Interference Avoidance with Dynamic Inter-Cell Coordination for Downlink LTE System

Transcription

1 Inerference Avoidance wih Dynamic Iner-Cell Coordinaion for Downlink LTE Sysem Mahmudur Rahman # Halim Yanikomeroglu # and William Wong * # Deparmen of Sysems and Compuer Engineering Carleon Universiy Oawa Canada {mmrahmanhalim}@sce.carleon.ca Absrac The invesigaion of co-channel inerference miigaion echniques (such as inerference cancellaion hrough receiver processing inerference randomizaion by frequency hopping and inerference avoidance hrough resource usage resricions imposed by frequency and power planning has become a key focus area in achieving dense specrum reuse in nex generaion cellular sysems such as G LTE LTEadvanced and WiMAX. In his paper we propose an inerference avoidance scheme for LTE downlink ha uses dynamic iner-cell coordinaion faciliaed hrough X inerface among neighbouring evolved UTRA nodebs (ebs i.e. LTE base saions. roposed scheme is evaluaed by exensive simulaions and compared wih a number of reference schemes available in he lieraure. I has been observed ha he proposed scheme aains superior performance in erms of cell-edge and secor hroughpu compared o hose in he reference schemes. Key Words Inerference avoidance iner-cell coordinaion G LTE Hungarian algorihm OFDM OFDMA. I. ITRODUCTIO Orhogonal frequency division muliplexing (OFDM has been widely acceped as a promising air-inerface echnology for fuure generaion sysems by various sandardizaion bodies and forums for example he hird generaion parnership projec long erm evoluion (G LTE [] worldwide ineroperabiliy for microwave access (WiMAX [] and he wireless world iniiaive new radio (WIER []. Equipped wih OFDM s inheren robusness agains frequency selecive fading orhogonal frequency division muliple access (OFDMA offers flexibiliy for radio resource allocaion [4]. The finer resource granulariy of he OFDMA allows each resource uni o be allocaed and modulaed adapively o obain frequency as well as muliuser diversiy. In order o mee user demands in erms of daa-rae ubiquiously dense reuse of available specrum is envisaged in fuure sysems. The obvious pifall of such a dense reuse is he srong iner-cell inerference which limis nework as well as cell-edge hroughpu. To obain he full poenial of OFDMA in a dense reuse environmen appropriae radio resource managemen (RRM algorihms for inerference miigaion are necessary. Inerference miigaion is one of he key issues currenly under invesigaion in differen sandardizaion bodies and forums. Based on approaches used miigaion echnique is generally caegorized ino hree major classes i.e. i inerference cancellaion ii inerference averaging and iii inerference avoidance echniques. The basic principle of inerference cancellaion echnique is he receiver signal * Communicaions Research Cenre of Canada (CRC Oawa Canada william.wong@crc.ca processing o esimae inerference and subrac i from he desired signal componen. Inerference averaging echnique such as frequency hoping (FH ensures user equipmens (UEs o access a range of channels raher han a narrow se in a specific paern so ha inerference effec is averaged ou for all UEs. Finally he inerference avoidance echnique focuses on finding an opimal effecive reuse facor ofen achieved hrough resricions on frequency and power allocaions o achieve nework performance goals. The benefis of each of hese schemes are muually exclusive; herefore a combinaion of he above sraegies is expeced in fuure sysems. The focus of his paper is on inerference avoidance hrough dynamic iner-cell coordinaion. UEs a he cell border experience high inerference from neighbouring ransmiers in addiion o high pah-losses. Scheduling schemes wih objecive o achieving maximized nework hroughpu find hese UEs wih poor channel condiions less aracive as hey do no conribue much o he oal hroughpu. A blun approach o improve cell-edge UEs raes by assigning more resources (i.e. prioriizing cell-edge UEs in a cerain way would jeopardize he raes for oher UEs in he cell-cenre and hence such an approach is highly undesirable. The objecive of inerference avoidance is o provide beer services o cell-edge UEs wihou sacrificing cell-cenre hroughpu. In his paper an inerference avoidance scheme ha uses dynamic iner-cell coordinaion hrough X inerface is invesigaed. Iner-cell inerference can be reduced significanly wih he radiional frequency reuse sraegy as presened in he classical aricle [5] on cellular clusering. Fig. shows reuse of along wih reuse of and oher saic pariion-based schemes. The higher he cluser size he greaer he reducion in iner-cell inerference. However his improvemen in inerference can only be realized wih he reducion in cell hroughpu. Alhough such a radiional reuse scheme migh have been good enough o suppor raffic demands of he early neworks he rae requiremens in he fuure sysems warran more aggressive reuse. In recen years a large number of available sudies in he lieraure consider reuse pariioning in which cell-edge UEs are assigned resources wih higher reuse facors compared o he UEs a he cell-cenre o obain an effecive reuse which is somewha greaer han bu no oo high. In general hese schemes are referred o as fracional The erms cell and secor are used inerchangeably in his paper; iner-cell inerference implies he inerference power received from any secor anenna o a UE in a secor of ineres.

