Content uploaded by Liam Murphy
Author content
All content in this area was uploaded by Liam Murphy on Aug 03, 2014
Content may be subject to copyright.
Interference Aware Relay Station Location Planning for
IEEE 802.16j Mobile Multi-hop Relay Network
Yang Yu
School of Computer Science
and Informatics
University College Dublin
Dublin, Ireland
yang.yu@ucd.ie
Seán Murphy
School of Computer Science
and Informatics
University College Dublin
Dublin, Ireland
sean.murphy@iname.com
Liam Murphy
School of Computer Science
and Informatics
University College Dublin
Dublin, Ireland
liam.murphy@ucd.ie
ABSTRACT
In this paper, an interference aware algorithm for multi-cell
Relay Station (RS) location planning problem is proposed in
order to investigate the capacity gains possible using trans-
parent mode 802.16j. The study focuses on the problem
of determining the best choice of locations for RS from a
given set of candidate sites - Base Station (BS) and Sub-
scriber Station (SS) locations are assumed given. The pro-
posed algorithm evaluates each RS based on the channel
conditions of the links between the RS and its surrounding
Subscriber Stations(SS) and the throughput gain it could
deliver, taking into account co-channel interference. The
performance statistics were collected over a large amount of
experiments with many different configurations, including
Base Station(BS) antenna types, number of candidate RS
sites, RS transmitting power, with/without spatial reuse,
etc. The results can be divided into two sections. Firstly,
the algorithm was tested for optimality and scalability and
was shown to perform well compared to classical Integer Pro-
gramming solution techniques. Secondly, the results gen-
erated by the algorithm when applied to different scenar-
ios were analyzed. Results show that using spatial reuse
could achieve a much higher throughput gain, 3-5 time, then
without using spatial reuse, although sectorised BS antenna
limits the separation of co-channel RSs in a sector. The
results demonstrate the characteristics of the trade-off be-
tween number of RSs and throughput gain for different con-
figurations.
Categories and Subject Descriptors
C.2.1 [Computer-Communication Networks]: Network
Architecture and Design—Wireless communication, Network
topology
General Terms
Algorithm, Design
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
PM2HW2N’09, October 26, 2009, Tenerife, Canary Islands, Spain.
Copyright 2009 ACM 978-1-60558-621-2/09/10 ...$10.00.
Keywords
IEEE 802.16j, Spatial Reuse, Network Planning, Relay Sta-
tion Location Problem
1. INTRODUCTION
As demand for wireless broadband continues to increase at
a staggering rate, there is a clear need for the deployment of
more advanced, spectrally efficient radio access technologies.
This has led to much heated debate within the community
over the last year, with particular focus on the relative merits
of WiMAX and LTE technologies. Much of this debate,
however, has been very partisan and often with a mindset
that there must be a single winning technology, rather than
recognizing that there may be room for both.
At heart, both of these technologies are very similar in
terms of how they behave at the radio interface; however,
LTE roadmaps are much more aligned with those of large
cellular network operators, which does give it a significant
advantage in terms of large scale rollout. On the other hand,
WiMAX technology is more advanced both in terms of stan-
dardization and technology development: hence, it is of in-
terest to operators wanting to roll out a high capacity access
network today.
Here, we take the view that WiMAX will have some role
to play in future networks. Further, we contend that the
emerging WiMAX standards will see light in the form of
real products and we focus here on one specific initiative
with the WiMAX standardization domain - IEEE 802.16j.
IEEE 802.16j [1] is an enhancement to previous 802.16
standards to provide support for relays, thus providing for
increased capacity and/or coverage, depending on the sce-
nario. The standard does not permit changes to SSs, hence
the changes introduced by the standard focus on communi-
cations between (enhanced) BS and the new RSs.
As this is still an emerging standard, it is still somewhat
unclear what gains it can deliver, although there have been
some contributions in this space which are starting to pro-
vide some answers. One issue which arises in this context is
how to approach network planning – a multi-hop radio ac-
cess network gives rise to new problems which have not been
addressed in previous radio planning approaches. Here, we
propose an algorithm for planning RS locations given a set
of BSs, SSs and candidate RS sites.
The contributions of this paper can then be identified as
(i) a new algorithm for selecting RS locations amongst a
given set of candidate sites and (ii) an analysis of the trade-
off between throughput increase and number of required RSs
for different scenarios.
