Iterferece Maagemet i LTE Femtocell Systems Usig a Adaptive Frequecy Reuse Scheme Christos Bouras 1,2, Georgios Kavourgias 2, Vasileios Kokkios 1,2, Adreas Papazois 1,2 1 Computer Techology Istitute & Press Diophatus ad 2 Computer Egieerig ad Iformatics Departmet, Uiversity of Patras, Greece bouras@cti.gr, kabourgias@ceid.upatras.gr, kokkios@cti.gr, papazois@ceid.upatras.gr Abstract Log Term Evolutio (LTE) has developed a ew techology i order to ehace idoor coverage. This ew techology is called femtocells ad is achieved with the use of access poits istalled by home users. However, iterferece problem betwee the femtocell ad the macrocell decreases the system's capacity ad as a result users' throughput. I this paper we study the iterferece mitigatio techiques i femtocell/macrocell etworks ad we propose a frequecy reuse mechaism that leads to icreased overall system performace. I particular, the mechaism aims to maximize throughput via a variety of combiatios betwee ier cell radius ad frequecy allocatio to the macrocell. Additioally, a positio mided frequecy allocatio to the femtocells targets to further optimize the total throughput of the cell. Keywords- frequecy reuse; iterferece mitigatio; log term evolutio; femtocell; I. INTRODUCTION I moder societies the demad for higher data rates is quite itese ad users expect to achieve comparable data rates i both wired ad mobile etworks. This has triggered the desig ad developmet of ew data-mided cellular stadards of which the Third Geeratio Partership Project s (3GPP s) Log Term Evolutio (LTE) appears to be the most promisig cadidate. The LTE is techologically based o Orthogoal Frequecy Divisio Multiple Access (OFDMA) to achieve higher data rates ad ehaced spectral efficiecy. Kowig that radio spectrum is limited ad becomig isufficiet owadays, it is ievitable to use precious spectrum resource wisely. Oe of the efficiet ways of spectrum resource utilizatio is to apply Frequecy Reuse, however if same subcarriers are used by differet users amog adjacet cells, Co- Chael Iterferece (CCI) problem occurs especially for cell edge users. Appropriate iter-cell iterferece coordiatio techique should be required to ehace the system capacity [1]. The LTE has developed a femtocell for idoor coverage extesio. Femtocells, also called home base statios (BSs) or Home Node-Bs (HNBs), are short-rage low-cost low-power BSs istalled by the cosumer that work i the licesed frequecy bads ad they are coected to broadbad Iteret backhaul [2], [3]. A femtocell allows service providers to exted service coverage idoors, especially where access would otherwise be limited or uavailable. Therefore, the subscriber is satisfied with higher data rates ad reliability ad simultaeously the operator reduces the amout of traffic o expesive macrocell etwork. Femtocells ad macrocells could suffer severe CCI from each other if cell plaig or spectrum maagemet is ot appropriately cosidered. Operatig femtocells o a dedicated frequecy bad is a simple solutio, but the frequecy resources are ot utilized effectively. Co-chael model is suitable for practical deploymet ad overcomes the problem of misuse, but the CCI problem should be solved. I [4] a extesive aalysis of the differet iterferece scearios is provided cosiderig closed access for OFDMA techologies while i [5], a proposal for iterferece avoidace is based o Dyamic Frequecy Plaig (DFP) i a survey coducted i a WiMAX etwork itegratig femtocells. Frequecy is dyamically assiged, cosiderig user badwidth demad, aimig to decrease iterfereces ad therefore improve the etwork capacity. However the cetralized approach of this solutio is agaist the femtocell techology where the HNBs are supposed to be powered up ad maaged by a persoal ower. The authors of [6] study various frequecy reuse schemes i the OFDMA based etwork, such as the LTE, to overcome the CCI problems. Whole frequecy bad is divided ito several sub-bads ad each sub-bad is allocated i the cells i differet ways so as to mitigate iterferece. Fractioal Frequecy Reuse (FFR) ad Two Level Power Cotrol (TLPC) schemes with appropriate settigs of ier regio radius ad power ratio provide the best performace whe a scheduler fair i throughput is assumed. As a result, itra-cell iterferece is removed, ad iter-cell iterferece is substatially reduced. Fially, the