Understanding Static Inter-Cell Interference

Coordination Mechanisms in LTE

Ashley Mills, David Lister, and Marina De Vos
AbstractThis work identies the factors which determine
the behaviour of static interference avoidance schemes: SINR
distribution shift, MCS mapping, and proportional MCS usage.
The work goes on to challenge the common assumption that it is
best to give resources with a high reuse factor to those at the
cell-edge, by showing for a xed rate service class, that it is best
to be greedy and give these resources to those at the cell-centre.
The work is performed using monte-carlo simulations, only in
the downlink direction, on a London scenario with realistic path
loss and network data. All work is statistically quantied using
appropriate tests.
Index TermsLTE, Interference Coordination, Soft Frequency
Reuse.
I. INTRODUCTION
T
HE next generation wireless technology, Long Term
Evolution (LTE), has been designed to deliver higher
spectral efciency and increased cell-edge throughputs relative
to HSPA [1]. It is expected that LTE will be deployed in a
reuse one conguration, in which all frequency resources are
available to use in each cell. Although LTE can operate at
SINRs as low as -6.5dB [2], concern still persists over cell-
edge performance.
This has led to the proposal of numerous inter-cell interfer-
ence coordination mechanisms. A large number of these are
dynamic in nature and usually assume communication between
basestations [3][15]. These schemes have tended toward
taking more and more cells into account, and it would appear
that the industry is converging toward multi-cell processing
with a centralised RAN architecture [16], [17].
Despite this progress and innovation, interest still persists
in static schemes that it is assumed can be deployed within
LTE without modication of the extant standards and without
signicant modication of extant equipment.
Static schemes usually fall into one of three broad cate-
gories: traditional hard frequency reuse, soft frequency reuse
[18], and partial frequency reuse [19]. Notwithstanding vari-
ants and other techniques that do not t the classication, this
taxonomy will serve the argument advanced here.
A. Soft Frequency Reuse
Soft Frequency Reuse was proposed by Huawei in [18];
supplemented in [20]. This proposal is effectively reiterated by
Ashley Mills and David Lister are with Group R&D Technology, Vodafone
Group Services Limited, The Connection, Newbury, Berkshire, RG142FN,
UK
Marina De Vos is with The University of Bath, UK
Manuscript received April 14, 2011; revised May 1, 2011; accepted June
30, 2011
Ericsson in [21] and by LG Electronics in [19] although the
latter augments the description with a priority based frequency
planning scheme. Alcatel propose a method very similar to soft
reuse in [22], albeit with a reuse factor higher than three at
the cell edge. Semi-static variants of soft reuse are proposed
in [23]. Soft frequency reuse is usually portrayed as depicted
in Figure 1.
Frequency
Power
Power
Power
A
B C
B
C
A
A
B C
A
B C
A
B C
A
B C
A
B C
A
B C
Fig. 1. Soft frequency reuse as conventionally presented
The left side of Figure 1 shows a hexagonal conguration
of cells, colour coded and labelled according to a tessellating
pattern to indicate which parts of the frequency band are
allowed in each part of each cell. The right of the gure shows
the frequency-power transmit proles for each of the three
types of cell that arise.
The general concept is that BSs transmit at reduced power
over the whole transmission band, to create spatially separated
cell centres that do not interfere with each other. At the cell
edges, a boosted reuse three pattern is used so that received
signals are orthogonal between otherwise interfering cell-edge
UEs.
The mean cell throughput under soft reuse for the downlink
is examined in [24]. It is claimed that 5th percentile throughput
can be improved relative to reuse one in trade off for a
reduction in total cell throughput by applying soft reuse. Since
the work only examines the mean cell throughput, it provides
no insight into the behaviour of more realistic schedulers.
Partial reuse differs from soft reuse in that the tessellated
part of the spectrum is kept disjoint from the reused part of
the spectrum [19].
B. Conicting Results
Examining results on soft and partial reuse [14], [18], [22],
[25][30] reveals some conicting statements.
312 JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011
2011 ACADEMY PUBLISHER
doi:10.4304/jcm.6.4.312-318
For example both [26] and [30] claim that partial reuse,
relative to reuse one, gives improvements in throughput at
the 5th percentile point, yet [29] concludes that the basic
partition-reuse scheme studied was not capable of improving
the rate at the 5% CDF point.
And [27] claims that soft reuse provides gains in both cell-
edge and total throughput when compared with reuse one
yet [28] concludes that With the expected link performance
no improvement can be found with static downlink reuse
schemes.. Furthermore [26] shows completely the opposite:
losses in both cell-edge and total throughput for some scenar-
ios.
The discussions in [29] and [28] go some way to explain,
for their own results in isolation from others, why they turn
out the way they do, yet no general explanation is proffered.
We contribute to this body of work in two ways: rstly
