 2nd ed.,
Prentice Hall, ISBN 013216292X, New Jersey
Texas Instruments. R1-091843: Discussion on UL DM RS for SU-MIMO, 3GPP TSG RAN WG1
Meeting #57, San Francisco, USA, May 4-8, 2009.

Advances in Vehicular Networking Technologies

Uploaded by

Copyright:

Available Formats

Advances in Vehicular Networking Technologies

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advances in Vehicular Networking Technologies

Uploaded by

Copyright:

Available Formats

ADVANCES IN

, (ii) the ingress time t

, and (c) 35 [m/s] v =

by applying a Data Rate Reduction (DRR) factor (i.e.,

= for messages propagating multi-hop within a cluster in the S direction at

= 0 dB, so that we have

takes values around 0.5 independently from the

= 0 dB, the algorithm of power

0 dB. Actually, having

can be derived from

> 0 dB, having respectively, C

= 40 dB are given on Figures 12 and 13.

U619, June 2005.

U133, September 2004.

RTT is the location estimator of RTT measurements,

RTT. The error term

), and the MLE of the scale-W

are calculated from the surface

. Thus, at low SNRs, where the error rates are high

is the mean excess delay.

= denotes as the source signal, which is constructed by

represent the covariance matrix, so we can get

is the general inverse matrix of covariance matrixC

, and C is the variance matrix.

. On this basis, the matrix constructed by variance and covariance

. So we can get the

. The proof is the end.

, which is the initial separation matrix.

, which averagely needs more iteration to search

is an eigenvector matrix when SINR takes the

( ) is the estimated value

= 0 for = k and , hence, the intracell CCI is completely eliminated (nulled).

, is an integer number several scheduling scenarios could result in the same

H(k, l) = Y(k, l) conj(r

) that can be optimized in practical implementations. In

You might also like