authors of work [7] propose a mechaism for optimal FFR scheme selectio based o the mea user throughput or user satisfactio. This survey is coducted i a cellular etwork that does ot support the existece of femtocells, however the use of user satisfactio as a metric of evaluatio appears to be very iterestig. I this paper, a frequecy reuse selectio mechaism is proposed, aimig to reduce cross-tier CCI betwee macrocells ad femtocells. Each macrocell is divided ito ier ad outer regio ad frequecy is allocated with reuse factor 1 ad reuse factor 3 respectively. The frequecy bad that will be used i each regio is calculated for every possible combiatio of ier cell radius ad subcarrier allocatio. For femtocells, frequecy allocatio is doe accordig to the femtocells s 978-1-4577-0580-9/12/$26.00 2012 IEEE
positio i the etwork. Each femtocell chooses a frequecy bad which is differet from the sub-bad already used for the macrocell located i the same area ad this frequecy allocatio is expected to further improve etwork s performace. The mechaism computes the total throughput of the cell ad user satisfactio for each pair of radius a frequecy ad fially selects the optimal frequecy allocatio scheme. Ier cell radius ad the umber of subcarriers to be allocated i each regio are selected so as to maximize the value of user satisfactio. A method to evaluate the proposed scheme is via the estimatio of the adjacet cell iterferece. For the calculatio of the desired results a aalytical model aalysis is used, that takes ito accout the path loss ad propagatio i order to estimate the SINR ad therefore the adjacet cell iterferece due to femtocells. Fially, those results are compared with other frequecy reuse schemes. The remaiig of this paper is structured as follows. The proposed etwork model is described i Sectio II. Sectio III aalyzes the simulatio mechaism. Sectio IV presets the equatios which will be used for the implemetatio of the proposed scheme ad Sectio V gives performace evaluatio ad compariso of the proposed method with other frequecy reuse schemes. Fially, Sectio VI icludes the coclusios of this study ad ideas for future work. II. PROPOSED NETWORK MODEL The primary goal of this survey is to ehace throughput for both femto ad macro users. Oe way to achieve this is to mitigate iterferece with frequecy reuse schemes. To this directio FFR is applied to each macrocell of the topology, as show i Figure 1, i order to reduce iterferece betwee adjacet cells ad ehace throughput. Figure 2. Frequecy bad divisio. Each hexagoal cell of radius R is divided ito ier (gray color) ad outer regio (colors red, gree ad blue). Oe part of the available frequecy bad is dedicated to the cell cetre users with reuse factor of 1 (frequecy bad W1), while the rest of the spectrum is equally divided i three sub-bads ad assiged to the cell edge users with reuse factor of 3 (frequecy bads W2, W3 ad W4). The frequecy bad divisio is depicted i Figure 2. If a femtocell is located i the outer regio of a macrocell, the the sub-bad used for the ier regio ca be reused for the femto users. O the cotrary, a femtocell dropped i the ier regio caot reuse the sub-bad which was assiged to the cell edge users of the macrocell. The reaso is the trasmit power of the BS i each case. Ier cell users are closer to the BS, which meas that lower trasmit power is required. O the other had, the BS should trasmit i maximum power i order to satisfy cell edge users. For example, accordig to Figure 1, if a femtocell is operatig i the outer regio of the cetral cell (colored with gray ad gree) it ca use ot oly sub-bads W3 ad W2 (red ad blue that are assiged to the outer regios of the adjacet cells) but also sub-bad W1 (gray) eve if it is already used i this macrocell. Uder these coditios, the iterferece from the macrocell to the femtocell ad vice versa will be reduced. Figure 1. Frequecy bad allocatio usig FFR with reuse factor 3. III. MECHANISM DESCRIPTION The mechaism receives as iput the macrocell eviromet dimesioig, the umber of femtocells ad their positios as well as other characteristics of the etwork such as the macrocell ad femtocell BS trasmissio power. Cosequetly it calculates the received power from servig as well as from the iterferig cells ad based o these values ad by takig ito accout the white Gaussia oise, the