we statistically quantify our results to provide condence in
them, something that none of the cited works do, and secondly:
we explain clearly how different results can manifest from the
application of the same or very similar schemes by identifying
the principal factors involved and explaining their interactions
(Section IV).
C. Challenging a common assumption
A common assumption in the works cited above, is that
it is better to give the resources with a higher reuse factor
to the UEs at the cell-edge. This is evidenced by the
observation that none of the work suggests doing the opposite.
And although in [31] a convincing mathematical argument is
advanced as to why the cell edge may benet more from inter-
ference coordination than the cell-centre, this says nothing
of the trade-off in general.
Without strong empirical support, it is far from clear that
giving the better resource to the cell-edge UEs is always
the best scheduling strategy. And it must be observed that
subbands with higher reuse factors offer improved SINR to
all UEs, not just cell-edge UEs. So it isnt clear apriori what
the best scheduling strategy is for a given performance metric.
Against this backdrop we decided to examine scheduling
strategies that favour high SINR UEs even when a soft reuse
scheme has been applied, and were surprised to nd, contrary
to intuition, that a net gain in number of satised UEs could be
obtained. This is explained with reference to the determinant
factors identied in Section IV.
D. Document outline
The rest of this document is organised as follows. In
Section II, the soft reuse terminology used here is dened.
Experimental assumptions are explained in Section III. In
Section IV, the factors complicit in causing static reuse results
to differ are drawn out and explained through the medium of
mean-rate experiments. In Section V, a feasible scenario is
examined where favouring the cell-centre UEs gives a better
outcome than favouring the cell-edge UEs. The implications
of the presented results are discussed in Section VI and
Section VII draws the work to a close with the conclusion.
1.3
TABLE I
RELATIVE TX POWER PER VRB ON THE ASB AND BSB.
Index 1 2 3 4 5 6 7 8 9 10
ASB TX 0
1
9
2
9
3
9
4
9
5
9
6
9
7
9
8
9
1
BSB TX 3
25
9
23
9
21
9
19
9
17
9
15
9
13
9
11
9
1
II. SOFT REUSE
A problem with the presentation in Figure 1 is that it
confounds the physical aspects of soft frequency reuse with
the virtual aspects of resource allocation by implying that the
boosted resource should be given to the cell edge UEs.
Since this work looks at giving the boosted resource to
the cell-centre UEs, soft reuse is presented neutrally as a
tessellating pattern with a boosted part and an attenuated part,
in the manner of Figure 2.
Fig. 2. Soft Frequency Reuse Conguration. The band is divided into atten-
uated and boosted regions. The scheduler decides which UEs are allocated to
which regions.
The available bandwidth is partitioned into an Attenuated
Sub Band (ASB) and a Boosted Sub Band (BSB) in the propor-
tion of 2:1. The position of the BSB third in the overall band
is changed on a per cell basis to create a tessellating pattern.
The only difference from Figure-1 is that cell geography is
not shown since we wish to avoid communicating apriori
geographical biases on the usage of the ASB or BSB.
The relative transmit powers of the ASB and BSB determine
how soft the overall reuse factor is. The power ratios shown
in Table-I were examined in this work.
This range of soft reuse power ratios is bounded by two end
points: reuse three at index 1, and reuse one at index 10. The
points in between linearly interpolate across the space dened
by these end points.
Observe that for Index 1, since the TX power on the ASB is
0, all UEs are assigned to the BSB. The BSB in this case uses
1/3rd of the total bandwidth at 3 times the transmit power.
In the following sections, the impact of applying each of
these soft reuse power ratios is examined. Different scheduling
strategies are considered to demonstrate the interaction be-
tween soft reuse power ratio and scheduling strategy. The cell
performance is measured for each condition, to understand, if
at all, where each soft reuse power ratio performs best.
III. EXPERIMENTAL METHODOLOGY
A. Overview
A realistic central London scenario is used to assess the
gains of applying the static soft reuse power ratios shown
in Table-I. The gains are measured in terms of scheduling
JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011 313
2011 ACADEMY PUBLISHER
performance for two scheduling approaches: mean rate, and
xed rate.
B. LTE System Assumptions
The left of Figure 3 illustrates the essential components
of a 10MHz LTE DL frame. In time, the frame consists of
10 subframes which each last 1ms. Half of a subframe is
called a slot. In frequency, each subframe is split into 50
Virtual Resource Blocks (VRBs). Each VRB is comprised of
a pair of physical resource blocks (PRBs). One VRB is the
smallest unit of allocation in LTE [32]. Each PRB spans 12
subcarriers in frequency and 7 symbols in time (shorter cyclic
prex was used). Each element of a PRB is called a Resource
Element (RE). An RE spans one subcarrier in frequency and
one symbol in time. An RE has a frequency width of 15kHz
and lasts approximately 70s.
S 0 S 1 S 2 S 3 S 10
VRB 0
VRB 1
VRB 2
VRB 49
VRB
PRB PRB
10Mhz LTE Frame
10 subframes (1ms each)
5
0

