2011 IEEE Symposium on Computers & Informatics

Resource Allocation Algorithm for WiMAX Systems

Mr. Wankhede Vishal A. Mr. Jha Rakesh Prof. Dr. Upena Dalal
Electronics Department Electronics Department Electronics Department
SVNIT / SNJBs COE SVNIT SVNIT
Surat / Chandwad, India Surat, India Surat, India
[email protected] [email protected] [email protected]

Abstract division multiple access (OFDMA), where subcarriers and

time slots are allotted to different users. To increase multi
WiMAX presents a very challenging multiuser communication user diversity designer have to increase freedom in
problem: Different users in same geographic area requiring scheduling the users. OFDMA come with a few outlay like
high on-demand data rates in a finite bandwidth with low overhead in downlink and uplink as base station(BS) needs
latency. To make resource allocation more practical, in mobile channel information for its users , and Mobile station(MS)
WiMAX instead of individual subcarriers a group of them is needs to know which subcarrier it has been assigned.
allocated to different users in both frequency and time. There Multiple access techniques like Frequency division multiple
are several ways to allocate subchannels in mobile WiMAX,
access (FDMA) , Time Division Multiple Access (TDMA)
three for downlink:FUSC (Fully Utilised Subchannelisation),
PUSC (Partially Utilised Subchannelisation) and AMC
and Code Division Multiple Access (CDMA) entails its own
(Adaptive Modulation and Coding). This paper aims to study pros and cons so recompense of OFDMA is that best features
the performance of low complexity adaptive resource of each technique can be achieved. OFDMA is hybrid of
allocation in downlink of OFDMA for WiMAX systems with FDMA and TDMA where users are vigorously assigned
variable rate requirements. Objective of the proposed subcarriers (FDMA) in different time slots (TDMA). So
algorithm is to maximize total throughput while maintining OFDMA comes with advantages like robust multipath
rate proportionality among the users. The proposed algorithm suppression and frequency diversity. In addition , OFDMA is
prioritizes the user with the high sensitivity to the subcarrier a flexible technique that can put up many users with varying
allocation. Simulation results shows that proposed method data rates and Quality of Service (QoS) requirements.
performs better than previous method in terms of significantly
decreasing the computational complexity and achieve higher To lay a hand on variety of traffic from different
capacities, while being applicable to a more general class of applications in WiMAX network , it is essential to find a
systems. resource management elucidation. Assigning subcarriers and
power to users in Mobile WiMAX system has recently been
Index terms— Downlink OFDMA, Subcarrier allocation , power an area of active research. In [1] novel downlink resource
allocation. management framework for multiservice WiMAX networks
was proposed where the demands of both WiMAX service
I. INTRODUCTION providers and subscribers are taken into account. For
optimization problems algorithms like fairness-constrained
Multiple-access techniques typically endeavor to provide greedy revenue for downlink adaptive power allocation and a
orthogonal , noninterfering , communication channels for
utility-constrained greedy approximation for downlink call
each established link. The most common way to divide the
admission control was developed in [1]. In [2] adaptive
available dimensions for different user is through the use of
time , frequency and also using code division multiplexing. bandwidth allocation and admission control scheme for
WiMAX has very challenging multiuser communication polling service in WiMAX is proposed using game
problem means users in one geographic area needs high on- formulation. Solution for non cooperative game is
demand data reates in a finite bandwidth with very less formulated by Nash equilibrium for the amount of bandwidth
latency. Multiple-access techniques allow different users to offered to a new connection. Resource allocation algorithms
share the available bandwidth by allocating individual user are core part of Physical layer of WiMAX , their aim is to
small part of the total system resources. In WiMAX pursue efficiency in resource usage in terms of power
multiple-access problem is quite complicated as WiMAX consumption, reducing interference and capacity
traffic- VoIP, data transfer and video streaming- and for provisioning. Resource allocation issues is been covered in
system deployment- mobility , neighboring cells , [3]. OFDMA and Adaptive Modulation and Coding (AMC)
requirement of bandwidth efficiency. It is difficult to mode is provided to manage physical layer resources and
implement efficient and flexible multiple-access strategy to leaving scope for the development of algorithms in [3]. For
improve WiMAX system performance. OFDM is a “last mile” Internet access OFDMA-Time Division
modulation technique that creates independent streams of Duplexing (TDD) system was proposed in[4] to provide high
data which can be used by different users , but OFDM is not spectrum efficiency on both uplink and downlink channels .
a multiple-access tactic. For this reason Mobile WiMAX use
different approach which is known as orthogonal frequency