2 frequency reuse (FFR schemes. Evolved from he idea of classical reuse clusering FFR was firs inroduced in [6] for Global Sysem for Mobile (GSM sysems. Varians of FFR such as sof frequency reuse (SFR and parial frequency reuse (FR have been adoped in he WiMAX and G LTE sysems (see [7] and [8] for example. This idea is also explored exensively in he WIER projec [9]. While his approach improves inerference o he cell edge UEs here is a grea poenial o lose overall cell hroughpu due o resource loss resuled from pariioning. An opimal pariioning depends on he disribuion of he UEs arrived raffic and channel dynamism. Therefore any saic reuse pariioning scheme would be a highly sub-opimal soluion. A dynamic pariioning based on UEs raffic load as well as muual inerference siuaion may provide balanced improvemens for boh he cell-edge and cell-cenre UEs raes. Such a dynamic reuse adapaion requires dynamic iner-cell coordinaion. An inerference avoidance scheme wih dynamic iner-cell coordinaion is presened for he WIER sysem in [0]. The scheme requires a cenral eniy such as radio nework conroller (RC. Cenralized processing of RRM algorihms is no encouraged in LTE due o he absence of RC. However iner-cell coordinaion among neighbouring cells faciliaed hrough X inerface [] is suppored in LTE. We modify he scheme presened in [0] o sui o LTE downlink as well as furher improve by separaing handling of inerference originaed from he secors of own eb and ha from he oher ebs. In paricular we presen a novel approach o uilize Hungarian algorihm by devising uiliy marix in a muli-cell fashion o handle inercell inra-eb inerferers. The remainder of his paper is organized as follows. Secion II gives overview of he available saic schemes in he lieraure. The proposed scheme is described in Secion III. Secion IV provides LTE simulaion environmen and parameers. Simulaion resuls are discussed in V followed by conclusions in VI. II. STATIC ITERFERECE AVOIDACE SCHEMES A. Sof frequency reuse Sof frequency reuse scheme is a variaion of FFR where he reuse facor is and equal or greaer han in he cellcenre and cell-edge areas respecively. I was proposed in [] and [] under G LTE framework o provide a higher rae o disadvanaged UEs such as hose near he cell boundary. Fig..c shows an example SFR scheme for cell sies wih secorizaion. For -secor cell sies he cell edge band is usually / of he available specrum and is orhogonal o neighbouring cells. The cell-edge subcarriers are called major subcarrier group while he cell-cenre frequency band is ermed as minor subcarrier group. The oal ransmi power is se fixed and each group is assigned ransmission power depending on desired effecive reuse facor which is deermined by he power raio of cell-cenre o cell-edge groups. Transmissions use higher power on he major band as shown in he righ side of Fig..c. Le us consider ha he power per physical resource block (RB is in he case of reuse of and power per RB for he cell-edge (major band is α for he SFR scheme. Then power per RB in he minor band would be (-α/ giving a power raio of (-α/α. Minor band is available o cell-cenre UEs only and major band can also be used for cell cenre areas. Adjusing power raio from 0 o effecively moves he reuse facor from o. Therefore for a ri-secor cell SFR is a compromise beween reuse and. UEs are caegorized ino cell-edge and cellcenre based on user geomery. B. arial frequency reuse The idea of he parial frequency reuse (FR was firs presened in [4]. While somewha similar o SFR in erms of frequency planning he effecive reuse facor of his scheme is always greaer han. FR and is varians are sudied in he G and WIER projecs (see for example [9] and [5]. Fig..d shows an example of FR