interference mitigation and avoidance in uplink ofdma with collaborative distributed intracell...

5
Int. J. Electron. Commun. (AEÜ) 65 (2011) 937–941 Contents lists available at ScienceDirect International Journal of Electronics and Communications (AEÜ) jou rn al h omepage: www.elsevier.de/aeue Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling Elias Yaacoub , Zaher Dawy Department of Electrical and Computer Engineering, American University of Beirut, P.O. Box: 11-0236, Beirut, Lebanon a r t i c l e i n f o Article history: Received 14 November 2010 Accepted 10 March 2011 Keywords: OFDMA Uplink Distributed resource allocation Interference avoidance Interference mitigation a b s t r a c t A distributed collaborative uplink scheduling model in OFDMA systems is extended to a multicell sce- nario. On the intercell level, the scenarios with and without base station cooperation are investigated. In the cooperative case, base stations collaborate by performing pricing-based power control based on exchanging interference information. In the absence of intercell collaboration, probabilistic transmis- sion is applied in each cell for interference avoidance. Simulations show that the proposed intercell interference mitigation and avoidance schemes lead to enhanced results. © 2011 Elsevier GmbH. All rights reserved. 1. Introduction Distributed resource allocation has become an increasingly interesting research topic in beyond 3G and 4G OFDMA systems. In Ref. [1], a cooperative distributed intracell scheduling scheme is presented, where the users in each cell cooperate by exchang- ing channel state information (CSI) and implementing a distributed scheduling algorithm. To limit the overhead due to this exchange, an efficient CSI quantization method is used to reduce the number of required feedback bits. In a multicell scenario, intercell interfer- ence should be taken into account. In Ref. [2], a scheme based on base station (BS) collaboration is presented in order to apply the approach of Ref. [1] in a multicell scenario. Several techniques for reusing the radio frequencies are inves- tigated in the literature to limit the effects of intercell interference in multicell scenarios. Static reuse schemes are based on fractional frequency reuse (FFR) where a cell is divided into an inner area with the same frequencies reused in all cells and an outer area where a subset of the frequencies is reused, e.g., Ref. [3]. More efficient schemes consist of dynamic frequency reuse where all the frequen- cies are allowed to be used in all cells and elaborate techniques are applied for interference mitigation or avoidance. In Ref. [4], pricing is considered in ad hoc networks, where each user sets a price for other users to compensate for the interference they are causing. The prices are used as a sort of power control scheme to reduce transmission power. However, users are assumed to transmit on Corresponding author. E-mail addresses: [email protected] (E. Yaacoub), [email protected] (Z. Dawy). the same carriers and pricing is used for power control and not for scheduling. In Ref. [5], multicell uplink OFDMA scheduling is con- sidered. Pricing is imposed by the network and each user performs power control in a distributed manner using the pricing informa- tion. Pricing is applied on the total transmit power of the user, not on the power transmitted on each individual subcarrier. In this paper, we extend the results of Ref. [1] to a multicell scenario. Therefore, we propose a cooperative intercell interfer- ence mitigation scheme where the BSs in each cell cooperate by exchanging interference information. A pricing based power con- trol scheme is implemented using the exchanged information. The proposed scheme is implemented in conjunction with scheduling, and the prices are exchanged for each individual subcarrier. For multicell scenarios without BS collaboration, we propose a proba- bilistic interference avoidance technique. The proposed technique is based on shutting down each subcarrier with a probability that increases with the received interference level on that subcarrier. The proposed techniques are compared to the transparent pricing scheme presented in Ref. [2] and to the case without interference mitigation techniques. This paper is organized as follows. The system model and the problem formulation are presented in Section 2. The intercell inter- ference mitigation/avoidance schemes are discussed in Section 3. The simulation results are presented in Section 4. Finally, conclu- sions are drawn in Section 5. 2. System model The system studied consists of a set of BSs, each covering a small area. The reduced coverage area is assumed, as in Ref. [1], because 1434-8411/$ see front matter © 2011 Elsevier GmbH. All rights reserved. doi:10.1016/j.aeue.2011.03.004