The rest of the paper is structured as follows. Previous
work related to the performance of IEEE 802.16j system is
presented next. In Section 3, aspects of the system architec-
ture under study, such as system model, transparent mode,
path selection, cell throughput, propagation and interference
models are described. The proposed planning algorithm is
then introduced, followed by a discussion of its scalability
and the quality of solution it obtains in Section 4. In Sec-
tion 5, numerical results are presented and discussed and
finally Section 6 concludes the paper.
2. RELATED WORK
There are some activities on evaluating the performance of
802.16j MMR networks. In [2], simulations were performed
to evaluate the system capacity in comparison to the conven-
tional single-hop deployment. The research demonstrates
the effects to the system performance of the RS height and
transmission power, the number of RSs that are transmit-
ting simultaneously as well as that the relay system suffers
from narrow channel size due to high overhead. A through-
put gain of around 50% was obtained for a system with and
without RSs.
In [3, 4], the IEEE 802.16j system’s performance was eval-
uated under some specific assumptions. In [3], a detailed
system level performance evaluation was performed and the
network coverage and throughput was compared under sev-
eral system configuration. The results show that a maximum
increase of 42% of the system capacity or 89% of the network
coverage can be achieved. In [4], the work was specifically
focused on the uplink. Through both analytical and simula-
tion modelling, the maximum gain obtained without spatial
reuse is 40%.
In the context of applying spatial reuse in IEEE 802.16j,
there are also some contributions focusing on the perfor-
mance gain. In [5], an analytical model was used to investi-
gate the system capacity of IEEE 802.16j transparent mode
with varying number of RSs and their transmission power.
In the study an overall throughput gain of 125% and 55%
can be obtained with and without spatial reuse respectively
in comparison with 802.16e system.
In [6], an analytical model was studied on investigation
cell capacity with two-hop coverage extension. The paper
shows the tradeoff between coverage extension and the ca-
pacity of a cell. It also shows that spatial reuse could miti-
gate the capacity loss.
In [7], the authors proposed several algorithms to leverage
the gain get by the users from diversity and spatial reuse.
The performance analysis shows that for small number of
users, the diversity gain would be sufficient to sustain the
traffic load, but with large number of users, spatial reuse
must be exploited. The results show that the throughput
gain of spatial reuse could be increased by 50% over the
diversity case.
While there have been a number of contributions in the
area, most of them are concerning about the performance
that IEEE 802.16j could bring. There have not been con-
tributions on planning for relay systems which is considered
very important to the operators.
3. SYSTEM ARCHITECTURE
IEEE 802.16j MMR network is a versatile system, it sup-
ports many options for setting up a network. In this sec-
tion, some assumptions made for the system studied are de-
scribed. Related aspects such as IEEE 802.16j transparent
model, path selection, cell throughput and the propagation
and interference models are also presented.
3.1 System Model
The problem studied in this work is to plan RS locations
in existing multi-cell single-hop networks. There are several
assumptions for the system architecture:
•The existing single-hop network topology is in regular
hexagon layout.
•Frequency reuse factor 3 is used for both sectorised
and omni antenna scenarios.
•All RSs are operate in transparent model.
•The multi-hop is limit to two hops for the planning,
i.e. all SSs are transmitting either through an RS to a
BS or directly to a BS.
3.2 Transparent Mode
In IEEE 802.16j standards[1], two relay modes are de-
fined: transparent mode and non-transparent mode. In non-
transparent mode, RSs transmit frame control information
to their downstream SSs while in transparent mode, which
is the assumption of this work, RSs do not transmit the
control information and the SSs have to rely on the control
information they get from the BSs. Because of this, trans-
parent mode does not support coverage extension but only
capacity enhancement.
In transparent mode with TDD, each frame is divided into
DL subframe and UL subframe. The DL subframe consists
of the DL access zone and an optional transparent zone.
The access zone is for BS transmitting the data bursts to its
directly connected SSs and RSs and subsequently in trans-
parent zone, the RS forwards the data to its downstream
SSs. Decode and forward RSs is assumed to be used in this
work so that the BS-RS link and the RS-SS link may use
different modulation and coding schemes.
3.3 Path Selection
A path selection mechanism is also used. Whether an SS
is receiving data from a BS through an RS is determined
by the link condition of each link on the transmission path.
Table 1 shows the 7 Modulation and Coding Schemes(MCS)
defined in the standard [1] for OFDMA. A certain MCS is
chosen at a receiver is based on its SNR/SINR value. A
basic weights scheme is used to reflect the link efficiency.
wb,wrand wsare used to represent the weights of BS-SS,
BS-RS and RS-SS links respectively. The weights are the
inverse of the bit rate in symbol/bit. A data transmission
through an RS only occurs when the link through the RS is
more efficient than the direct link, i.e. wb> wr+ws.