for sies wih secorizaion. Le us consider ha he oal bandwidh is β and i is divided o inner zone and ouer zone wih β and β respecively. For a reuse facor of in he ouer zone he frequency band assigned o each secor s ouer zone is β /. Therefore he effecive frequency reuse is given by β/( β + β /. Like SFR he power used for he ouer zone RBs can be amplified as shown in he figure. ower a Reuse ower c SFR d FR Fig. : Reuse and oher saic pariion based schemes ower b Reuse III. DESCRITIO OF E ROOSED SCHEME ower Iner-cell inerference is caegorized ino wo groups i.e. inra-eb and iner-eb inerference as shown in Fig.. UEs receive dominan inerference from he firs-ier of inerferers from own-eb and 4 from cells of oher neighbouring ebs as illusraed in he figure. Based on muual inerference siuaion and UE rae requiremens resource resricions are prepared using wo algorihms one for inra-eb and he oher for iner-eb inerferers. Inerference originaed from cells of own eb should be handled separaely as eb can ake appropriae measures iself wihou he need for inereb communicaion hrough X inerface. Boh algorihms involve preparaion of uiliy marix and applying Hungarian algorihm [6] on he uiliy marix in an ieraive manner in order o find RBs o be resriced in he neighbouring cells. Hungarian algorihm is opimal for one-o-one RB o UE

3 allocaion; however i becomes sub-opimal when i is used ieraively o assign more han one RB o a UE. The resricions on he usage of RBs are deermined from ime-o-ime a a ime-inerval wihin he channel coherence ime i.e. depending on he speed of he mobile. This inerval is denoed as resource resricion refresh inerval. Once he RB resricion lis is available a a secor he scheduler can perform RB scheduling based on is own crieria. We sudied wo variaions of he proposed scheme; resriced RBs are no used a all in one and hese RBs are used however only wih reduced power (for example wih 0 db lower in anoher variaion. Fig. : Inra and iner-eb iner-cell inerference A. Inra-eB iner-cell inerference avoidance For a paricular RB a secor can resric only one dominan inra-eb inerferer. A uiliy marix covering all hree secors of an eb is consruced as follows. Secor Secor Secor Secor Uinra = U U UU U U UU U ( where each elemen of he above composie marix is iself a marix of size M ; and M are he number of RBs and he number of UEs per secor respecively. The firs hree elemens are he uiliy marices for secors and respecively given remaining secors (i.e. {} {} and {} use RBs concurrenly. Subsequen subses of marices show uiliy when one of he inra-eb secors is resriced o use RBs. These uiliy marices condiioned on he possible concurren inra-eb iner-cell inerferers are given by: U i i i u u u u u u M M = u u u M u u u M The uiliy measure u nm is he produc of he achievable rae on RB n if i is assigned o UE m (r mn and he curren demand facor of UE m (d m when boh inra-eb inerferers are acive. The demand facor of a UE is defined as he raio of he received hroughpu o average secor hroughpu. When an inra-eb secor is resriced a penaly (in erms of rae is inroduced in he calculaion of he uiliy as below.. Tri-secor BS anenna Inra-eB inerference Iner-eB inerference ( rnm dm ; all ransmi unm = ( ( rnm rp d m;one resriced where r p is he rae penaly considered in he uiliy measure o accoun for resource loss due o resricion imposed o one of he inra-eb secors. In simulaions r p =.5 bps/hz is considered such ha average number RBs resriced by inraeb avoidance algorihm is around 0% of he sysem RBs. I should be noed ha in calculaing r mn all oher iner-eb inerferers are considered acive. Enries in U inra corresponding o UEs having demand facors less han (i.e. rae saisfied in he pas are se o zero so ha resricions