V
R
B
s
Resource
Elements (REs):
Fig. 3. The essential components of a DL LTE frame.
The right of the gure shows a VRB in detail, containing the
RE types modelled here. Primary and secondary synchronisa-
tion, and broadcast channels, only occur on the three VRB
either side of the central carrier. The former only occur in
frames 0 and 5 and the latter only in frame 0. Their detailed
action is not modelled: the channels only consume space that
would otherwise be occupied by data REs. In the majority of
the frame, only pilot, control, and data REs are present.
Pilot symbol positions and associated RSRP computation is
modelled accurately according to [32]. Control channels are
assumed to consume the rst 3 symbols of every subframe,
their action is not modelled, and they only consume space that
would otherwise be occupied by data symbols. The average
number of data REs per VRB was computed as 124.8720. This
number is at the root of all throughput computations.
C. MCS Codeset
To map SINR to throughput, a lookup curve obtained from
Vodafone Group [33] was used. Figure 4 shows the curve
relative to the 3GPP reference curve which uses a single
antenna (SISO) transmit/receiver [2]. The Vodafone curve uses
2x2 MIMO and assumes optimal switching between STBC and
spatial multiplexing. The fading at the link level was based on
the ITU Pedestrian B channel at 3km/h [34].
D. Deployment scenario
A realistic London scenario was used for all simulations.
The data represents an area of central London. Antenna
settings and terrain data reect the actual network settings used
in 2004 for the Vodafone UMTS macro deployment. Figure 5
10 5 0 5 10 15 20 25 30
0
1
2
3
4
5
6
7
SINR (dB)
R
a
t
e

(
b
i
t
s

p
e
r

s
y
m
b
o
l
)
Vodafone obtained LTE lookup curve (3km/h)
Compared to 3GPP reference curve
Shannon
Vodafone MIMO
3GPP SISO
Fig. 4. SINR to MCS bitrate lookup curve used here in comparison to
Shannon and a 3GPP reference curve.
Fig. 5. Atoll predictions and antenna locations for the London scenario
scenario used in all simulations. Circled antennas were excluded from results
collection.
shows the London area studied along with generated pathloss
predictions.
Pathloss was calculated at a resolution of 10m
2
using the
Pace3D ray tracing software module in Atoll [35]. Pace3D
accurately models the effects of building penetration losses,
reection, and refraction effects and provides a realistic picture
of the actual pathloss variation experienced in each cell. To
mitigate border simulation affects, results were not collected
for the circled cells in Figure 5.
In all experiments, each cell transmits continuously, so that
the worst case interference scenario is represented.
IV. MEAN RESULTS
To generate mean cell results the following procedure was
used:
ForEach ( Sof t Reuse Power Ra t i o ) : Do
ForEach ( Ce l l ) In ( Sc e na r i o ) : Do
ForEach ( Squar e i n Ce l l Area ) : Do
A < ASB Bi t r a t e
B < BSB Bi t r a t e
C < Mean : 2/ 3 A + 1/ 3 B
Done
Compute mean of A, B, C over Squar es
314 JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011
2011 ACADEMY PUBLISHER
Done
Compute mean of A, B, C over Ce l l s
Recor d A, B, C f o r Sof t Reuse Power Ra t i o
Done
In the rst line, the transmit power prole is applied to
the network. This means every cell transmits at exactly the
power specied in the prole on each VRB. A full interference
model is examined which means all cells are simultaneously
transmitting on all VRBs. The algorithm then iterates over all
cells in the scenario, and averages the measures A, B, and C
for each cell at a resolution of 10m
2
. This is the meaning of
Square in the algorithm. The algorithm ends by averaging
the results across all cells.
Figure 6 plots the mean cell throughput as a function of soft
reuse power ratio.
1 2 3 4 5 6 7 8 9 10
6
6.5
7
7.5
8
8.5
9
x 10
6
Soft Reuse Power Ratio Index (R3 > R1)
T
h
r
o
u
g
h
p
u
t