Upload: elias-yaacoub

Post on 02-Sep-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling

Id

ED

a

ARA

KOUDII

1

iIiisaoeba

tiftascaioTt

1d

Int. J. Electron. Commun. (AEÜ) 65 (2011) 937– 941

Contents lists available at ScienceDirect

International Journal of Electronics andCommunications (AEÜ)

jou rn al h omepage: www.elsev ier .de /aeue

nterference mitigation and avoidance in uplink OFDMA with collaborativeistributed intracell scheduling

lias Yaacoub ∗, Zaher Dawyepartment of Electrical and Computer Engineering, American University of Beirut, P.O. Box: 11-0236, Beirut, Lebanon

r t i c l e i n f o

rticle history:eceived 14 November 2010ccepted 10 March 2011

a b s t r a c t

A distributed collaborative uplink scheduling model in OFDMA systems is extended to a multicell sce-nario. On the intercell level, the scenarios with and without base station cooperation are investigated.In the cooperative case, base stations collaborate by performing pricing-based power control based on

eywords:FDMAplinkistributed resource allocation

nterference avoidance

exchanging interference information. In the absence of intercell collaboration, probabilistic transmis-sion is applied in each cell for interference avoidance. Simulations show that the proposed intercellinterference mitigation and avoidance schemes lead to enhanced results.

© 2011 Elsevier GmbH. All rights reserved.

nterference mitigation

. Introduction

Distributed resource allocation has become an increasinglynteresting research topic in beyond 3G and 4G OFDMA systems.n Ref. [1], a cooperative distributed intracell scheduling schemes presented, where the users in each cell cooperate by exchang-ng channel state information (CSI) and implementing a distributedcheduling algorithm. To limit the overhead due to this exchange,n efficient CSI quantization method is used to reduce the numberf required feedback bits. In a multicell scenario, intercell interfer-nce should be taken into account. In Ref. [2], a scheme based onase station (BS) collaboration is presented in order to apply thepproach of Ref. [1] in a multicell scenario.

Several techniques for reusing the radio frequencies are inves-igated in the literature to limit the effects of intercell interferencen multicell scenarios. Static reuse schemes are based on fractionalrequency reuse (FFR) where a cell is divided into an inner area withhe same frequencies reused in all cells and an outer area where

subset of the frequencies is reused, e.g., Ref. [3]. More efficientchemes consist of dynamic frequency reuse where all the frequen-ies are allowed to be used in all cells and elaborate techniques arepplied for interference mitigation or avoidance. In Ref. [4], pricings considered in ad hoc networks, where each user sets a price for

ther users to compensate for the interference they are causing.he prices are used as a sort of power control scheme to reduceransmission power. However, users are assumed to transmit on

∗ Corresponding author.E-mail addresses: [email protected] (E. Yaacoub), [email protected] (Z. Dawy).

434-8411/$ – see front matter © 2011 Elsevier GmbH. All rights reserved.oi:10.1016/j.aeue.2011.03.004

the same carriers and pricing is used for power control and not forscheduling. In Ref. [5], multicell uplink OFDMA scheduling is con-sidered. Pricing is imposed by the network and each user performspower control in a distributed manner using the pricing informa-tion. Pricing is applied on the total transmit power of the user, noton the power transmitted on each individual subcarrier.

In this paper, we extend the results of Ref. [1] to a multicellscenario. Therefore, we propose a cooperative intercell interfer-ence mitigation scheme where the BSs in each cell cooperate byexchanging interference information. A pricing based power con-trol scheme is implemented using the exchanged information. Theproposed scheme is implemented in conjunction with scheduling,and the prices are exchanged for each individual subcarrier. Formulticell scenarios without BS collaboration, we propose a proba-bilistic interference avoidance technique. The proposed techniqueis based on shutting down each subcarrier with a probability thatincreases with the received interference level on that subcarrier.The proposed techniques are compared to the transparent pricingscheme presented in Ref. [2] and to the case without interferencemitigation techniques.

This paper is organized as follows. The system model and theproblem formulation are presented in Section 2. The intercell inter-ference mitigation/avoidance schemes are discussed in Section 3.The simulation results are presented in Section 4. Finally, conclu-sions are drawn in Section 5.