are made only for he rae deprived UEs. Then he Hungarian algorihm is applied o U inra. In each ieraion algorihm selecs bes RBs for UEs in hree differen secors so ha he sum of uiliies of he chosen RBs is maximized. oe ha an enry from a marix wih resricion is chosen only when he rae improvemen due o inerference suppression exceeds he penaly r p. U inra is updaed afer each ieraion as follows. If a chosen enry is from he marix where RB resricion U uiliy enries for he is no required for example corresponding RB for all UEs in U U U U U and U in addiion o U would have o be replaced by zeros. However uiliy enries in U and U remain unchanged in order o allow fuure ieraion o reselec his RB. If a chosen enry is from he marix where a RB resricion is required for example U (secor has resricion on he seleced RB following acions are required. o lace he RB index in resricion lis for secor o Enries corresponding o his RB for all UEs in all marices excep U have o be replaced by zeros. In his case fuure ieraions will allow his RB o be used only by a UE in secor. The above seps are repeaed unil all enries in U inra are zero. In his process each secor prepares RB resricions for is inra-eb neighbours. The number of required ieraions varies depending on he number of rae deprived UEs as well he uiliy values of he marix. B. Iner-eB iner-cell inerference avoidance Similar o [0] each secor prepares RB resricions for iner-eb inerferers based on dominan received inerference from four iner-eb inerferers. This algorihm involves preparaion of uiliy marix U iner by using heurisics and applying Hungarian assignmen algorihm o his marix. Then neighbouring ebs are communicaed abou hese resource resricions over X inerface. The deails of his algorihm are discussed below. reparaion of uiliy marix ( i Le us consider rmn and r (min m n o be he achievable raes (max for UE m and RB n a secor i when none and all s-ier iner-eb inerferers are resriced respecively. However

4 moving from ( i rm n o r (min m n implies increasing penaly o he (max inerfering secors as more and more inerferers are o be resriced. A hreshold-based sraegy is used o deermine which inerferers are o be resriced. Based on is demand facor and channel condiions a UE can resric wo mos dominan inerferers a mos. This limis he number of resuling RB resricions in he neighbouring secors. In order o consruc iner-eb uiliy marix he following seps are repeaed for each UE and RB. Four iner-eb dominan inerferers are sored in descending order ino a dominan inerferer se. rm n(min is calculaed considering he presence of all inercell iner-eb dominan inerferers and aking inra-eb resricions ino accoun. o If rm n r (min no inerferers are o be resriced if RB n is assigned o UE m irrespecive of is demand facor. In his case UE m is hen eiher having a srong desired link from serving BS or is experiencing weak inerference from all dominan inerferers on RB n. In simulaions we have used r =. 5 bps/hz. o Else calculae he new rae r m n( new wih he mos dominan inerferer being resriced. If r r r UE m will reques m n( new m n(min his dominan inerferer o be resriced irrespecive of is demand. r = bps/hz has been used in simulaions. ( i Else if d m > (UE m has been rae derived in he pas r = 0 and r 0 m n(min m n > ( new he mos dominan inerferer is o be resriced. In some cases boh ( i r m n and r (min m n( new can be zero. If resricing wo mos dominan inerferers provides achievable rae hese will be marked for resricion. oe ha he above hreshold values are chosen such ha he number of RBs resriced by iner-eb algorihm is on average around 5-0% of he available sysem bandwidh. Afer finding he iner-eb dominan inerferer(s o be ( i resriced on each RB and each