C
o
n
t
r
i
b
u
t
i
o
n

(
b
p
s
)
Mean Cell Throughput Against Soft Reuse Power Ratio Index
(London,VF Codeset)
Mean cell throughput (Total)
Fig. 6. Mean cell throughput
The general trend observed is a decline in throughput when
moving away from reuse one toward reuse three. To compare
the two extremes: mean throughput is 40% greater under
reuse one than it is under reuse three. This difference is signif-
icant with p < 1 10
7
under a right tailed, unequal variance,
ttest (Satterthwaites approximation was used to address the
Behrens Fischer problem [36]).
It is desired to understand exactly why this result is ob-
served. It is straightforward to examine the two end points:
reuse three and reuse one. The question is why the reduction
in bandwidth in the reuse three case is not compensated,
to borrow terminology from [37], by the improved SINR
conditions.
To see why, consider the SINR distributions under reuse
three and reuse one, as illustrated in Figure 7.
The SINR distribution is right-shifted under reuse three
compared to reuse one. For any monotonically increasing MCS
lookup curve, improved SINR results in improved bitrate. Yet
since bandwidth is reduced by a factor of three the effective
bitrate observed will be reduced by a factor of three. For a
given UE to benet, it must thus obtain more than a factor
three improvement in bitrate.
10 0 10 20 30 40
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SINR (dB)
C
D
F

F
r
e
q
u
e
n
c
y
Empirical CDF of SINRs Under Reuse One and Three
(London,VF Codeset)
Reuse One
Reuse Three
Fig. 7. Cell SINR CDFs under reuse three and reuse one.
Figure 8 illustrates which parts of the cell obtain such an
improvement when switching from reuse one to reuse three
and which do not. The gure plots the effective bitrate under
reuse three, for sets of UEs dened by MCS index under reuse
one. To make this clear, consider MCS index 10. The reuse
one rate plotted is simply the rate for that MCS scheme. The
reuse three rate, is the effective mean bitrate of all UEs under
reuse three, that under reuse one were served by MCS index
10. A way to think of this is that each MCS serves an area
of the cell under reuse one, and the plot shows how the mean
bitrate changes over each MCS-area when switching to reuse
three.
0 5 10 15 20 25 30 35
0
1
2
3
4
5
6
7
MCS Index
M
C
S

R
a
t
e

(
b
i
t
s

p
e
r

s
y
m
b
o
l
)
Effective MCS Rate Under Reuse Three
Compared to Reuse One (London,VF Codeset)
Reuse one rate
Reuse three rate (>R1)
Reuse three rate (<R1)
Fig. 8. Effective change in bitrate index when switching from reuse one to
reuse three.
The points labelled with up arrows are those which result in
an effective bitrate improvement and those labelled with down
arrows an effective bitrate loss. As can be seen, some parts
of the cell, namely those dened by low index MCS schemes,
do benet from switching to reuse three. It follows that the
JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011 315
2011 ACADEMY PUBLISHER
mean cell result is determined by the relative proportional use
of each MCS scheme. Figure 9 plots the relative proportional
use of each MCS scheme under reuse one.
0 5 10 15 20 25 30 35
0
0.02
0.04
0.06
0.08
0.1
0.12
MCS Index
F
r
e
q
u
e
n
c
y