mechaism is able to make a estimatio of the SINR ad throughput at ay give positio of the examied LTE etwork. Fially, user satisfactio for each cell is calculated via the values of throughput of each user. The mechaism follows the approach preseted below: Calculate ier cell radius Give the macrocells characteristics, differet radiuses from 0 to R are examied i order to fid the best dimesios of the ier cell that optimizes user satisfactio for the cell users. Fid the optimum frequecy bad divisio The aim of this step is to calculate user satisfactio for every possible combiatio of spectrum divisio. The available spectrum will be allocated to the users accordig
to the combiatio that maximizes user satisfactio. There are two disjoit sets of subcarriers. Set I that cotais the subcarriers of the ier regio ad set O that cotais the subcarriers of the outer regio. There are totally 26 differet cases ad every time the frequecy bad that is allocated to the outer regio is equally divided betwee W1, W2 ad W3. Iitially, set I is a empty set ad all the subcarriers are cotaied i set O, which meas that all the subcarriers are allocated to the outer regio ad each oe of W1, W2 ad W3 equals to 25/3. I each oe of the followig cases oe subcarrier is removed from set O ad is added to set I. Fially, i the 26 th case, set I cosists of 25 subcarriers ad set O is a empty set. Allocate frequecy bad to the femtocells I this step frequecy is allocated to the femtocells with the process preseted i Sectio II. The The pseudo-code that follows presets briefly the geeral idea of the mechaism. create_topolgy() %defies cells ad drops femtocells ad users for r = 0: R %ier cell radius for f = 0: 25 %frequecy bad divisio allocate frequecy bad for macrocells if r > = distace_femto %femtocell belogs to ier cell allocate frequecy bad for femtocell else %femtocell belogs to outer cell allocate frequecy bad for femtocell ed for u = 1 : U %all users sir = calculate_sir(u) capacity = calculate_capacity(u) throughput = calculate_ throughput(u) ed US = calculate_user_satisfactio(r, f) ed ed defie_ffr(max_ user_satisfactio) IV. ANALYSIS OF MECHANISM This sectio presets the simulatio model that will be used i our aalysis. It estimates the cross-tier iterferece ad the throughput i every poit of the LTE etwork itegratig femtocells ad macrocells. Based o the proposed etwork scheme the model takes ito accout the path loss ad propagatio models i order to estimate the SINR ad therefore the adjacet cell iterferece of the itegrated LTE etwork. I order to evaluate the performace of a lik we ca use either block error ratio (BLER) or throughput. I this paper we will cotiue our aalysis with the help of throughput performace. To formulate the equatio for the system throughput, first of all we will eed a equatio for the SINR calculatio. The received SINR for a macro user i o a subcarrier ca be expressed as [8] : SINR i, PM, Gi, M, = N Δ f + P G + P G 0 M ', i, M ', F, i, F, Μ ' F (1) where, P M, ad P M, is trasmit power of servig macrocell M ad eighborig macrocell Μ o subcarrier, respectively. G i,m, is chael gai betwee macro user i ad servig macrocell M o subcarrier ad G i,μ, correspods to chael gai from eighborig macrocell Μ. Trasmit power of eighborig femtocell F o subcarrier is deoted by P F, ad G i,f, represets chael gai betwee macro user i ad eighborig femtocell F o subcarrier. Fially, N 0 is white oise power spectral desity ad Δf is subcarrier spacig. I case of a femto user f, the received SINR o a subcarrier ca be similarly give by [8] : SINR f, PF, Gf, F, = N Δ f + P G + P G 0 M, f, M, F', f, F', Μ F ' where, F is the set of iterferig femtocells. The chael gai G is domiatly affected by path loss, which is differet for macro users ad femto users. The path loss betwee a macro BS ad a User Equipmet (UE) is modeled as [9] : 15.3 37.6log 10( ) ow (2) PL = + d + L (3) For the case of a outdoor user, L ow is set to 0. The path loss betwee a femto BS ad a UE i the same house is give by the followig equatio [9] : PL = 38.46 + 20 log 10 ( d) + 0.7d2 D, idoor + (( + 2)/( + 1) 0.46) 18.3 + ql ad i the case that the UE is ot i the same house, path loss is give by [9] : PL = max(15.3 + 37.6log ( d),38.46 + 20log ( d)) + 0.7d 18.3 ql L L 10 10 (( 2)/ ( 1) 0.46) 2 D, idoor + + + + iw + ow1 + ow2 For the case of a outdoor user, L ow2 is set to 0. I the equatios of path loss, d is the distace betwee the trasmitter (Tx) ad the receiver (Rx) i meters ad L ow ad L iw correspods to peetratio loss of a outdoor ad idoor wall respectively which are set to L ow = 20 db ad L iw = 5 db. Especially for path loss betwee a femto BS ad a user, is the umber of peetrated floors, q is the umber of walls separatig apartmets betwee the femto BS ad the UE ad the term 0,7d 2D,idoor takes accout of peetratio loss due to walls iside a apartmet with d 2D,idoor represetig the distace iside the house. So, the chael gai ca be expressed as follows [8] : G iw (4) (5) = 10 PL/10 (6)