UE achievable raes r m n are calculaed. ow he uiliy of RB n for UE m can be expressed as: ( i ( i ( i u m n = rm n dm. (4 ( ( i i The uiliy marix is given by Uiner = u mn. Each enry of ( i Uiner is associaed wih corresponding inerferer(s o be resriced in addiion o he achievable rae and demand when RB n is used by UE m. Applying Hungarian algorihm o uiliy marix ( i Hungarian algorihm is applied o Uiner in an ieraive manner similar o Secion III(A. In each secor seps given below are followed o prepare iner-eb RBs resricions. Apply Hungarian algorihm o ( i U iner. If any seleced uiliy enry has a corresponding inerferer resricion he resricion lis will be updaed wih he marked inerferer for he RB. Updae he columns of he uiliy marix corresponding o assigned RBs wih zeros. ow apply he Hungarian algorihm o he updaed uiliy marix. Repea above seps unil all enries of he uiliy marix are zero. The number of required ieraions is bounded by M. Iner-eB communicaion using X inerface For a paricular RB Fig. shows an example scenario of iner-eb iner-cell resricions. In his figure he green (solid line and red (dashed line arrows indicae ha iner-eb iner-cell inerference received a he arrow-originaing-secor from he arrowhead-secor is accepable and unaccepable (o be resriced respecively. For example for a RB of ineres secor B can olerae inerference from secor A bu he opposie is no rue as here is a red (dashed arrow from secor A oward B. In his case eiher secor A or B has o be resriced for his RB. In his case eb corresponding o secor A communicaes wih he eb corresponding o secor B using X inerface abou resricing he RB. I is expeced ha a pair of secors will have resricions for he same RB in some cases (i.e. red arrows o each oher. In such cases he secor ha achieves higher uiliy survives. The negoiaion of resoluion of his ype of conflicing resricions is also carried over X inerface. B A C Fig. : Graphical represenaion of iner-eb RB resricion IV. SIMULATIO SYSTEM AD ARAMETERS Considered simulaion and sysem parameers are aken mosly from [] and [7] as summarized in Table I. Time-frequency correlaed 6-aps exended spaial channel model (SCME wih power delay profile as defined in [8] is considered. Independen lognormal shadow fading wih a sandard deviaion of 8 db has been assumed.

5 0 MHz sysem bandwidh consiues 00 RBs. A RB consiss of subcarriers (each of 5 khz in frequency and 7 OFDM symbols in he ime dimension. We assume ha 4 and OFDM symbols per RB are used for downlink reference and conrol signals respecively giving 77 OFDM symbols per RB for daa raffic. Therefore a RB can carry 77 informaion bis wih QSK rae / modulaion and coding scheme (MCS. TABLE I SIMULATIO ARAMETERS arameer Assumpion Cellular layou Hexagonal grid 9 cell sies secors per sie Iner-sie disance 500 m Carrier frequency / Bandwidh.0 GHz / 0 MHz (00 RBs Disance-dependen pah loss L = log0( R Lognormal shadowing Shadowing sandard deviaion UE speeds of ineres eneraion loss Anenna paern (horizonal Anenna configuraion BS anenna gain UE anenna gain UE noise figure AMC modes Channel model Toal secor TX power Downlink iner-cell inerference modelling Minimum disance beween UE and BS Traffic model Scheduler Independen among links 8 db 0 km/hr 0 db θ A ( θ = min Am θ db θ db = 70 degrees A m = 0 db Single-Inpu-Single-Oupu 4 dbi 0 dbi 7 db QSK 6-QAM and 64-QAM wih varying raes 6-Tap SCME 46 dbm Explici modelling (all links are simulaed 5 m Full buffer Hungarian algorihm based on uiliy marix Adapive modulaion and coding (AMC is used wih various MCS modes wih quadraure phase shif keying (QSK 6- and 64-quadraure ampliude modulaion (QAM and coding raes ranging from /8 o 4/5. Link adapaion is performed using aenuaed and runcaed form of Shannon bound mached o link level performance curves wih above menioned modulaion level and coding raes as follows [9]. 