3.4 Cell Throughput
The cell throughput can then be derived as follow:
Ui,j =Nused
Ts∑n
k=1 Wk
n(1)
where Ui,j is the cell/sector throughput of the sector jof
BS iand in the case of omni antenna, it is considered that
Table 1: OFDMA MCS, Bit Rate and Weights
Modulation QPSK 16-QAM 64-QAM
Coding rate 1/2 3/4 1/2 3/4 1/2 2/3 3/4
Rx SNR(dB) 5 8 10.5 14 16 18 20
Bitrate(b/s) 1 3/2 2 3 3 4 9/2
Weight 9/2 3 9/4 3/2 3/2 9/8 1
the BSs have only 1 sector. Nused is the number of data
sub-carriers, Tsis the symbol duration, nis the number of
SSs in the cell, Wkis the sum of weights of the more efficient
transmission path from SS kto the BS: min(wr+ws, wb).
Single-hop cell throughput is represented as U′
i,j using the
weights of BS-SS direct links wbfor Wkin Formula (1).
3.5 Radio Propagation Model
In wireless communication systems, the radio propagation
model is a key component. There are many factors attenuate
the radio power on its pathway. These factors together with
the interference determines the signal power strength at a
receiver. In this section, the radio propagation model and
interference model used in this work is described.
The propagation model used in this work is a modified ver-
sion of the common SUI model recommended by the IEEE
802.16j standard body[8]. Only the path loss is considered in
this work with no fast fading and shadowing effects. With
the path loss, the received signal strength at the receiver
from a particular transmitter can then be get from:
Pt,r =PtGtGrAt,r
Lt,r
(2)
where Pt,r is the received signal power in Watt from trans-
mitter tto receiver r,Ptis the transmission power of the
transmitter, Gtis the antenna gain of transmitter, Gris the
antenna gain of receiver, Lt,r is the path loss between the
transmitter and the receiver and At,r is the antenna atten-
uation for sectorised antenna based on the formula given in
[8].
3.6 Interference Model
Interference coming from co-channel transmitters reduces
the signal quality at receiver. It is assumed in this work that
all the transmissions are frame-synchronism, i.e. the access
zone and the transparent zone are transmitting at their own
time slots. Under this assumption, the interference of the
transmissions in access zone is only coming from co-channel
BSs and the interference of the transmissions in transparent
zone is only coming from co-channel RSs.
Thus, with the interference, which MCS is used by a re-
ceiver is determined by the SINR of its received signal:
SI N Rt,r = 10 log10
Pt,r
BNfN0+∑t′Pt′,r
where SI N Rt,r is in dB, t′denotes the concurrent co-channel
transmitters, Pt,r is the received signal power and Pt′,r is the
received interference power as defined in Formula (2), Bis
the band noise, Nfis noise figure, N0is thermal noise.
The maximum cell rang Rcan then defined as:
R= max(dt,r),∀t, ∀r, S IN Rt,r > α
where dt,r is the distance between transmitter and receiver
and α= 5 is the SINR threshold for the lowest efficient MCS
as shown in Table 1. The distance Dbetween two adjacent
BSs is D=√3R.
4. PLANNING ALGORITHM
The proposed interference aware RS location planning al-
gorithm was designed for adding new RSs to an existing
single-hop network for both spatial reuse and non-spatial
reuse. In this section, the problem formulation and the pro-
posed algorithm is described.
4.1 Problem Formulation
The following inputs to the problem are defined as follow:
•B={B1, B2, ..., Bb}: set of BSs in existing network
•R={R1, R2, ..., Rr}: set of candidate RSs
•S={S1, S2, ..., Ss}: set of SSs in existing network
The objective of the problem is to determine a selection of
RSs from the candidate sites, and assign them to BSs and
SSs, in order to maximise the throughput gain brought by
the RSs.
Before introducing the proposed algorithm, it is also nec-
essary to define some parameters that are used in the pro-
posed algorithm:
•Weights of the links:
wb
i,j ,∀i∈Band ∀j∈S
wr
i,j ,∀i∈Band ∀j∈R
ws
i,j ,∀i∈Rand ∀j∈S
•Weight difference between BS-SS link and BS-RS-SS
link, it shows how much gain an RS could bring to each
existing BS-SS link:
Di,j,k =wb
i,k −(wr
i,j +ws
j,k),∀i∈B , ∀j∈Rand ∀k∈S
•Possible BS-RS-SS link, it shows whether an RS could
bring gain to each existing BS-SS link:
Li,j,k ={1, Di,j,k >0
0, Di,j,k ≤0,∀i∈B, ∀j∈Rand ∀k∈S
•Link map of antenna sector, it shows by which sector
of BS antenna a node is covered:
Sec ={{1,2,3},for sectorised antenna
1,for omni antenna
Ts
i,j =Sec, ∀i∈B , ∀j∈S
Tr
i,j =Sec, ∀i∈B , ∀j∈R
4.2 The Proposed Algorithm
The essential idea of the algorithm is to choose the best
RS sequentially. There are two criteria for choosing an RS:
1. It should cover the TPs with as good communication
links as possible
2. It could bring as much capacity gain as possible
An RS evaluation matrix was build based on the above
criteria is presented as follow:.