2. System model

The system studied consists of a set of BSs, each covering a smallarea. The reduced coverage area is assumed, as in Ref. [1], because

Page 2: Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling

938 E. Yaacoub, Z. Dawy / Int. J. Electron. Commun. (AEÜ) 65 (2011) 937– 941

se of t

ttutdtiitseptps

m

s

P

˛

wpisaP

Fig. 1. System model in the ca

he users are required to be in a relatively close proximity so thathey can collaborate by exchanging information. Although we willse the term BS throughout the paper, a BS can represent in prac-ice: a BS serving a small coverage area, a remote antenna in aistributed BS system, an access point in a local area network, a cen-ral controller in a cognitive radio (CR) network, or a femto BS in anndoor scenario. We investigate interference mitigation techniquesn the scenarios with and without BS cooperation. An example withwo collaborating BSs is shown in Fig. 1. Letting Isub,kl

be the set ofubcarriers allocated to user kl in cell l, N the number of subcarri-rs, L the number of BSs, Kl the number of users in cell l, Pkl,i,l

theower transmitted by user kl over subcarrier i in cell l, Pkl

the totalransmission power of user kl, Pkl,max its maximum transmissionower, and Rkl

its achievable throughput, the maximization of theum of user utilities can be formulated as

axL∑

l=1

Kl∑kl=1

U(Rkl(Pkl

, Isub,kl)) (1)

ubject to

kl,i,l≥ 0; ∀kl = 1, . . . , Kl, i = 1, . . . , N, l = 1, . . . , L (2)

kl,i,l≥ 0; ∀kl = 1, . . . , Kl, i = 1, . . . , N, l = 1, . . . , L (3)

N

i=1

Pkl,i,l≤ Pkl,max; ∀kl = 1, . . . , Kl, l = 1, . . . , L (4)

Kl

kl=1

˛kl,i,l= 1; ∀i = 1, . . . , N, l = 1, . . . , L (5)

here U(Rkl) is the utility of user kl as a function of the through-

ut Rk, and ˛kl,i,l= 1 if subcarrier i is allocated to user kl in cell l,

.e., i ∈ Isub,kl; otherwise, ˛kl,i,l

= 0. Hence, (5) represents the con-traint that each subcarrier can be allocated to a single user only in

given cell during one transmission time interval (TTI). In addition,kl

represents a vector of the transmitted powers Pkl,i,lof user kl

wo cooperating base stations.

on each subcarrier in cell l. Hence, the objective is to perform util-ity maximization with respect to the optimization variables: thetransmit powers Pkl,i,l

and the subcarrier allocation variables ˛kl,i,l.

The throughput of user kl in cell l is given by:

Rkl(Pkl

, Isub,kl) =

N∑i=1

˛kl,i,lB

Nlog2(1 + ˇ�kl,i,l

) (6)

where B is the total bandwidth and ̌ represents the signal to noiseratio (SNR) gap. The variable ̌ indicates the difference between theSNR needed to achieve a certain data transmission rate for a prac-tical M-QAM system and the theoretical limit (Shannon capacity)[1,6]. It is given by ̌ = − 1.5/ln(5Pb), where Pb denotes the bit errorrate (BER). In addition, �kl,i,l

is the signal to interference plus noiseratio (SINR) of user kl over subcarrier i in cell l. It is given by:

�kl,i,l=

Pkl,i,lHkl,i,l∑L

j /= l,j=1

∑Kjkj=1˛kj,i,j

Pkj,i,jHkj,i,l

+ �2i,l

(7)

where Hkl,i,lis the channel gain of user kl over subcarrier i in cell

l, taking into account pathloss, lognormal shadowing, and Rayleighfading. In addition, �2

i,lis the noise power over subcarrier i in cell l.

Without loss of generality, we let �2i,l

= �2i

= �2.