o
f

O
c
c
u
r
a
n
c
e
Frequency of MCS Use Under Reuse One
(London,VF Codeset)
Fig. 9. Frequency of MCS use under reuse one.
When the MCS distribution shown in Figure 9 is considered
against the effective bitrate changes shown in Figure 8, the
outcome is that approximately 50% of the UEs gain from
switching to reuse three, and approximately 50% do not gain.
When reuse three wins out, a mean improvement of 0.13
bits per symbol is obtained, whereas when reuse one wins out,
a mean improvement of 1.88 bits per symbol is obtained.
The overall result is that reuse one wins out when the whole
cell is taken into account, as was illustrated in Figure 6. It
should now be clear that the mean cell result is determined by
three items:
1) The change in SINR distribution bought about by the
coordination of interference.
2) How this corresponds to a change in MCS schemes used
due to the gradient of the MCS lookup curve.
3) The relative proportional use of each MCS scheme
before coordination.
Clearly then, these items are critical, and anything that
changes them can change the outcome of the competition
between reuse three and reuse one, or in general reuse one
and some other interference scheme such as soft reuse that
improves SINR. From this it follows that scheduling policy
plays a crucial role in determining whether a given interference
coordination scheme brings about a benet or not. The sched-
uler decides which UEs receive resources, and thus modies
the relative proportional use of MCS schemes, which is the
third item above.
For schedulers which bias the resource allocation to low
SINR UEs, or for cells which have a very large percentage of
low SINR UEs, there is likely to be a benet from statically
applied soft reuse schemes. In other cases, there will not
be. However, the exact scheduling strategy and exact UE
distribution will determine the overall result and should be
examined on a case by case basis.
To summarise: in the mean, no net benet is obtained
from the application of any soft reuse scheme tested. The
reason for this has been clearly explained in terms of the
interaction between SINR distributions, MCS lookup curve,
and proportional use of MCS schemes.
In the next section soft reuse is applied to a xed rate
scheduler to get some idea how in practise, scheduling shifts
the relative proportional use of MCS schemes, and whether or
not this results in an overall benet under soft reuse.
V. FIXED RATE RESULTS
In this section a semi-realistic scheduler is examined whose
goal is to satisfy as many UEs as possible, where each UE
has the same xed bitrate target. The scheduler operates as
described below:
ForEach (UE i n Schedul i ng Or der ) : Do
Al l o c a t e VRBs from t he BSB Un t i l :
No VRBs r emai n
OR UE i s s a t i s f i e d .
Al l o c a t e VRBs from t he ASB Un t i l :
No VRBs r emai n
OR UE i s s a t i s f i e d .
Updat e s a t i s f i e d UE count a c c o r d i n g l y
Done
The scheduler always allocates the best resource, the BSB,
rst. This means that the scheduling order is important. To
investigate the impact of which UEs get preference for the
BSB, three scheduling orders were considered:
1) Greedy - The UEs are scheduled according to wideband
SINR in descending order from best to worst.
2) Random - The UEs are scheduled in random order.
3) Leftist - The UEs are scheduled according to wideband
SINR in ascending order from worst to best.
Note that the third of these is the approach which is usually
promoted in the literature (see for example [18], [26], [38],
[39]), namely that the boosted part of the spectrum should be
given to the cell-edge UEs.
The complete process for obtaining results, which is exe-
cuted for each xed rate target, and each scheduling strategy,
is as follows:
ForEach ( Sof t Reuse Power Ra t i o ) : Do
ForEach ( Ce l l ) In ( Sc e na r i o ) : Do
ForEach ( Random s eed i n 1 t o 1000 ) : Do
Drop 25 UEs a t random
Schedul e t he UEs
Recor d number of s a t i s f i e d UEs
End
Compute mean over a l l UE dr ops
End
Compute mean over a l l c e l l s
End
Figure 10 plots the number of satised UEs, under the best
soft reuse power ratio, for each bitrate target, and for each of
the scheduling strategies.
The results are surprising and show that giving the BSB to
the cell-edge UEs, actually results in the worst performance.
It turns out that it is always best to be greedy and give
the BSB to the best SINR UEs, at least for the scheduler
examined. Note that the number of UEs satised never reaches
316 JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011
2011 ACADEMY PUBLISHER
500 450 400 350 300 250 200 150 100 75 50 25
0
5
10
15
20
25
Bitrate Target (kbps)
N
u
m
b
e
r