978-1-61284-691-0/11/$26.00 ©2011 IEEE 753

 Fig. 1. Block diagram of Downlink OFDMA for K Users.

In [5] problem of dynamic subcarrier allocation, pk,n Power assigned to particular user
adaptive power allocation, admission control and capacity B Bandwidth 1MHz
gk,n Channel gain of kth user in nth subcarrier --
planning is tackled using Hungarian method transforms for with Additive White Gaussian Noise
wireless metropolitan area networks. A low complexity (AWGN)
adaptive resource allocation in downlink of OFDMA σ2 =N0 B / N where N0 is noise spectral
systems with fixed and variable rate was proposed in [6]. In density.
N0 Noise Power spectral density
[6] total throughput of the system increases while
hk,n Subchannels signal-to-noise ratio (SNR) gk,n2/ σ2
maintaining rate proportionality among the users, also Ptot Total Power 1W
priority is assigned to users to allocate more resources. Rk Total data rate for user
Φ Proportionality constant
II. SYSTEM MODEL
Block diagram for resource allocation in OFDMA is For proposed system kth user’s received SNR on
shown in Fig. 1. Proposed system is only applicable for subcarrier n is defined as γk,n=pk,n hk,n. For proposed system
downlink traffic. In the downlink Base Station (BS) of wireless channel is modeled as time varying frequency
WiMAX system applies the combined subcarrier and power selective Rayleigh channel .
allocation algorithms. BS assign subcarriers to live users In downlink transmitter of typical OFDMA system K
and the number of bits per OFDM symbol from each user to users information bits are allocated to N subcarriers and
be transmitted on each subcarrier. The Power calculation for each subcarrier n of user k following constraint 1≤k≤K is
each subcarrier is depends on number of assigned bits and assigned a power pk,n. While subcarriers are allotted to
type of modulation scheme used at that instant. Table I different users allotted subcarrier shouldn’t be allot to
shows the list of notations used to design mathematical another users till that user finish his work. For this purpose
model of the system with values used in our simulation. control channel is used which is responsible for keeping
track of allotted subcarriers to different users. E.g. User 1
TABLE I. NOTATIONS USED TO DESIGN MATHEMATICAL MODEL decodes allotted subcarriers only. We follow two
Notation Description Value assumption for proposed system first is that each user
K Number of Users 16 experiences independent fading and second is that each user
N Number of Subcarriers 64 is able to estimate the channel perfectly also these estimates
n Each or Individual Subcarrier (n) , 1 to 64 are made known to transmitter through feedback network.
following 1≤n≤N
k Individual user following 1≤k≤K 1 to 16
Channel gain of user k in subcarrier n is denoted as gk,n with