Ei,j =
1
∑s
k=1 Li,j,k ∑s
k=1 Pj,kLi,j,k
∑s
k=1 Di,j,kLi,j,k
,∀i∈Band ∀j∈R(3)
where the numerator is the average path loss between RS j
and all its potential SSs that are connected to BS iand the
denominator is the sum of the weight difference brought by
the RS. In the above formula, the weights used for RS-SS
links are measured based on the SNR values while for BS-RS
and RS-SS links are based on the SINR values. SNR values
used for RS-SS links is because of that at the beginning of
the planning process, the interference are not fixed yet and
will be calculated during the planning process.
Obviously, from formula (3) the smaller the value in the
RS evaluation matrix, the better the RS is. The RSs are then
examined based on the value in the RS evaluation matrix
from small to big. For each RS, together with all the selected
RSs the interference and the gain is calculated. If the gain is
over a certain threshold γ, the RS is then marked as selected.
The RS evaluation matrix is updated after each RS is chosen.
A detailed algorithm pseudo code is shown in Table 2.
Before the algorithm starts, several variables used in the al-
gorithm is initialised. Two levels of nested loops are used
in the algorithm, the outer loop is for spatial reuse and the
inner loop is for searching the suitable RSs within each spa-
tial reuse group. Obviously, for the scenarios without spatial
reuse, only the inner loop is applied. Each round of the in-
ner loop, an RS of the minimum value in the RS evaluation
matrix and its corresponding BS/sector is examined. With
the new RS and the RSs that have already been selected
in the cell/sector, the SINR and the weights of RS-SS links
are calculated. The new cell/sector throughput can then
be derived. If the new RS brings more then γgain to the
cell/sector, the RS is selected and some updates to the pa-
rameters are performed. Whether or not the RS is selected,
the corresponding value in the RS evaluation matrix is set
to infinity so that this RS is excluded from the search. In
order for the incompetent RSs to be considered again in the
next spatial reuse group, a duplication of the RS evaluation
matrix is made only for updating the RSs that are selected.
The output of the algorithm is a matrix consisting of RSs,
BS sectors and the spatial reuse group assigned to each RS.
4.3 Performance Analysis
A network planning problem is known as an NP-hard
problem and can be solved typically by using integer pro-
gramming approach. However, with integer programming
approach, even for a median size problem, it would take un-
realistic time. The proposed algorithm solves the problem
is a greedy way and is only linear time complexity:
T(n) = k×r×n=O(n)
where nis the size of the problem, ris the number of can-
didate RS sites per cell and kis the number is spatial reuse
groups. The problem size is the number of BSs. The number
of RSs and SSs are proportional to this number. In worst
case, the inner loop goes through all the RSs. The number
of spatial reuse groups is the number of outer loops which
is not dependent on the problem size, shown as kin the
formula. The actual execution time for the problems stud-
ied are less then one second to a few seconds on a normal
Table 2: Planning Algorithm Pseudo Code
Initialisation
1. An empty set keeping track of the RSs that have been
chosen: C← {ϕ}
2. An empty set keeping track of the SSs that have been
connected with an RS: D← {ϕ}
3. Gi,j,k ←0,∀i∈B, ∀j∈R, ∀k∈Sec for keeping track
of the assignment of spatial reuse groups
4. g←0, for current processing spatial reuse group
5. Duplicate RS evaluation matrix: E′
i,j ←Ei,j
Algorithm
begin
loop //loop for spatial reuse
g←g+ 1
loop //loop for RS selection
(b, r)←getpos(min(Ei,j ))
t←Tr
b,r
M← ∀j, Gb,j,t =g
K← ∀s, Lb,r,s = 1 & Ts
b,s =t
K←K−D
update SI N Ri,j ,∀i∈M , ∀j∈K
update ws
i,j ,∀i∈M, ∀j∈K
gainold ←Ub,t /U′
b,t −1
calculate new cell/sector throughput u←Ub,t
gainnew ←u/U ′
b,t −1
if gainnew −gainold > γ
C←C∪ {r}
N← ∀j, Db,i,j ≤0,∀i∈M, ∀j∈K
D←D∪K−N
Gb,r ←g
Ub,t ←u
update Ei,j , E′
i,j ,∀i∈B, ∀j∈R−C, ∀s∈S−D
end if
Eb,r ←Inf
loop until min(Ei,j ) =Inf
Ei,j ←E′
i,j
loop until min(Ei,j ) =Inf |non-spatial reuse case
end
Output
Gi,j
desktop computer depending on the problem.