3. Intercell interference mitigation

This section describes the techniques used for the mitigation ofintercell interference when cooperation between BSs is allowed, inaddition to the avoidance of intercell interference in the absenceof BS cooperation. Each BS computes a price reflecting the inter-ference level it is receiving from its neighbors on each subcarrier.The price cil computed by the BS of cell l to reflect the interferenceon subcarrier i is selected to take a value in the interval [0 1] and is

given by:

ci,l = min

(Ii,lPref

, 1

)(8)

Page 3: Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling

n. Com

wPtP4

I

ut

3

tpaHpsp

U

pHmTwcnotem

masStncy

P

wdRte∑t

P

E. Yaacoub, Z. Dawy / Int. J. Electro

here Ii,l is the interference on subcarrier i measured at BS l, andref is a power threshold above which the price is set to one, i.e.,he BS in cell l does not tolerate more interference on subcarrier i.ref is determined according to the approach described in Section. The expression of the interference is given by:

i,l =L∑

j /= l,j=1

Kj∑kj=1

˛kj,i,jPkj,i,j

Hkj,i,l(9)

It should be noted that BS l does not need to know all the individ-al terms in the double summation of (9). It only needs to measurehe total received interference power on subcarrier i.

.1. Intercell cooperation: pricing based power control scheme

In this section, we present a power control scheme based onhe pricing information exchanged between the BSs. Users areenalized in the scheduling algorithm by reducing their utility byn amount that increases with the interference they are causing.ence, a transmission “price” is included in the utility in order toenalize high power transmissions. For greedy sum-throughputcheduling, the following utility is used instead of the through-ut:

(Rkl(Pkl

, Isub,kl))

=N∑

i=1

˛kl,i,lB

N

⎛⎝log2(1 + ˇ�kl,i,l

) −L∑

j /= l,j=1

ci,jPkl,i,l

⎞⎠ (10)

Hence, (10) represents the throughput scaled down by theower cost, following the concepts presented in Refs. [4] and [5].owever, conversely to Ref. [5], a price is imposed on the trans-it power on each subcarrier, not on the total transmit power.

his is in line with the approach of Ref. [4] for ad hoc net-orks. However, the utility is used here in greedy scheduling,

onversely to Ref. [4] where scheduling is not used. It should beoted that in the case where the interference is negligible, the sec-nd term in (10) goes to zero, and the problem becomes equivalento a sum-rate maximization problem without pricing, where inach cell, the single cell scheduling approach of Ref. [1] is imple-ented.With the power control scheme, the price

∑Lj /= l,j=1ci,j of trans-

itting on subcarrier i in cell l is not embedded in the pilot powers in the transparent pricing scheme of Ref. [2]. Hence, usershould be informed of the price of transmitting on each subcarrier.ince the prices are known at the BS, they can be communicatedo the users on the downlink via an appropriate control chan-el. The power on each subcarrier when power control is usedan be obtained by setting the derivative of (10) to zero. Thisields:

kl,i,l= [

1

ln(2)(∑L

j /= l,j=1ci,j

) − 1

ˇHkl,i,l/�2

i

]+

(11)

here [y]+ = max(0, y). In this case, users in each cell performistributed scheduling according to the algorithm presented inef. [1], but with the power allocation performed accordingo (11) instead of equal power allocation over the subcarri-rs as in Ref. [1]. If, for certain users, the result is such that

Ni=1Pkl,i,l

> Pkl,max, the power constraint is enforced by set-

ing:

kl,i,l= Pkl,i,l

·Pkl,max∑Ni=1Pkl,i,l

(12)

mun. (AEÜ) 65 (2011) 937– 941 939

The solution to ˛kl,i,lis obtained as described in the single cell

algorithm presented in Ref. [1]: the algorithm consists of allocatingsubcarrier i to user kl in a way to maximize the difference:

�kl,i,l= U(Rkl

(Pkl, Isub,kl

∪ {i})) − U(Rkl(Pkl

, Isub,kl)) (13)

where the marginal utility, �kl,i,l, represents the gain in the utility

function U when subcarrier i is allocated to user kl (i.e. when ˛kl,i,l=

1), compared to the utility of user kl before the allocation of i. Sub-carrier i is allocated to the user k∗

lsuch that k∗

l= argmaxkl

�kl,i,l,

i.e., ˛k∗l,i,l = 1, and ˛kl,i,l

= 0 for all kl /= k∗l.

In the case without interference, the prices are set to ci,j = 0.However, to avoid division by zero in (11) during numerical imple-mentation, we set ci,j = ε instead, with ε a very low value (e.g.,ε = 2.2 × 10−16 in Matlab). In this case, applying (12) after (11)leads to equal power allocation over the subcarriers allocated touser kl, which retrieves the result of Ref. [1] in the single cell sce-nario.