754
AWGN. At the end resource allocation algorithms receive and j ϵ {1,2,3,….,K}
channel estimates as a input. i≠j for k=1,2,3,…,K.
It is assumed that systems can give service to K number where R is the data rate for the user and Φ is proportionality
of users with N subcarriers. Frequency selective Rayleigh constants only one thing should keep in mind that sum of
fading channel is modeled. Additive white Gaussian proportionality constant for users in the range 1 to K should
noise(AWGN) with power spectral density of Noise be equal to one. It is simply like a probability for different
σ2=N0B/N. Bit error rate can be approximated to BER≤10-3. codes which we want to transmit, sum of probability of all
Resource allocation allow each user to choose better available codes is equal to 1.
subcarrier for him. Number of bits transmitted in downlink Total data rate (Rk) for kth user given as
can be formulated as, B N
γ k ,n Rk = ∑ ck , n rk , n
N n =1
(3)
rk ,n = log 2 (1 + ) = log 2 (1 + p k ,n H k ,n ) (1)
Γ
Where Γ =-1*ln(5BER)/1.6 is a constant signal to noise III. PROPOSED ALGORITHM
ratio gap and Hk,n=hk,n/ Γ is the effective subchannel or We utilize a combination of the approaches in [7,8,9] to
group of subcarrier’s signal to noise ratio. The resource develop the nature of the problem also we are trying to
allocation algorithms objective formulated as follows where reduce complexity while preserving performance of the
ck,n acts as an indicator for subcarrier allocation so that when system. Following are the steps of proposed algorithm.
ck,n=1 then only subcarrier n is assigned to user k. a. Calculate number of subcarriers Nk assigned to
B K N each user.
max max
c k ,n , p k ,n
∑
N k =1
∑c
n =1
k ,n log 2 (1 + pk , n H k , n ) ( 2) b. Assign subcarriers to each user to achieve
proportionality.
c. Assign total power Pk to each user maintaining
To follow aim in (2) system have to strictly pursue proportionality.
following constraints d. Assign pk,n for each users subcarriers subjected to
Constraint 1 : ck,n ϵ {0 , 1} for all values of k and n. Pk constraints.
First constraint tells to algorithm that if subcarrier is In proposed algorithm proportionality constraints are
allocated to particular user then set value of ck,n to 1 so that used to differentiate different services so that service
another user cant get that subcarrier. If value of ck,n is 0 provider can give priority to specific users based on
means subcarrier can be taken by any user. different billing methods.
In proposed algorithm proportionality constraints are
Constraint 2: pk,n ≥ 0 for all values of n. used to differentiate different services so that service
Second constraint tells to algorithm that value of kth users nth provider can give priority to specific users based on
subcarrier’s power should always be greater than or equal to different billing methods. Assuming a flat transmit power
0. over the entire bandwidth, each subcarrier adds fraction of
Constraint 3: total power equal to Ptot/N. Proposed algorithm aims to
K maximize throughput per user with lesser complexity.
∑c
k =1
k ,n =1 for all values of n. Instead of prioritizing the user with lesser throughput the
user with highest data rate to achieve is giving priority to
Third constraint tells to algorithm that allocate kth users nth choose available subcarrier. Above a to g steps can be
subcarrier to user ranges in between k=1 to K only once. represented mathematically as follows.

Constraint 4: Step 1:
K N To satisfy proportionality constraints subcarriers assigned to
∑∑ c k ,n p k ,n ≤ Ptotal each user should approximately same as data rates after
k =1 n =1 power allocation.
Fourth constraint tells to algorithm that sum of the product N1:N2:-----:NK=Φ1: Φ2:----- ΦK.
of subcarriers allotted to user ranges in between 1 to K and Nk=[ ΦkN]
power allotted to each user’s allotted subcarriers should be K
less than or equal to total available power. This constraint
This may lead to N var = N − ∑k =1 N k , where Nvar
also called as power constraint. means unallocated subcarriers. Second step takes care of
how Nk subcarriers for kth user and Nvar subcarriers.
Constraint 5:
Ri Φ Step 2: Initialization
= i , for all values of i. ck,n=0, for all the values of k ϵ {1,…,K} and for all values of
Rj Φ j n ϵ {1,…,N}.

755
Rk=0, for all values of k ϵ {1,…,K}
P=Ptot/N2.
Set of Number of Subcarriers (Nset)={1,2,…,N}

Step 3: Running Iterations from k=1 to K.