Additioally, for the throughput calculatio, we use the capacity of a user i o a specific subcarrier, which ca be estimated via the SINR from the followig equatio [8] : C = W log (1 + asinr ) (7) i, 2 i, where W is the available badwidth for subcarrier divided by the umber of users that share the specific subcarrier ad α is a costat for a target bit error rate (BER) defied by α = -1.5 / l(5ber). Here we set BER to 10-6. So the expressio of the total throughput of the servig macrocell M is [8] : T = p C (8) M i, i, i where p i, is a assigmet idex variable with p i, = 1 implyig that subcarrier is assiged to user i ad p i, = 0 otherwise. I the specific paper the study is doe for LTE systems, which meas that each subcarrier ca be oly occupied by oe user i the same cell i each time slot. Supposig that N i is the umber of users i a macrocell, this implies that: N i pi, = 1 (9) i= 1 The expressios for the capacity (equatio (7)) ad the total throughput (equatios (8) ad (9)) are similar i the case of a femto user. I [7] the term User Satisfactio (US) is itroduced. The authors have proved that US guaratees slight differeces i users throughput values. US physically presets how close the user s throughput is to the maximum throughput of a user dropped i the same cell. Whe US approaches 1, all users i the cell experiece similar throughput. O the cotrary, whe there is big differece i the throughput achieved by the users i the cell, US value approaches to 0. US ca be expressed as [7] : Ti i= 1 US = max_ user _ throughput U U (10) where U is the total umber of users i the specific cell, T i is the throughput of each user (either macro or femto user) ad max_user_throughput is the maximum throughput value achieved by a user i this cell. The mathematical aalysis preseted previously will be the basis for the implemetatio of the simulatio mechaism that will help us to evaluate the solutio proposed for the scearios which will be examied i this paper. V. PERFORMANCE EVALUATION A. Scearios Presetatio The simulatio parameters that will be used for the experimet are listed i Table I. The trasmit power of the femto BS is 20 mw ad the macro BS trasmits for the edge regio with 22 W ad for the cetre zoe with 15 W. Furthermore we cosider a system with 10 MHz of badwidth divided ito 25 subcarriers of 375 khz of badwidth ad 15 khz of subcarrier spacig. A fixed umber of 100 femtocells radomly located i the etwork area is used oly for the calculatio of the optimum parameters for the proposed scheme. I the experimets for the compariso with other frequecy reuse schemes varyig umber of femtocells from 0 to 5 femtocells per macrocell were used. Parameters TABLE I. SIMULATION PARAMETERS Macro Values Femto Number of cells 19 100 Radius 250 m 20 m BS trasmit power 15, 22 W 20 mw Badwidth 10 MHz Subcarriers 25 Subcarriers badwidth Subcarrier spacig Carrier frequecy Chael model Power Noise Desity 375 khz 15 khz 2 GHz 3GPP Typical Urba -174 dbm / Hz The followig scearios will be used as a meas to evaluate the proposed scheme preseted previously: I the same etwork topology Iteger Frequecy Reuse (IFR) with reuse factor 1 is applied which meas that all the available badwidth is used i each macrocell. Furthermore, each femtocell uses the same frequecy bad. The available spectrum is divided ito 3 equal sub-bads ad IFR with reuse factor 3 is applied to the macrocells of the etwork. As a result, each macrocell uses a subbad which is differet from all the eighborig macrocells. Frequecy allocatio to the femtocells is doe accordig to their positio, which meas that each HNB ca oly trasmit i the two sub-bads that are ot used by the macrocell where the femtocell belogs.