The drawback of the power control approach is that it requiresthe user utility to be “channel separable” [4]. This is easily achievedin the case of greedy maximization, since the user utility is the sumof the user throughput on the individual subcarriers. However, dueto the logarithm operation in proportional fair (PF) scheduling [6],the utility is not channel separable except in the special case ofallocating a maximum of one subcarrier to each user. Hence, thepower control approach will only be investigated in the case ofgreedy scheduling.

3.2. Interference avoidance in the absence of intercellcooperation: probabilistic transmission

The proposed intercell interference avoidance technique isbased on probabilistic transmission without BS cooperation. Inthe proposed scheme, each BS measures the received interfer-ence power on each subcarrier. Then, each BS l computes a priceci,l on each subcarrier i based on the measured interference, with0 ≤ ci,l ≤ 1, according to (8) and (9). Since ci,l is selected to take avalue in the interval [0 1], it can be used as a probability mea-sure. Hence, at each scheduling interval, the BS in cell l decidesto stop using subcarrier i with probability ci,l, or continues using itwith probability (1 − ci,l). Consequently, when the interference ona given subcarrier increases, its chances of being shut down alsoincrease. The intracell scheduling algorithm presented in Ref. [1]is then implemented in each cell on the subcarriers that are stillon.

4. Results and discussion

We consider a scenario consisting of two BSs, representing abroadband wireless access hot spot where the proposed schemeis implemented. In each cell, users are located within a dis-tance of Rc = 100 m from the BS. This range is relatively limitedin order to allow the signals transmitted by users on the celledge to be within the reference sensitivity level of users at thediametrically opposed edge (the reference sensitivity level is dis-cussed in Ref. [7]). Scenarios with four and eight users per cellare studied. Due to space limitations, and since the same con-clusions are reached in both cases, only the results of the fourusers case will be presented. We adopt the simulation modelof the distributed scheduling scenario in Ref. [1]. The thresholdpower Pref is considered to be the power received from a usertransmitting at its maximum power from a reference distance dref

from the BS, when only pathloss is present (shadowing and fad-ing are not considered in Pref in order to have a constant valuesince it is a power threshold). We assume that the channel gainremains approximately constant during the duration of 10 TTIs.
Page 4: Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling

940 E. Yaacoub, Z. Dawy / Int. J. Electron. Co

1 2 30

20

40

60

80

100

120

Number of CSI bits (last value corresponds to full CSI)

Sum

−T

hrou

ghpu

t (M

bps)

No interference mitigationTP − d

ref=0.5R

c

TP − dref

=Rc

PC − dref

=0.5Rc

PC − dref

=Rc

Fig. 2. Sum-throughput in the greedy scheduling case with four users.

Atbc[

itooab

Fig. 3 shows that the best results with PF scheduling are achieved

t each TTI, the two BSs measure the interference and computehe pricing information, then scheduling is performed in each cellased on the proposed techniques. The proposed techniques areompared to the transparent pricing method presented in Ref.2].

Fig. 2 shows the sum-throughput results for greedy schedul-ng using the proposed power control scheme. The results of theransparent pricing scheme of Ref. [2] are denoted by TP, thosef the power control scheme are denoted by PC, and the results

btained without applying any interference mitigation techniquere represented by “-”. The results for different CSI quantizationits according to the approach of Ref. [1] are presented.

1 2 0

2

4

6

8

10

12

Number of CSI bits (last v

Thr

ough

put P

rodu

ct

1 2 0

5

10

15

20

Number of CSI bits (last v

Ave

rage

num

ber

of a

lloca

ted

subc

arrie

rs

Fig. 3. Throughput product and allocated s

mmun. (AEÜ) 65 (2011) 937– 941

In Fig. 2, both the TP and PC interference mitigation schemesprovide enhancements over scheduling without interference miti-gation. It can be seen that all PC results outperform the case withoutinterference mitigation, and outperform the TP approach of Ref. [2]when the number of feedback bits is greater than one. For the caseof one CSI bit, the CSI quantization is too coarse for the PC approachto outperform the case without interference mitigation. In fact, 1bit CSI quantization leads to a rough estimate of the first term of(10) corresponding to the throughput, whereas the second termcorresponding to power control is not quantized. This explains thebehavior of the PC approach in Fig. 2 for 1 bit CSI, where it is outper-formed by the case without interference mitigation, which in turnis outperformed by the TP approach. However, when the number ofbits increases, the results of the PC approach become closer to thefull CSI value, and hence the accuracy of the quantization becomessufficient for PC to outperform both the TP approach in addition tothe case without interference mitigation.