0 ; γ < γmin (5 η = αs ( γ ; γmin < γ < γmax η ; γ > γ max max where η is he specral efficiency in bps/hz γ is he signal-oinerference-plus-noise raio (SIR seen on RB and α (0.75 used in simulaion is he aenuaion facor applied o he Shannon bound given by S ( γ = log ( + γ which achieves η max (4.8 bps/hz a γ max (9. db or beyond and 0 a γ min (- 6.5 db or lower. Auomaic repea reques (ARQ has no been considered in simulaions. For fair comparison oal ransmi power per secor is kep consan in all schemes which is. Accordingly power allocaed o RBs for differen schemes are shown in Table II. In he able (46 dbm (00 res and elig are he oal power per secor sysem bandwidh in erms of number of RBs and he number of resriced (from inra and iner-eb and unresriced RBs respecively. As shown in he able resriced RBs are unused in proposed scheme and used wih 0 db lower power in proposed scheme. TABLE II OWER ALLOCATIO FOR DIFFERET SCHEMES Scheme Allocaed power on RBs Reuse R rb = Reuse R rb = FR eff reuse FR. FR.. rb ou =. 5 rb in =. SFR.5 SFR.5.5 rb ou =.5 rb in = 0.75 SFR.0 SFR.0.0 rb ou =.0 rb in = 0.5 SFR.5 SFR.5.5 rb ou =.5 rb in = 0.5 SFR.75 SFR rb ou =.75 rb in = 0. roposed RO = RO 0 scheme rb elig ( res rb res = roposed RO RO scheme rb elig = 0 ( 0elig + res rb res = ( 0elig + res V. SIMULATIO RESULTS AD DISCUSSIOS The performance of he proposed inerference avoidance scheme is compared o ha of he reference schemes in erms of cell-edge and average secor hroughpu. Cell-edge hroughpu is defined as he 5 h percenile poin of CDF of UE hroughpu. The reference schemes simulaed are reuse reuse SFR (wih effecive reuse of and.75 and FR (effecive reuse of.. Saisics are colleced from a oal of 00 drops. In each drop UEs are uniformly disribued according o a densiy of UEs/secor. Simulaion ime span is 5 ms (50 OFDM symbols in each drop. Resource resricion refreshmen inerval considered is 6 OFDM symbols (wihin he channel coherence ime for he UE speed of 0 km/hr. UEs are placed in he cenral secors among he oal 57 secors of 9 sies. For he proposed scheme allocaion algorihms are run in hese secors. Saisics are colleced from he cenral eb ( secors only. Inerference is calculaed using cenral cell approach. Fig. 4 shows CDF of UE hroughpu for he proposed as well as reference schemes. The lower ail of CDF is zoomed in Fig. 5 in order o view he cell-edge hroughpu clearly. Fig. 6 compares cell-edge and average secor hroughpu among all simulaed schemes wih reference o reuse scheme. I is observed ha while reuse scheme achieves 68.4% improvemen in cell-edge hroughpu compared o reuse scheme i suffers from secor hroughpu degradaion by 49.%. The FR scheme improves cell-edge hroughpu by 49.7% however wih 6% reducion in secor hroughpu.

6 SFR schemes wih effecive reuse of.0.5 and.75 show cell-edge hroughpu improvemen by.4% 69.4% and 4.7% wih degradaion of secor hroughpu by 4.% 8.6% and 6.% respecively. SFR scheme wih effecive reuse of.5 neiher improves cell-edge nor secor hroughpu compared o reuse scheme. Comparing wih reuse scheme proposed scheme improves cell-edge hroughpu by 66.% wih an improvemen in secor hroughpu by.9%. roposed scheme shows inferior cell-edge and superior secor hroughpu compared o hose in proposed scheme as resriced RBs are used wih reduced power which favors cell-cenre UEs while harms hose a he cell-edge. However compared o reuse scheme proposed scheme obains 7.% improvemen in cell-edge hroughpu while improving he secor hroughpu by 7.