Wireless Sensor Networks: Protocols, Optimization and Applications

Wireless Sensor Networks: Protocols,
Optimization and Applications

Kin K. Leung
EEE and Computing Departments
Imperial College, London, U.K.
www.commsp.ee.ic.ac.uk/~kkleung/
2
Current Wireless Communications and Sensor Research
IP backhaul
Network
BS
WLAN
WiMax
Control
Server
Gateway
Access Point
Public
Internet
Application
(Web) Servers
Corporate
Computers
3G/4G
IEEE
Ad-Hoc
Sensors
MEMBRANE
MESSAGE
WINES/
SmartEN
PULSERS
ITA
M-VCE
MoD UDRC
Mobile-AWSN
Smart Infrastructure:
Wireless sensor network system for condition assessment and
monitoring of infrastructure
Kenichi Soga (PI), Engineering
Robert Mair, Engineering
Campbell Middleton, Engineering
Ian Wassell, Computer Lab
Frank Stajano, Computer Lab
Peter Bennett, Engineering
Nigel Graham (PI), Civil Engineering
Cedo Maksimovic, Civil Engineering
Kin K. Leung, Electrical Engineering
Yike Guo, Computing
Ivan Stoianov, Civil Engineering
Aging Engineering Infrastructure
Water Supply and Sewer Systems
Thames Water
31,000 km of pipelines
more than 100 yrs old, 1/3 more than
150 yrs old, ~30% leakage
Difficulties in implementing RTC with
conventional technologies
Tunnels
London Underground (LUL)
Tunnels 75 100 yrs old
Deterioration of linings
Minimal clearance to tunnel wall
Risks from 3rd party construction
Four of the UK's busiest road tunnels are
among the 10 most dangerous in Europe
(Blackwall Tunnel)
Bridges
Highway Agency/LUL/ Humber Bridge
~150,000 bridges in UK
Critical links in road/rail infrastructure
Deterioration
Many structures below required strength
Generic/Pervasive Sensor Networks
Major goal of this project: Generic/Pervasive sensor networks
Sharing of equipment for monitoring of multiple types of infrastructures
Exploit common characteristics of different infrastructures to advance sensor network design
Sensors
Low-power, low-cost
Reliable performance
Communications
Tiered structure and adaptive
network topology
Scalable protocol design
Efficient, secure and robust
Data analysis
Device, network
& service management
to Internet
Advantages of Wireless
Low-cost and fast deployment, especially in difficult-to-access areas
Scalable: Enable dynamic system growth and extension
Adaptive network configuration and operation in case of failure and
unexpected events, resulting into improved reliability
Take advantage of low-cost and low-power sensors
Two Small-scale Deployments as Proof-of-Concept
Scalability and adaptability
Cross-layer protocol design
Protocols linking WSN and Internet for management and control
Efficiency
Limited power supply
Harsh radio propagation environments
Tradeoffs between communication and computation
Security and reliability
Distributed network architecture with no single point of failure
Protection measures against attacks and for privacy
Low-power public key cryptography
Testing and deployment in real operating infrastructures
Not an easy task!
Asset owners have committed to provide assistance
Research Challenges for Large-Scale
Wireless Sensor Networks (WSN)
8
MAC Protocols: Monitoring Scenario
Assumptions
A single data sink
Multi-hop network
Small batteries
Relatively slow-changing wireless links
Globally time synchronization
Event-triggered reporting of large volumes of
data
Application: large infrastructure
Fracture detection using acoustic emissions
Wires of the main cable from suspension bridge
over Humber (Suspension) Bridge
Concrete and steel bridges and tunnels
Vibration monitoring in tunnels and bridges
A B C
D E F
Sink
9
In-network data aggregation
Assuming that data from neighboring nodes is correlated, thus can be
aggregated and compressed inside the network
Every node generally executes the following steps
Receive data from its neighbors
Aggregate received data with its own data
Forward compressed data towards the sink
We propose two protocols. Their respective objectives are to decide:
The route followed by the packets to be aggregated, which is a tree
The schedule for packet transmissions
A B C
D E F
Sink
1 2 3
4
TDMAframeconsistingoftransmissionslots
1 2 3 4 1 2 3 4 1 2 3 4
10
Fast Aggregation Tree (FAT) Protocol
Goal of FAT
Quickly construct a data aggregation tree in a duty-cycled network
Functioning
Radio transceivers of sensor nodes are turned on periodically with
period T
s
.
There is an offset of the schedules of nodes in different tiers
Key advantage
Time to construct the tree is divided by the number of tiers
Therefore, nodes can sleep for longer periods and save energy
11
FAT Performance
FATs tiered architecture restricts
possible parents, not optimal
Traversal time is the time to
transmit data, a measure of the
quality of the aggregation tree
SPT is the shortest path tree
The algorithm Centralized1 is only
good for high aggregation ability
FAT is relatively good across all
degrees of aggregation ability
12
Two MAC Protocols: RandSched and TBSP
Problems of the existing scheduling algorithms
Some of them are centralized
The obtained schedule may be infeasible
The k-hop interference model fails occasionally
The joint interference from multiple nodes may be infeasible
Our simulation results are in the table below
BFk neglects the interference caused more than k hops away
13
RandSched: Scheduling for data aggregation
Distributed scheduling protocol
Initialization phase
Testing phase
In CFi it is decided which nodes gain access to TFi
A node only gains a transmission slot if it has been proved that it can
tolerate other nodes interference
Data transmission phase
14
Properties of RandSched
Medium overhead, but scale well because RandSched is a
distributed protocol
12 slots per Contention Frame (CF) are sufficient to decide the
transmitters of a certain slot
This number of slots is independent of node density and network size
Shorter schedule than BFk lower latency and higher throughput
(See figure below)
M is the number of slots of the schedule
N is the number of nodes in the network
15
Test Based Scheduling Protocol (TBSP)
Differences with RandSched
Only supports uncompressed traffic (no data aggregation)
It is adaptive (it enables parts of the schedule to be recomputed
without affecting other nodes schedules)
Targeted applications
Periodic data gathering with slowly-varying traffic
Latency of 15 TDMA frames to acquire a slot can be tolerated
Advantage of TBSP over comparable protocols
Lower energy consumption (no need to monitor other nodes
schedules)
Lower probability of dismissing a neighbor as unreachable
16
Conclusions on MAC Protocols
FAT constructs an aggregation tree in a duty-cycled
environment quickly
RandSched produces a TDMA schedule for data aggregation
reliably
TBSP adapts a TDMA schedule for uncompressed traffic with
little power consumption
Uncompressed traffic is necessary in a preliminary data-collecting
stage in order to determine how data can be compressed
Opt imal Resource Allocat ion for Bat t ery
Limit ed Wireless Sensor Net works
Yun Hou ( I mperial College)
Kin K. Leung ( I mperial College)
Tom LaPort a ( PSU)
I TA Proj ect ( Sponsored by
U. S. Army & U. K. MoD)
-18-
Background
Wired networks
Max U over (Rate)
Wireless networks
Max U over (Rate, Power)
Wireless single radio networks
Max over (Rate, Power, Per-node Airtime)
Because: Mult iple flows served by t he
same node share t he airt ime [ 1]
NUM:
Max U(x)
x= resources
This new work:
Battery limited scenarios, how long a flow can be active is related to
transmission power
Flow duration added into as another optimization variable
Max U over (rate, power, airtime, flow-duration)
[1] Yun Hou, Kin K. Leung and Archan Misra, Enhancing Congestion Control with Adaptive Per-Node Airtime Allocation for
Wireless Sensor Networks, Proc. of IEEE PIMRC 2009, September 13-16, Tokyo, Japan.
-19-
Sensor networks battery limited
Current NUM objective function
U
f
= U (X
f
)
Utility as a function of flow rate only
But:
Large flow rates high transmission power battery runs out quickly!!
We introduce:
A new utility to consider both flow rate and duration
max U (X
f