Fig. 3 shows the comparison with TP of the results of PF schedul-ing using the proposed probabilistic scheduling (denoted by PS)scheme. Since with PF the utility used is the logarithm of thethroughput [6], the slow increase of the log function will not reflectthe enhancement obtained by the proposed approach, converselyto the product of the throughput, which is equivalent to the sumof logarithms when used as a utility, as discussed in Ref. [1]. There-fore, the upper part of Fig. 3 shows the product of the throughput,in Mbps, of the users in both cells. The values are multiplied by 10−6

in order to avoid excessively large numbers. The lower part of Fig. 3shows the average number of allocated subcarriers per cell per TTI,out of 16 subcarriers.

when the number of allocated subcarriers is the least. This indicatesthat the proposed PS scheme was able to enhance the total networkutility by shutting down the subcarriers corresponding to relatively

3

alue corresponds to full CS I)

No interference mitigationTP − d

ref=0.5R

c

TP − dref

=Rc

PS − dref

=0.5Rc

PS − dref

=Rc

3alue corresponds to full CS I)

ubcarriers in the PF scheduling case.

Page 5: Interference mitigation and avoidance in uplink OFDMA with collaborative distributed intracell scheduling

n. Com

hc

5

tltsioale

A

fp

[

[

[

[

[

[6] Lim J, Myung HG, Oh K, Goodman DJ. Proportional fair scheduling of uplink

E. Yaacoub, Z. Dawy / Int. J. Electro

igh interference and allocating the power on the remaining sub-arriers.

. Conclusions

Distributed uplink scheduling in OFDMA systems was inves-igated in the case of intracell user cooperation. On the intercellevel, the scenarios with the presence and absence of base sta-ion collaboration were both studied. Cooperation between basetations was implemented using the exchange of interferencenformation via pricing. In the absence of base station collab-ration, a probabilistic scheme was proposed for interferencevoidance. Both the cooperative and non-cooperative schemesed to enhanced results in the presence of intercell interfer-nce.

cknowledgements

The authors would like to thank the anonymous reviewersor their comments that helped in enhancing the quality of theaper.

[

mun. (AEÜ) 65 (2011) 937– 941 941

This work was supported by the American University of Beirut(AUB), the AUB Research Board, Dar Al-Handassah (Shair and Part-ners) Research Fund, and the Rathman (Kadifa) Fund.

References

1] Yaacoub E, Dawy Z. Achieving the nash bargaining solution in OFDMA uplinkusing distributed scheduling with limited feedback. International Journal ofElectronics and Communication AEUE (Elsevier) 2011;65:320–30.

2] Yaacoub E, Dawy Z. A transparent pricing scheme for interference mitigationin uplink OFDMA with collaborative distributed scheduling. In: InternationalConference on Telecommunications (ICT 2010), April. 2010.

3] Fujii H, Yoshino H. Theoretical capacity and outage rate of OFDMA cellular systemwith fractional frequency reuse. In: IEEE VTC-Spring. 2008.

4] Huang J, Berry RA, Honig ML. Distributed interference compensation forwireless networks. IEEE Journal on Selected Areas in Communications2006;25(May):1074–84.

5] Jing Q, Zheng Z. Distributed resource allocation based on game theory in multi-cell OFDMA systems. International Journal of Wireless Information Networks2009;16:44–50.

single-carrier FDMA systems. In: IEEE PIMRC, September. 2006.7] 3rd Generation Partnership Project (3GPP): 3GPP TS 25.101 UMTS User Equip-

ment (UE) radio transmission and reception (FDD), version 8.1.0, Release 8;2008.