A performance comparison of integer programming ap-
proach and the proposed algorithm is also shown in Figure
1 and in Table 3. The comparison is based on some tests
of planning for omni antenna without spatial reuse. The in-
teger programming approach could produce optimal results.
From the comparison, it can be seen that the proposed algo-
rithm gives slightly worse result then the optimal: with 0.6
more RSs on average to achieve 1.3% less throughput gain.
A more detailed planning experiments of the proposed
algorithm are discussed in the next section.
5. EXPERIMENTAL RESULTS
Experiments simulating the network planning process were
performed to investigate the performance of different system
configurations. In this section, the experiments scenarios are
introduced at the beginning, followed by some numerical re-
sults. In each of the numerical results subsection, only a
15% 20% 25% 30% 35% 40%
0%
10%
20%
30%
40%
Throughput Gain Comparison
Gain (%)
PDF
Integer prog algorithm
Proposed algorithm
Figure 1: Throughput gain comparison between in-
teger programming algorithm and proposed algo-
rithm.
Table 3: Comparison between Integer Programming
Algorithm and Proposed Algorithm (values are av-
erage)
Potential % SS No. RS Thruput
SSs(%) covered per cell gain
IntProg Alg 54% 50.8% 13.6 28.0%
Proposed Alg 52.5% 14.2 26.7%
single parameter varies in order for understanding the in-
fluence of certain parameters. The varying parameters are
the number of RS candidate sites per cell, RS transmission
power and the gain threshold of the proposed algorithm.
5.1 Scenario
The network planning were performed in two scenarios
based on two different types of BS antenna: omni antenna
and sectorised antenna with 120 degree angle. There are to-
tal 7 and 27 BSs respectively in sectorised and in omni an-
tenna scenario. In the scenarios, each cell uses the maximum
transmission range which makes the size of the scenarios are
4km×5.8km and 7km×7.6km respectively. Only the centre
cells of the scenarios – 1 for sectorised antenna scenarios and
7 for omni antenna scenarios – are used to collect statistical
information in order to avoid the border effect.
The system parameters used for both scenarios are shown
in Table 4. The default and all the possible values of the
three varying parameters during the experiments are as shown
in Table 5.
The locations of all nodes were randomly generated in a
uniform distribution within the planning area. It models the
actual constraints that a real world network operator would
typically have: the devices can only be placed at a certain
location of the selected sites, such as top of buildings, towers
or street lights, etc.
5.2 Gains of Spatial Reuse over non Spatial
Reuse
The purpose of this section is to discover how much gain
would spatial reuse bring after adding RSs to a single-hop
network. It is compared with non-spatial reuse planning
as well. Some other statistics of the planning results are
Table 4: System Parameters
Parameter Name Value
Frequency 2.5GHz
Channel bandwidth 20MHz
FFT size 2048
No. of data sub-carriers 1440
G(length of cyclic prefix) 1/4
Nf(noise figure) 12dB
N0-174dBm
BS
Height 40m
Transmission power 40dBm
Antenna gain 8dB
(for sector antenna)
RS Height 20m
Antenna type Omni
SS height 2m
BS-SS A, Macro-cell suburban,
ART to BRT for hilly
SUI model RS-SS terrain with moderate
terrain type to heavy tree densities
BS-RS D, Macro-cell suburban,
ART to ART
Table 5: Variables for the Experiments
Variable Name Default Value Possible Values
No. of RS candidate 80 40, 80, 160
sites per cell
RS transmission 30dBm 20dBm, 30dBm,
power 40dBm
Algorithm gain 0 0, 1, 2, 3, 4, 5
threshold(%)
also displayed. In the tests, all variables remain default as
described in the previous section.