f
)
A new energy constraint
s.t. P
n

f
E
n
Mot ivat ion
Flow duration
Residual energy
-20-
Syst em set t ings
A node can t r ansmi t f or one f l ow at
a t i me
Mul t i pl e f l ow s goi ng t hr ough t he
same node
Fl ow s ar e schedul ed
one by one
Single-radio Sensor:
Flow 1
Flow 2
Al l f l ow s shar e t he ai r - t i me of t he node
1,1
o
(n, f) link originated from node n on flow f
airtime fraction of flow f on node n
, n f
o
1,2
o
,
: ( )
1
n f
f n Path f
o
e
=
-21-
Syst em set t ings
Periodic scheduling, 1 slot per node throughout the period
A node divides its slot to transmit all flows using the same
power, then
Total number of slots a node can transmit:
Capacity C
n,f
for various flows at a same node are different
Amount of data (n, f) can send in one slot
transmission power of node n
number of time slots that flow f lasts
length of one time slot residual energy of node n
n
P
f
t
s
T
n
E
( )
n n s
E P T
, , s n f n f
T C o
New constraints
-22-
Problem formulat ion
{ }
, ,
( )
min
f n f n f
n Path f
X C o
e
s
and
( ) { }
( )
min
f n n s
n path f
E P T t
e
s
Two Constraints:
They are equivalent to:
, , f n f n f
X C o s and
( )
f n n s
E P T t s and ( ) f n path f e
( )
, , ,
max log
f s f f
X P
f f
U T X
o t
t
(
=