%. The performance gain in he proposed scheme is achieved wih he cos of increased overhead required for UE-BS feedback as well as eb-eb communicaion over X inerface. robabiliy (Throughpu <= Abscissa robabiliy (Throughpu <= Abscissa roposed ( res =0 0.5 roposed ( res =0.X elig 0.4 Reuse Reuse 0. FR (eff reuse. 0. SFR (eff reuse.5 SFR (eff reuse.0 0. SFR (eff reuse.5 SFR (eff reuse Average UE Throughpu (Mbps Fig. 4: CDF of UE hroughpu roposed ( res = roposed ( res =0.X elig 0.04 Reuse Reuse 0.0 FR (eff reuse. 0.0 SFR (eff reuse.5 SFR (eff reuse SFR (eff reuse.5 SFR (eff reuse Average UE Throughpu (Mbps Fig. 5: CDF of UE hroughpu (zoomed o show cell-edge hroughpu VI. COCLUSIOS We have presened and evaluaed an inerference avoidance scheme for downlink LTE sysem ha uses dynamic iner-cell coordinaion suppored by X inerface. Two variaions of he proposed scheme have been compared wih he reference reuse scheme as well as he saic pariion based inerference avoidance schemes in he lieraure. I has been observed ha alhough saic schemes achieve improved cell-edge hroughpu hey suffer seriously in erms of secor hroughpu. On he oher hand he proposed schemes no only achieve higher cell-edge hroughpu bu also show improvemen in average secor hroughpu compared o hose in any saic scheme. Average Secor Throughpu (Mbps Reuse CE: 99 KbpsS:.8 Mbps SFR eff.reuse.0 CE:+.4%S:-4.% roposed ( res =0.X elig SFR eff.reuse.5 CE:-.%S:-.4% SFR eff.reuse.5 CE:+69.4%S:-8.6% SFR eff.reuse.75 CE:+4.7%S:-6.% roposed CE:+7.%S:+7.% ( res =0 CE:+66.%S:+.9% FR eff reuse. CE:+49.7%S:-6% Reuse CE:+68.4%S:-49.% Cell Edge Throughpu (Kbps Fig. 6: Cell-edge vs. average secor hroughpu ACKOWLEDGMET This research is financially suppored in par by he Communicaions Research Cenre of Canada (CRC. REFERECES [] G TR 5.84 V7..0 hysical Layer Aspecs for Evolved Universal Terresrial Radio Access (UTRA (Release 7 Sepember 006. [] WiMAX Technical Repor Mobile WiMAX - ar I: A Technical Overview and erformance Evaluaion Augus 006. [] A. Alexiou WIER: Designing a new radio inerface for nexgeneraion sysems Bell Labs Technical Journal vol. pp. 9-5 Augus 007. [4] J. Chuang and. Sollenberger Beyond G: wideband wireless daa access based on OFDM and dynamic packe assignmen IEEE Commun. Mag. vol. 8 pp July 000. [5] V.H. MacDonald The cellular concep Bell Sysem Technical Journal vol. 58 no. pp. 5-4 January 979. [6] K. Begain G.I. Rozsa A. fening and M. Telek erformance analysis of GSM neworks wih inelligen underlay-overlay in roc. Sevenh Inernaional Symposium on Compuers and Communicaions (ISCC 00 July 00 pp [7] G R Inerference Miigaion - Consideraions and Resuls on Reuse Siemens Sepember 005. [8] G R-0609 OFDMA Downlink Iner-Cell Inerference Miigaion okia February 006. [9] WIER II Deliverable D4.7. Inerference Avoidance Conceps June 007 available a hp:// [0] M. Rahman and H. Yanikomeroglu Inerference avoidance hrough dynamic downlink OFDMA subchannel allocaion using inercell coordinaion in roc. IEEE VTC Spring 008 pp May 008. [] G TS 6.00 V8.4.0 E-UTRA and E-UTRA Overall descripion; Sage (Release 8 March 008. [] G R Sof Reuse Scheme for UTRA LTE Huawei May 005. [] G R Furher Analysis of Sof Reuse Scheme Huawei Sepember 005. [4] M. Sernad T. Oosson A. Ahlen and A. Svensson Aaining boh coverage and high specral efficiency wih adapive OFDM downlinks in roc. IEEE VTC 00 Fall Oc. 00 pp [5] G R-0605 Inerference Miigaion by arial Reuse Siemens January 006. [6] H.W. Khun The Hungarian mehod for he assignmen problem aval Research Logisic Quarerly vol. pp [7] G TS 6. V8..0 EUTRA hysical Channels and Modulaion (Release 8 March 008. [8] D.S. Baum J. Salo M. Milojevic. Kyösi and J. Hansen An inerim channel model for beyond-g sysems in roc. IEEE VTC 005 Spring pp. 6 May 005. [9] G TR 6.94 V..0 E-UTRA Radio (RF Sysem Scenarios (Release 8 June 007.