Figure 2 shows the throughput gain of the relay systems
after planning process. From the figure, it can be seen that
for both omni antenna and sectorised antenna scenarios, the
throughput gain per cell of spatial reuse scenarios is signif-
icantly higher than that of non-spatial reuse scenarios: the
gains are around 3 times and 5 times more and can reach
100% and 170% for the two scenarios respectively. The rea-
son why the gain of sectorised antenna cases is not as signif-
icant as omni antenna cases is that, in each sector because
of the interference the RSs do not have enough space to be
placed separately enough and thus the interference is more
severe. The number of spatial reuse groups is not affected
much by the sectors: the number of spatial reuse groups in
both cases are very close and in average, there are 4.34 and
4.89 for sectorised antenna and omni antenna cases respec-
tively.
Figure 3 shows the number of SSs that are covered by RSs.
The curve of total potential SS illustrates the total number
SSs that has a potential of gaining from a relay of all the
SSs within each cell. As can be seen from the figure, there
are at least 70% of the potential SSs are covered by an RS.
Generally speaking, there are fewer relay covered SSs in sec-
torised antenna cases than that in omni antenna cases. This
is because in sectorised antenna cases there are less choices
of candidate site within each sector, thus as the number of
candidate RS sites increasing the difference should become
10% 30% 50% 70% 90% 110% 130% 150% 170%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90% Throughput Gain over Single−hop Network
Gain (%)
PDF
Sector Ant Spatial Reuse
Sector Ant no Spatial Reuse
Omni Ant Spatial Reuse
Omni Ant no Spatial Reuse
Figure 2: Throughput gain of spatial reuse and non-
spatial reuse after planning RSs to a single-hop net-
work.
10% 20% 30% 40% 50% 60% 70%
0%
10%
20%
30%
40%
50% % SS per Sector/Cell Benefits from RS
% SS
PDF
Total Potential SS
Sector Ant Spatial Reuse
Sector Ant no Spatial Reuse
Omni Ant Spatial Reuse
Omni Ant no Spatial Reuse
Figure 3: The number of SSs that are covered by an
RS.
smaller as will be discussed in Section 5.3. Also, the number
of SSs covered by RSs in spatial reuse cases is smaller than
that of non-spatial reuse cases, for the reason that within
each spatial reuse group there will be coverage gaps between
the RSs and it is very hard guaranteed that another spatial
reuse group happen to cover the gap.
As shown in Figure 4, the number of RSs selected for each
cell is way too many and this is because the algorithm gain
threshold is set to 0% for the tests. In Section 5.5, there are
tests trying to figure out a balance point where there are
a suitable number of RSs selected and also an acceptable
throughput gain could be obtained for each cell. However,
when looking at the trend of the number of RSs per BS,
the number in spatial reuse cases is significantly more than
non-spatial reuse which is very obvious: with spatial reuse,
the cell would be able to be fill in more RSs in many spatial
reuse groups and thus more throughput gain could be get.
5.3 Effect of the Number of RS Candidate Sites
In this section, the tests were performed with the number
of RS candidate sites varying from 40, 80, 160 per cell and
all other variables remain default. Results show the effect
of the number of RS candidate sites on the performance
10 15 20 25 30 35 40 45 50
0%
10%
20%
30%
40%
50%
60%
70%
80% Number of RS per BS
Number of RSs
PDF
Sector Ant Spatial Reuse
Sector Ant no Spatial Reuse
Omni Ant Spatial Reuse
Omni Ant no Spatial Reuse
Figure 4: The number of RSs selected for each cell.
10% 30% 50% 70% 90% 110% 120%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90% Throughput Gain per Sector
Gain (%)
PDF
40 Candidate RS/BS SR
40 Candidate RS/BS no SR
80 Candidate RS/BS SR
80 Candidate RS/BS no SR
160 Candidate RS/BS SR
160 Candidate RS/BS no SR
Figure 5: Throughput gain per sector for the num-
ber of RS candidate sites experiments in sectorised
antenna scenario.
gain. There are also other statistics shown in table. Since in
the networks currently running, sectorised antenna is more
commonly concerned, the details of omni antenna scenario
information are not graphically shown in figure.
More candidate RS sites means that there is higher prob-
ability the RSs being at a better location, i.e. there would
be more candidate RSs that can provide good link to more
SSs as well as bring a greater throughput gain for the SSs.
This is approved by the results as follows. From Figure 5,
it can be seen that the number of candidate RS sites sig-
nificantly effects the throughput gain of the planning result
in sectorised antenna scenario and the gain could reach over
120% highest and 73.5% on average.
Other statistical information are shown in Table 6. The
throughput gain of omni antenna scenario is also increasing
as the number of candidate sites increasing and can reach
a even higher number, 190% gain over single-hop network.