O
f

S
a
t
i
s
f
i
e
d

U
E
s
Number Of Satisfied UEs Under Best Soft Reuse Power Ratio
For Decreasing Bitrate Targets (London,VF Codeset)
Greedy Scheduler
Random Scheduler
Leftist Scheduler
Fig. 10. Number of satised UEs under the best performing soft reuse power
ratio, for each bitrate target.
the maximum 25, this is because the Monte Carlo simulation
samples some areas which cannot receive any throughput
under the MCS lookup curve used. In practise however, time
dependent variation, and partial loading, should allow all UEs
to be satised for reasonable bitrate targets.
Figure 10 plots the number of satised UEs under the best
soft reuse power ratio for each xed rate target. It does not
specify which soft reuse power ratio is best for each xed
rate target. For the greedy scheduler, the only scheduler of
interest given the above results, the answer is as follows: for
bitrate targets 500 down to 200, reuse one satises the greatest
number of UEs, and for bitrate targets 150 and below, reuse
three satises the greatest number of UEs.
Thus, no intermediate soft reuse scheme ever does better
than either reuse one or reuse three in this scheduling scenario.
Given that either reuse one or reuse three satises the greatest
number of UEs, Figure 11 plots the ratio of the number UEs
satised under reuse three, to the number satised under reuse
one.
When 25 UEs are trying to get 500kbps each, the system
is overloaded, and in this case reuse one satises upto 10%
more UEs than reuse three. For the lower load and saturation
states, reuse three satises upto 4% more UEs than reuse one.
The former gain comes about because the greedy scheduler
prioritises UEs with a high MCS which benet most from
having the full resource available to them. The latter gain
comes about because reuse three is able to serve UEs which
cannot be served under reuse one due to the cut-off point in
throughput caused by the lookup curve. In practise this gain
is likely to be diminished because time-dependent fading will
periodically bring cut-off UEs into service.
Furthermore, given the results of Section IV, it will be
observed that any reuse three gain will come at the cost of
reduced mean rate, and will only be apparent for low bitrate
targets. Thus it is unlikely in practise that any signicant
gain in xed rate satisfaction would be observed from the
application of soft-reuse in the general case.
500 450 400 3 5 0 300 250 200 150 100 75 50 25
0.85
0.9
0.95
1
1.05
R3/R1 Performance Ratio (London,VF Codeset)
Bitrate Target (kbps)
R
3
/
R
1