Sort Hk,n in ascending order
n=arg max n ϵ Nvar1 | Hk,n |
ck,n=1
Nk=Nk-1 , N set=Nset/{n}
B
Rk = Rk + log 2 (1 + pH k , n ) ( 4)
N
Step 4: Define Data Rate and calculation of unallocated
subcarriers.
In for loop’s condition check for || Nset || > Nvar.
Kset={1,2,…K}
k=arg min k ϵ Kset Rk/ ΦK.
n=arg max n ϵ Nset |Hk,n|
if Nk > 0
ck,n =1
Nk=Nk-1 , Nset=Nset/{n}
B
Rk = Rk + log 2 (1 + pH k ,n )
N
else
Kset=Kset/{k} Fig. 2. Flow chart for the simulated model

Step 5: Allocate remaining subcarriers.

Kset={1,2,…K}
for n=1 to Nvar
k=arg max k ϵ Kset |Hk,n|
ck,n=1
B
Rk = Rk + log 2 (1 + pH k ,n )
N
Kset=Kset/{k}
For power allocation among users we use same as in
[10,11]. Waterfilling across the subcarriers is been
performed. The flowchart of the proposed algorithm is
shown in Fig. 2. Above steps number 1 to 5 explained in
easy way using flowchart.
IV. SIMULATION RESULTS
The frequency selective multipath channel is modeled Fig. 3 Channel with only two users with eight subchannels.
of six Rayleigh multipaths. Doppler frequency of 50Hz is
TABLE II. THE CHANNEL CHARACTERISTICS OF A TWO USER SYSTEM
assumed with delay spread of channel 5µs. The total power SHOWN IN FIG. 3.
was assumed to be 1W and total bandwidth as 1MHz. Rest
of the parameters are already been mentioned in section II. User Index 1 2
Minimum channel gain 0.55 0.15
As shown in the Fig. 3., the channel gains of User 1, Maximum channel gain 0.66 0.33
have a less amount of channel gain from 0.55 to 0.66 i.e. Variance 0.11 0.18
0.11 while for User 2 channel gains varies from 0.15 to 0.33
i.e. 0.18 for first two subchannels. This information is Fig. 4 shows the normalized proportions of the
shown in following TABLE II. This is due to multiuser normalized rate for each user averaged over 100 channel
diversity as each user experiences different channel fading. samples. The normalized capacities are given by
However, how different users undergo fading could be 16
measured by their channel gain variance as shown in R/ ∑ k =1
Rk and observed for both proposed algorithm and
TABLE II.

756
ROOT-FINDING algorithm. This result is compared to the

normalized proportional constraints

{φ }16
k k =1 . It is been

observed that proportionality among the users for the

proposed method is no longer being strictly enforced as
compared to ROOT-FINDING.

Fig. 5 Users average capacity versus number of users. There are 64

subcarriers and BER=10-3.