The number of candidate RS sites also affects the number of
SSs covered as described in previous section. It effects the
number of RSs selected as well since there are more to be
chosen and more at a better location.
5.4 Effect of the Transmission Power of RSs
In this section, the RS transmission power was set to
Table 6: Experimental Results with the Number of RS Candidate Sites Varying
``````````
`
Mean vals
No. RS Spatial Reuse Non-Spatial Reuse
40 80 160 40 80 160
Throughput gain(%) Sector 38.9 59.2 73.5 16.0 20.1 22.5
Omni 95.7 123.2 146.1 22.0 24.2 26.4
No. SS covered(%) Sector 30.3 39.0 46.8 36.9 46.0 52.4
Omni 43.2 49.3 52.4 47.9 52.5 54.0
No. RS per cell Sector 22.5 31.3 37.6 19.9 24.6 27.9
Omni 16.2 18.9 19.2 13.6 14.2 13.3
No. reuse group Sector 3.5 4.3 5.3 / / /
per sector/cell Omni 4.4 4.9 4.4 / / /
No. RS per Sector 2.2 2.4 2.4 / / /
reuse group Omni 3.7 3.9 4.3 / / /
20dBm, 30dBm, 40dBm respectively for the experiments in
order to understand the relationship among the transmis-
sion power of RSs, the gain brought by RSs and the number
of selected RSs.
In the case of transmission power increasing, the coverage
area of each RS is getting larger and they are more likely to
provide better service to the surrounding SSs. Looking at
the throughput gain in figure 6, the throughput gain of non-
spatial reuse cases is increasing as the transmission power
increasing, as there are no interference among RSs without
spatial reuse. However with spatial reuse, the interference
would become severe when the power becomes too high and
thus impairs the link quality of links between the RSs in
spatial reuse group and their SSs. As a result the throughput
gain in 40dBm case is not as much as that in 30dBm case.
One the other hand, although small RS transmission power
reduces the interference but it also reduces the coverage and
results in introducing a lot more RSs than necessary.
Some other aspects of the tests results are shown in Table
7. As can be seen, as an effect of increasing the RS trans-
mission power, there tends to be more SSs covered by RSs
and less RSs chosen.
10% 30% 50% 70% 90%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90% Throughput Gain per Sector
Gain (%)
PDF
RS Pt 20dBm SR
RS Pt 20dBm no SR
RS Pt 30dBm SR
RS Pt 30dBm no SR
RS Pt 40dBm SR
RS Pt 40dBm no SR
Figure 6: Throughput gain per sector for the RS
transmission power experiments in sectorised an-
tenna scenario.
5.5 Effect of the Gain Threshold
The experiments in the previous sections were performed
with the gain threshold γbeing 0%. Obviously, not all RSs
10% 30% 50% 70% 90% 100%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90% Throughput Gain per Sector
Gain (%)
PDF
Gain Threshold 1% SR
Gain Threshold 1% no SR
Gain Threshold 4% SR
Gain Threshold 4% no SR
Gain Threshold 5% SR
Gain Threshold 5% no SR
Figure 7: Throughput gain per sector for the gain
threshold experiments in sectorised antenna sce-
nario.
bring high gain to the throughput, and only the RSs that
bring big enough gain are worth to put into the network. It
saves the cost of the RSs that can only add a little gain to
the throughput. The purpose of this section, is by doing the
experiments with varying gain threshold to find a balance
point where there are suitable number of RS selected while
the throughput gain is still acceptable.
From Figure 7, it can be seen that the throughput gain
does not drop much until the gain threshold reaches 5% with
spatial reuse in sectorised antenna scenario. More detailed
information are shown in Table 8. As the gain threshold
increases, the throughput gain as well as the number of RSs
selected drop. With not very significant gain drop and a
suitable number of RSs selected, the balance points for the
problem studied are 4% and 3% for spatial reuse and non-
spatial reuse cases respectively in sectorised antenna sce-
nario, 3% and 1% for spatial reuse and non-spatial reuse
cases respectively in omni antenna scenarios.
6. CONCLUSIONS
In this paper, an interference aware algorithm was pro-
posed for planning RS locations in IEEE 802.16j multi-hop
relay network with spatial reuse. The algorithm was shown
to be more efficient than IP-based approaches and produced
similar quality solutions for small problems.