P
e
r
f
o
r
m
a
n
c
e

R
a
t
i
o
Fig. 11. Ratio of satised UEs under reuse three to satised UEs under reuse
one, for each bitrate target.
VI. DISCUSSION
A. Summary
The conditions under which soft reuse can be expected to
benet have been clearly enumerated: SINR distribution shift
under ICIC, SINR to MCS mapping, and proportional usage
of each MCS scheme before and after ICIC. In addition, the
intuitive notion that the cell-edge UEs should receive the
boosted part of the spectrum has been demonstrated false in
the case of a xed bitrate service class presented.
B. Scope of results
It may be argued that the results presented here are too
specic, and that they overt the particular London scenario
examined. Given this possibility, the experiments presented
were repeated for a 57 cell hexagonal environment and re-
peated again for the 3GPP codeset shown in Figure 4.
In the hexagonal case, there are greater benets from coordi-
nation, but overall the mean rate still favours reuse one, and the
xed rate scheduling outcome shows the same trends described
here. Using the 3GPP codeset, the only differences observed
are expected lower throughputs, but no diversion from the
trends. In summary, there are no qualitative differences in the
results or the implications of the results. Note however that
the degradation observed when switching from hexagons to
the London scenario is likely to be even greater for femtocells
and highly irregular networks. This is because the spatial
orthogonality on which static reuse schemes depend, will be
eroded.
C. Contributions of this paper
The novel contributions of this work are threefold:
The primary factors which manifest static reuse results
are illustrated through a simple example. These are: SINR
distribution shift under ICIC, SINR to MCS mapping, and
proportional usage of each MCS before and after ICIC
due to user distribution and scheduling strategy. Different
JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011 317
2011 ACADEMY PUBLISHER
scheduling strategies may manifest fundamentally differ-
ent outcomes for a given ICIC approach.
Simulations are performed for a realistic London deploy-
ment, and all results are quantied statistically. This is in
contrast to former work cited.
The assumption that the cell-edge UEs should be assigned
the best resources, implied by former work, is challenged
and demonstrated false for the trafc class examined.
VII. CONCLUSION
We propose examining results in terms of relative MCS
improvement curves given the scheduler examined as we have
done here, rather than solely in terms of CDF shifts due to
the soft reuse scheme applied. The former approach captures
the important interactions between UE distribution, scheduling
strategy, and MCS codeset, whereas the latter only reects the
SINR change independent of these.
The notion that it is better to give cell-edge UEs the
resources having a high reuse factor has been demonstrated
false in the case examined here.
REFERENCES
[1] 3GPP TR 25.913, V9.0.0: Requirements for Evolved UTRA (E-UTRA)
and Evolved UTRAN (E-UTRAN), 2009.
[2] 3GPP TS 36.942 - V8.2.0 - Evolved Universal Terrestrial Radio Access
(E-UTRA); Radio Frequency (RF) system scenarios, May 2009.
[3] Mahmudur Rahman and Halim Yanikomeroglu and William Wong,
Interference Avoidance with Dynamic Inter-Cell Coordination for
Downlnink LTE System, in WCNC09: Proceedings of the 2009 IEEE
conference on Wireless Communications & Networking Conference.
Piscataway, NJ, USA: IEEE Press, 2009, pp. 12381243.
[4] Alexander L. Stolyar and Harish Viswanathan, Self-organizing Dy-
namic Fractional Frequency Reuse for Best-Effort Trafc Through
Distributed Inter-cell Coordination, in INFOCOM 2008. The 27th
Conference on Computer Communications. IEEE, April 2008, pp. 691
699.
[5] W. Wu, M. Gitlits, and T. Sakurai, Dynamic resource allocation
with inter-cell interference coordination for 3GPP LTE, Microwave
Conference, 2008. APMC 2008. Asia-Pacic, pp. 1 4, dec. 2008.
[6] Zhifeng Tao et al and Toshiyuki Kuze, Dynamic Inter-cell Interference
Coordination (ICIC) and Signaling, 2008.
[7] X. Zhang, C. He, L. Jiang, and J. Xu, Inter-cell interference coordi-
nation based on softer frequency reuse in ofdma cellular systems, jun.
2008, pp. 270 275.
[8] R. Chang, Z. Tao, J. Zhang, and C.-C. Kuo, Multicell ofdma downlink
resource allocation using a graphic framework, Vehicular Technology,
IEEE Transactions on, vol. 58, no. 7, pp. 3494 3507, sep. 2009.
[9] N. Himayat, S. Talwar, A. Rao, and R. Soni, Interference management
for 4g cellular standards [wimax/lte update], Communications Maga-
zine, IEEE, vol. 48, no. 8, pp. 86 92, aug. 2010.