V. CONCLUSION AND FUTURE SCOPE

The problem of multiuser downlink resource allocation
Fig.4 Normalized capacity ratios per user for SNR=30dB, K=16 and in Mobile WiMAX system was studied. Simulation results
BER=10-3. show that information fed back and association among
subcarriers play vital role to improve system performance.
Fig. 5. show the comparison of total capacities between A reduced complexity resource allocation was proposed and
the proposed method and ROOT-FINDING. As the number seems especially attractive as the number of users increases.
of users increases capacities increases. This is the effect of It is observed through simulation that the proposed method
multiuser diversity gain, which is outstanding in systems performs better than previous method in terms of
with bigger number of users. The proposed algorithm has a significantly decreasing the computational complexity and
consistently higher total capacity than the ROOT-FINDING achieve higher capacities, while being applicable to a more
method for all the numbers of users for this set of simulation general class of systems. As a possible future work, our
parameters. Normalized capacity ratios per user for 16 users algorithm , could be tested using Fuzzy logic and Genetic
averaged over 100 channels, proportions Φk taken as input algorithms. The fairness issue for the different user in the
for both algorithms for each user. The proposed algorithm same service class is another topic in our research schema.
has minimal deviation from the required proportions, but The effects of imperfect channel state information due to
ROOT-FINDING stick on to it much better. Proposed estimation error and feedback latency are also comes under
algorithm achieves the best performance for the most future study.
underserved user, with a slight gain for optimal power
allocation over its allocated subcarriers relative to an equal-
power allocation. Also proposed algorithm results in a REFERENCES
starved worst-case user; in fact , it is typical for several
[1] Bo Rong, Yi Qian and Kejie Lu, “Integrated Downlink Resource
users to receive no resources at all for substantial periods. Management for Multiservice WiMAX Networks”, IEEE
Transactions On Mobile Computing, VOL.6 , NO.6, JUNE 2007.
TABLE III. SUMMARY OF PERFORMANCE COMPARISON [2] Dusit Niyato and Ekram Hossain, “Radio Resource Management
Games In Wireless Networks: An Apporach to Bandwidth Allocation
and Admission Control for Polling Service in IEEE 802.16”, IEEE
Performance Criterion Proposed ROOT-FINDING Wireless Communications, February 2007.
Algorithm [3] Leonardo Badia, Andrea Baiocchi ,Simone Merlin, Silvano Pupolin,
Subcarrier Allocation O(KNlog2N) O(KNlog2N) Andra Zanella and Michele Zorzi, “On The Impact of Physical Layer
Complexity Awareness on Scheduling and Resource Allocation in Broadband
Power Allocation O(K) O(nK),n≈9 Multicellular IEEE 802.16 Systems”, IEEE Wireless Communication,
Complexity February 2007.
Achieved High Low [4] Bo Rong and Yi Qian, “Adaptive Power Allocation and Call
Capacity Admission Control in Multiservice WiMAX Access Networks”, IEEE
Adherence probability Slack Stretched Wireless Communications, February 2007.
Assumption on subchannel None Required [5] Syed Hussain Ali, Ki-Dong Lee and Victor C.M. Leung, “Dynamic
SNR Resource Allocation in OFDMA Wireless Metropolitan Area
Networks”, IEEE Wireless Communications, February 2007.

757
[6] Sanam Sadr, Alagan Anpalagan and Kaamran Raahemifar, [9] Fen Hou, James She, Pin-Han Ho and Xuemin Shen, “A Flexible
“Suboptimal Rate Adaptive Resource Allocation for Downlink Resource Allocation and Scheduling Framework for Non-real-time
OFDMA Systems”, Hindawi Publishing Corporation, International Polling Service in IEEE 802.16 Networks” , IEEE Transactions On
Journal of Vehicular Technology, Volume 2009. Wireless Communications, VOL. 8, NO. 2, FEBRUARY 2009.
[7] Mo-Han Fond, Robert Novak, Sean McBeath and Roshni Srinivasan, [10] Jianwei Huang, Vijay G. Subramanian, Rajeev Agrawal and Randall
“Improved VoIP Capacity in Mobile WiMAX Systems Using A. Berry, “Downlink Scheduling and Resource Allocation for
Persistent Resource Allocation”, IEEE Communications Magazine, OFDM Systems.”, IEEE Transactions On Wireless Communications,
October 2008. VOL. 8, NO. 1, JANUARY 2009.
[8] Spyros Xergias, Nikos Passas and Apostolis K. Salkintzis, [11] Nassar Ksairi and Walid Hachem, “Resource Allocation for
“Centralized Resource Allocation for Multimedia Traffic in IEEE Downlink Cellular OFDMA Systems-Part I: Optimal Allocation”,
802.16 Mesh Networks”, Proceedings of the IEEE, Vol. 96, No.1 , IEEE Transactions on Signal Processing, Vol. 58, No. 2, February
January 2008. 2010.

758