The algorithm was then applied to find good RS locations
Table 7: Experimental Results With RS Transmission Power Varying
hhhhhhhhhhhh
h
Mean vals
RS Pt(dBm) Spatial Reuse Non-Spatial Reuse
20 30 40 20 30 40
Throughput gain(%) Sector 42.9 59.2 52.4 12.4 20.1 26.6
Omni 92.4 123.2 127.7 17.5 24.2 31.9
No. SS covered(%) Sector 26.0 39.0 50.1 30.2 46.0 55.5
Omni 38.9 49.3 53.7 43.2 52.5 55.0
No. RS per cell Sector 33.4 31.3 20.0 31.3 24.6 12.2
Omni 28.0 18.9 10.9 24.8 14.2 6.3
No. reuse group Sector 3.7 4.3 3.8 / / /
per sector/cell Omni 4.9 4.9 3.7 / / /
No. RS per Sector 3.0 2.4 1.8 / / /
reuse group Omni 5.7 3.9 2.9 / / /
Table 8: Experimental Results With Gain Threshold γVarying
XXXXXXXX
X
Mean vals
γ(%) Spatial Reuse Non-Spatial Reuse
1 2 3 4 5 1 2 3 4 5
Throughput Sector 57.7 55.0 55.9 52.3 35.7 18.4 17.3 16.2 14.0 10.6
gain(%) Omni 120.4 110.7 97.4 53.6 / 22.1 19.6 15.0 4.8 /
No. SS Sector 36.2 33.1 33.3 31.4 23.3 35.9 33.0 30.5 26.4 19.5
covered(%) Omni 47.6 44.5 40.2 23.0 / 42.2 36.0 27.4 8.8 /
No. RS Sector 25.1 20.6 18.2 15.3 10.0 13.4 10.9 8.8 6.8 4.9
per cell Omni 16.0 12.7 10.4 5.3 / 7.5 5.8 4.0 1.1 /
No. RG per Sector 2.9 2.2 1.9 1.7 1.2 / / / / /
sect/cell Omni 3.2 2.4 1.9 1.0 / / / / / /
No. RS Sector 8.8 9.5 9.6 9.0 8.6 / / / / /
per RG Omni 5.1 5.3 5.4 5.3 / / / / / /
in a number of different randomly generated scenarios. The
results show that both sectorized systems and those based
on omni directional antennas can gain from the use of relays,
although higher gains are achievable in the case of omnidi-
rectional antennas. The results show that transparent mode
RSs can deliver significant gains in a number of different con-
texts, with increasing numbers of RSs delivering increasing
gains in all cases. It was found that increasing the number
of candidate sites does have some impact on the results and
this should be borne in mind when planning such networks.
The choice of RS transmit power also involves a trade-off
– larger power gives rise to more interference, but smaller
power means that the RS can only serve a small number
of SSs. In this work, it was found that higher power RSs
deliver greater gains and should be used more frequently.
7. ACKNOWLEDGMENTS
Y. Yu would like to acknowledge the support of the UCD
Ad Astra scheme for this work. S. Murphy would like to
acknowledge the support of the EU FP7 CARMEN project.
8. REFERENCES
[1] IEEE 802.16 Work Group. IEEE standard for local and
metropolitan area networks part 16: Air interface for
broadband wireless access systems. May 2009.
[2] Pavel Mach and Robert Bestak. Wimax throughput
evaluation of conventional relaying. Journal
Telecommunication Systems, 38(1-2):11–17, March
2008.
[3] Hui Zeng and Chenxi Zhu. System-level modeling and
performance evaluation of multi-hop 802.16j systems.
In Wireless Communications and Mobile Computing
Conference, 2008. IWCMC ’08. International, pages
354–359, 2008.
[4] E. Visotsky, Junjik Bae, R. Peterson, R. Berryl, and
M. L. Honig. On the uplink capacity of an 802.16j
system. In Wireless Communications and Networking
Conference, 2008. WCNC 2008. IEEE, pages
2657–2662, 2008.
[5] Vasken Genc, Se´an Murphy, and John Murphy.
Analysis of transparent mode IEEE 802.16j system
performance with varying numbers of relays and
associated transmit power. In Wireless
Communications and Networking Conference, 2009.
[6] G.S. Paschos, P. Mannersalo, and T.M. Bohnert. Cell
capacity for IEEE 802.16 coverage extension. pages
933–937, Jan. 2008.
[7] Karthikeyan Sundaresan and Sampath Rangarajan. On
exploiting diversity and spatial reuse in relay-enabled
wireless networks. In MobiHoc ’08: Proceedings of the
9th ACM international symposium on Mobile ad hoc
networking and computing, pages 13–22, New York,
NY, USA, 2008. ACM.
[8] IEEE 802.16j Multi-hop Relay Working Group.
Multi-hop relay system evaluation methodology
(channel model and performance metric). February
2007.