[10] Ericsson, R1-074444: On Inter-cell Coordination Schemes without/with
Trafc Load Indication, 3GPP TSG-RAN WG1 Meeting #50, Shanghai,
China, October 2007.
[11] Marc C. Necker, Scheduling Constraints and Interference Graph Prop-
erties for Graph-based Interference Coordination in Cellular OFDMA
Networks, vol. 14, no. 4, pp. 539550, 2009.
[12] M. Necker, Towards Frequency Reuse 1 Cellular FDM/TDM Systems,
in Proceedings of the 9th ACM/IEEE International Symposium on
Modeling, Analysis and Simulation of Wireless and Mobile Systems
(MSWiM 2006), October 2006.
[13] M. C. Necker, Integrated scheduling and interference coordination in
cellular ofdma networks, sep. 2007, pp. 559 566.
[14] F. Xiangning, C. Si, and Z. Xiaodong, An Inter-Cell Interference Co-
ordination Technique Based on Users Ratio and Multi-Level Frequency
Allocations, in Wireless Communications, Networking and Mobile
Computing (WiCom 2007), September 2007, pp. 799802.
[15] Gabor Fodor and Chrysostomos Koutsimanis and Andrs Rcz and Norbert
Reider and Arne Simonsson and Walter Muller, Intercell Interference
Coordination in OFDMA Networks and in the 3GPP Long Term
Evolution System, Journal of Communications, vol. 4, no. 7, pp. 445
453, August 2009.
[16] Gary Boudreau and John Panicker and Ning Guo and Rui Chang and
Neng Wang and Sophie Vrzic, Interference Coordination and Cancella-
tion for 4G Networks, IEEE Communications Magazine, vol. 47, no. 4,
April 2009.
[17] China Mobile Research Institute, C-RAN: The Road Towards Green
RAN. White Paper. Version 1.0.0. April 2010.
[18] Huawei, 3GPP TSG RAN WG1 Meeting #41, R1-050507 - Soft
Frequency Reuse Scheme for UTRAN LTE, 2005.
[19] Siemens, 3GPP TSG RAN WG1 Meeting #41, R1-050476: Evolved
UTRA uplink scheduling and frequency reuse, 2005.
[20] Huawei, 3GPP TSG RAN WG1 Meeting Ad Hoc Meeting, R1-050629
- Inter-cell Interference Mitigation.doc, 2005.
[21] Ericsson, 3GPP TSG RAN WG1 Meeting #42, R1-050764 - Inter-cell
Interference Handling for E-UTRA, 2005.
[22] Alcatel, 3GPP TSG RAN WG1 Adhoc Meeting, R1-050593 - Interfer-
ence coordination for evolved UTRA uplink access, 2005.
[23] T. Instruments, R1-051059: Inter-Cell Interference Mitigation for EU-
TRA, October 2005.
[24] Nokia, 3GPP TSG RAN WG1 Meeting #44, R1-060291: OFDMA
Downlink Inter-cell Inteference mitigation, 2006.
[25] N. D. Ericsson, TSG-RAN WG1 meeting #44, R1-060586 - Downlink
and uplink inter-cell inteference co-ordination/avoidance - impact on the
specications, 2006.
[26] Mohammad Abaii et al., IST-4-027756. WINNER II D4.7.2 v1.0.
Interference avoidance concepts.
[27] Texas Instruments, 3GPP TSG RAN WG1 Meeting #44, R1-060368
- Performance of Inter-Cell Inteference Mitigation with Semi-Static
Frequency Planning for EUTRA Downlink.txt, 2006.
[28] Ericsson, R1-061374: Downlink inter-cell interference co-
ordination/avoidance evaluation of frequency reuse, May 2006.
[29] IPWireless, 3GPP RAN WG1 Ad Hoc Meeting, R1-050652 - Attaining
the Cell Edge Performance Requirements for the LTE Downlink, 2005.
[30] Siemens, 3GPP RAN WG1 Ad Hoc Meeting, R1-060135 - Interference
Mitigation by Partial Frequency Reuse, 2006.
[31] Raymond Kwan and Cyril Leung, A Survey of Scheduling and Inter-
ference Mitigation in LTE. Article ID 273486. Journal of Electrical
and Computer Engineering, vol. 2010.
[32] 3GPP TS 36.211 - V8.7.0 - Evolved Universal Terrestrial Radio Access
(E-UTRA); Physical Channels and Modulation, May 2009.
[33] Vodafone Group Plc. Homepage. Accessed 28.09.10,
http://www.vodafone.com/hub page.html.
[34] I. Corporation, 3GPP TSRG1-01-0030, Further Results on CPICH
Interference Cancellation as A Means for Increasing DL Capacity.
[35] Atoll Overview. Forsk website. Accessed 10/04/10.
http://www.forsk.com/web/EN/11-atoll-overview.php.
[36] B D Hall and R Willink, Does Welch-Satterthwaite make a good
uncertainty estimate? Metrologica, vol. 38, no. 1, 2001.
[37] Andras Racz and Norbert Reider and Gabor Fodor, On the Impact of
Inter-Cell Interference in LTE, in Global Telecommunications Confer-
ence, 2008. IEEE GLOBECOM 2008. IEEE, 2008, pp. 16.
[38] Xiang Yikang, Luo Jijun, and Christian Hartmann, Inter-cell Interfer-
ence Mitigation through Flexible Resource Reuse in OFDMA based
Communication Networks, in In proceedings of 13th European Wireless
Conference (EW07), Paris, France, April 2007.
[39] X. Xiang, F. Liu, and Y. Ji, Simulation based performance evaluation of
ICI mitigation schemes for broadband wireless access networks, in CNS
08: Proceedings of the 11th communications and networking simulation
symposium. New York, NY, USA: ACM, 2008, pp. 181187.
318 JOURNAL OF COMMUNICATIONS, VOL. 6, NO. 4, JULY 2011
2011 ACADEMY PUBLISHER