The final formulat ion:
, , f n f n f
X C o s
and
f n n s
P E T t s
and ( ) f n path f e
s.t.
total amount of
data transmitted
on flow f
proportional
fair among
flows
rate constraint duration constraint
f
Flow rate and duration are determined by the minimum values along the path
-23-
Concavit y/ convexit y analysis
( )
, , ,
max log
f s f f
X P
f f
U T X
o t
t
(
=

, , f n f n f
X C o s and
f n n s
P E T t s and ( ) f n path f e s.t.
log log log
s f f
f f f
T X t + +

concave
Proved in ACITA 09
convex
Geometric programming
' '
log log
f f n n
P P t t = =
' '
0
f n
P
n s
e E T
t +
s
convex
To show: objective function is concave and constraints are convex
Objective function:
Constraints:
-24-
The algorit hm ( Ts = 1)
1. update the shadow prices for flow rate and duration
2. update the transmission power
, , ,
( ) ( )
1
( 1) ( ) ( ) ( ) ( ) ( )
( )
n n P n f n f e n f f
f Flow n e n f Flow n
n
P t P t t M t t t
P t
o t
e = e
| |
+ = +
|
\ .

Forwarding nodes:
Source nodes:
1. Update the flow rate
,
( )
1
( 1)
( 1)
f
n f
n Path f
X t
t
e
+ =
+
2. Update the flow duration

,
( )
1
( 1)
( 1) ( 1)
f
n f n
n Path f
t
t P t
t
e
+ =
+ +
( )
, , , ,
( 1) ( ) ( ) ( )
n f n f n f n f f
t t C t X t
o
+
(
+ =

( )
, ,
( 1) ( ) ( ) ( )
n f n f n f n
t t E t P t
t
+
(
+ =

and
2. update the airtime fractions
, , , , ,
( 1) ( ) ( ) ( ) ( )
n
n f n f n f n f n e
e F
t t t t t o o o q q
e
+
(
| |
+ =
( |
|
(
\ .

-25-
0.0092
0.0162 0.0103 0.0125 0.0095
0.4444
0.6163
0.2773
0.0904
0.1724
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
1 2 3 4 5
Transmissionpower(Watt)
durationawareNUM traditional
Numerical result s
0.7519
0.4199
0.4711
2.4162
0.9114
0.6785
0
0.5
1
1.5
2
2.5
3
1 2 3
Flowratecomparison(bits/sec/Hz)
399.0531
61.6724
97.0156
11.2521
1.6226 3.6057
0
50
100
150
200
250
300
350
400
450
1 2 3
Flowdurationcomparison(sec)
Utility:
Duration-aware = 9.48
Traditional = 1.22
Transmitted bits:
Duration-aware = 371.66
Traditional = 31.20
Smaller
flow rate
Much longer
flow duration
Much smaller
TX power
-26-
A new resource allocat ion t o consider flow
durat ion t oget her wit h flow rat e
The problem is formulat ed wit h four variables
( rat e, power, airt ime- fract ion, durat ion)
Concavit y of t he problem has been proved and
a dist ribut ed algorit hm has been developed
Simulat ion result s show
When t ot al amount of dat a is t o be maximized, t he new
NUM framework gives t he opt imal solut ion
When energy is limit ed, t he new NUM t ends t o give very
small power allocat ion t o prolong flow durat ion
Conclusion on Net work Ut ilit y Maximizat ion
-27-
Combine cont inuous and discret e dist ribut ed
opt imizat ion
Cont inuous: NUM, rat es, power, air t ime, flow durat ion,
et c.
Discret e: t ransmission schedule ( MAC) , rout ing, dat a-
aggregat ion pat h, et c.
Net work coding
How t o t ake advant age of net work coding for efficient dat a
t ransfer and aggregat ion?
Physical- layer net work coding possible?
Transport prot ocols
Simple t ransport prot ocol for reliabilit y and in- net work
dat a aggregat ion
.
WSN issues for fut ure research

Wireless Sensor Networks: Protocols, Optimization and Applications

Uploaded by

Copyright:

Available Formats

Wireless Sensor Networks: Protocols, Optimization and Applications

Uploaded by

Document Information

Original Title

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

Wireless Sensor Networks: Protocols, Optimization and Applications

Uploaded by

Copyright:

Available Formats

Wireless Sensor Networks: Protocols,

Optimization and Applications

2. Update the flow duration

You might also like