QoS Aware Adaptive QoS Aware Adaptive Subcarrier Allocation Subcarrier Allocation in OFDMA Systemsin OFDMA Systems

Mustafa Ergen & Sinem Coleri

{ergen,csinem}@eecs.berkeley.edu

University of California Berkeley

Introduction

Motivation Orthogonal Frequency Division Multiple Access(OFDMA) OFDMA System Resource Allocation Problem Algorithms

Optimal Suboptimal

Simulation Conclusion

Motivation

Broadband Wireless AccessEx: IEEE 802.16, Wireless MAN

OFDMEliminates InterSymbol Interference

OFDMA

OFDM Diagram

Multiuser OFDM

OFDM-TDMA OFDM-FDMA OFDMA

User 1

User 2

User 3

Subcarrier

Time

OFDM-TDMA

Subcarrier

Time

OFDM-FDMA

Subcarrier

Time

OFDMA

…

…

Resource Allocation

Goals: Dynamic subcarrier selection Improve system performance with adaptive

modulation More bits transmitted in large channel gain carriers

Provide QoS Rate and BER

Resource Allocation

Assumptions: Base station knows

the channel Base station informs

the mobiles for allocation

BaseStation

subcarrier

user

System

Application

Network

Resource Allocation

Physical Layer

rQoS=[rR,rBER] oQoS=[oR,oBER,oCoS]

oCoS=Ptotal for downlinkoCoS=Pu for uplink

[User x Subcarrier]

OFDMA

AWGNw(n)

Adaptive Modulation

IDFTGuard

InsertionP/S

Channel

+S/PGuard

RemovalDFT

Adaptive Demodulation

X(k)

Y(k)

x(n)

y(n)

h(n)

xf(n)

yf(n)

PathLoss

Channel Informationfrom

user kResouce Allocation Module

Subcarrierallocation

withDifferent

Modulation

SubcarrierExtraction

foruser

k

Adaptive Modulation

Adaptive Modulation

User 1 (Rate R1, BER1)

User 2 (Rate R2, BER2)

User K (Rate RK, BERK)

Maximum TotalPower

.

.

.

.

.

.

.

.

.

.

.

User k Adaptive Demodulation

Adaptive Demodulation

Resource Allocation

64-QAM

16-QAM

4-QAM

Channel

Subcarrier

Use

r

RATE: [12 6 6 8 ] BER: [1e-2 1e-2 1e-4 1e-4]

QoS

Resource Allocation

Notation

)12()4

(3

)(:

2,

:

},...,1,0{,

:

},...,1{:

},...,1{:

)(:

2

1

2,

,,

c

nk

nkkcnk

BERQ

NocfQAMM

nkgainchannel

Mnk

cbitassigned

Nnsubcarrier

Kkuser

cfPPowerTransmit

Optimal Integer Programming

K

k

M

ccnk

cnk

N

n

M

cnkk

cnkcnk

K

k

N

n

M

c nk

nk

nallforand

kallforcRtosubject

forcf

cnk

1 1,,

,,1 1

,

,,,,1 1 1

2,

,

.,10

,.

}1,0{)(

min,,

Subcarrier

Use

r

Subcarrier

Use

r

Subcarrier

Use

r

Subcarrier

Use

r

Pc2

Pc3

Pc1

Motivation for Sub-optimal Algorithms

IP is complex Allocation should be done within the

coherence time Time increases exponentially with the

number of constraints

Current Suboptimal Algorithms

2-step: Subcarrier Allocation

Assume the data rate for all subcarriersAssume modulation rate is fixedAssign the subcarriers

Bit LoadingGreedy approach to assign the bits of user

Current Suboptimal Algorithms Subcarrier Allocation

Hungarian algorithm Optimal, very complex

LP approximation to IP problem

Close to optimal

Bit Loading

Subcarrier

Use

r

Subcarrier

Use

r

Subcarrier

Use

r).(

1

)(minarg

:,

,,

,,

,,

nknk

nknk

nknkSn

k

cPevaluate

cc

cPn

timesRfollowingtherepeatkeachFor

k

Problems in Current Suboptimal Algorithms

Subcarrier assignment and bit loading are separatedUsers with bad channels may need higher

number of subcarriers Not iterative subcarrier assignment

Iterative Algorithm

Iterative algorithm based on Assignment of bits according to highest

modulation Finding the best places

Distributing the assigned bits to other subcarriers or to non-assigned subcarriers

Exchanging the subcarriers among user pairs for power reduction.

Iterative Algorithm

Fair Selection(FS) Greedy Release(GR) Horizontal Swaping(HS) Vertical Swaping(VS)

Iterative Algorithm

FAIR SELECTIONPtotal<Pmax

GREEDYRELEASE

Start

Modulation--

HO

RIZ

ON

TA

L S

WA

P

VE

RT

ICA

L S

WA

P

ASSIGNMENT ITERATION

Simulation Environment

Build the OFDMA system Modulations:4-QAM,16-QAM,64-QAM Independent Rayleigh fading channel

to each user Number of subcarriers =128 Nodes are perfectly synchronized

CDF of total transmit power without Pmax constraint

CDF of total transmit power with Pmax constraint

Average bit SNR vs. RMS delay spread

As RMS delay spread increases, the fading variation increases hence higher gains are obtained by adaptive allocation

Average bit SNR vs. number of users

As the number of users increases, the probability of obtaininggood channel at a subcarrier increases

Instantaneous Average bit SNR vs Time

Iterative Algorithm improves its Average Bit SNR by the time.

Conclusion

OFDMA Broadband Wireless Access

Resource Allocation Channel Information QoS Requirement

Optimal Algorithms complex

Iterative Algorithms