B. Numerical Evaluatio ad Compariso This sectio presets the operatio of the mechaism ad the applicatio of the proposed FFR scheme with the values that maximize user satisfactio. Figure 3 depicts the etwork topology preseted previously ad particularly the cetral cell (colored with gree outer regio). Users ad femtocells are distributed uiformly i the topology ad optimal ier cell radius ad subcarrier allocatio is calculated via the experimet. Figure 5. User satisfactio vs umber of femtocells. Figure 3. Network Topology with the applicatio of FFR for optimal radius. I the cetral cell there are 30 users ad 5 femtocells radomly located. With the proposed mechaism the ier cell radius that maximizes user satisfactio was calculated ad equals to 175m which meas that all the femtocells ad 24 users belog to the ier regio while the remaiig 6 users were dropped i the outer regio of the cell. Figure 4 depicts US while the umber of subcarriers allocated to the ier cell varies from 0 to 25 for the optimum ier cell radius preseted i Figure 3. Figure 4. User Satisfactio vs subcarriers allocated to ier regio for optimal ier radius. The coclusio of Figure 3 ad Figure 4 is that the maximum value of user satisfactio which is 0.3859 was achieved for ier cell radius 175m while allocatig 22 subcarriers for the ier regio ad the remaiig 3 for cell edge users which correspods to 8.25 MHz ad 1.125 MHz respectively. As the umber of subcarriers allocated to the ier cell icreases, user satisfactio icreases too ad maximizes for 22. This is ratioal ad ca be explaied by the fact that almost all the users belog to the ier regio, which meas that the majority of the subcarriers should be allocated to the cell cetre. Figure 5 shows how US chages as umber of femtocells per macrocell icreases from 0 to 5. I order to evaluate the proposed mechaism we preset a compariso with IFR with reuse factor 1 (IFR1) ad IFR with reuse factor 3 (IFR3). Accordig to Figure 5, the cases of IFR1 ad IFR3 achieve similar values of user satisfactio ad about three times smaller tha the proposed scheme whe there are more tha 1 femtocells. This is very importat because it idicates that there is fairess amog all the users ad the resources are distributed i a way that everyoe will have throughput values with little variatio from the maximum oe. The sudde drop of user satisfactio i the case of IFR1 ad IFR3 whe the first femtocell appears idicates the absece of a algorithm that will distribute appropriately the resources amog the users. As soo as there are femto users, big part of the available badwidth is allocated to them leadig to disproportioate values of throughput ad reducig user satisfactio. Figure 6 illustrates aother compariso betwee the proposed scheme ad the cases of IFR1 ad IFR3. The assumptio is agai that the umber of femtocells varies from 0 to 5, although this time throughput values are plotted. I each case is calculated the miimum (mi) ad maximum (max) value of throughput of a user as well as the average value of throughput for all the users beig served by the same cell. Through this experimet we aim to prove oce agai the fairess of the proposed mechaism. For that reaso Figure 6 focuses to a overall compariso of all the cases ad less attetio is paid to the values of throughput i each case separately.
Figure 6. Throughput vs umber of femtocells. The most valuable observatio from Figure 6 is that the average values of throughput of the proposed mechaism are close eough to the miimum ad maximum value of this techique. This slight differece idicates that all the users experiece values of throughput with little variace. This is exactly what was expected, because usig US as a metric for the applicatio of the FFR, all the users are served with comparable quality. O the cotrary, the cases of IFR1 ad IFR3 have average values of throughput close eough to the worst case, however the values of maximum throughput are extremely high givig a sese of ijustice i the way that the users are treated. I the experimetal process the characteristics of the FFR scheme to be applied were selected usig US. The value of ier cell radius ad the distributio of subcarriers were those that maximized US. The reaso is that US allocates frequecy i a way that eables all the users to achieve comparable values of throughput. Figure 7 presets the compariso of user throughput for two differet metrics. I each case throughput for every user is calculated based o maximum US or maximum total cell throughput. It should be highlighted that users 3, 4, 7, 10, 26 ad 29 are femto users while the rest of them are macro users. It is worth otig that user throughput is lower whe US is the metric, however high values of throughput are sacrificed i order to serve the users i a more fair way. O the cotrary, applyig FFR i a way that maximizes total cell throughput leads to a ufair distributio of the available resources. The biggest part of the available badwidth is allocated to the femto users ad as a result they achieve big throughput values - almost te times bigger tha the macro users - which maximizes total cell throughput. Additioally, Figure 7 shows that whe total cell throughput is used as a metric, there are users that are ot served at all. These are the users that belog to the cell cetre. As soo as ier cell icreases ad icludes all the femtocells, all the subcarriers are allocated to the outer cell. I this occasio there is the maximum profit for a femto user. O the cotrary, usig user satisfactio as a metric reassures ot oly to serve the ier cell users, but also with similar values of throughput with the cell edge ad the femto users. VI. CONCLUSIONS AND FUTURE WORK The research i this paper focused i the mitigatio of the iterferece i LTE etworks that itegrate femtocell overlay. Our techique is based o the idea of frequecy reuse i order to reduce SINR ad achieve higher values of throughput. The proposed mechaism selects the optimum values for the applicatio of the FFR based o the maximizatio of user satisfactio. Our mechaism seems to be far more effective whe it comes to fairess eve though it seems to be worse cocerig total cell throughput. For further evaluatio we preset the compariso whe maximizatio of total cell throughput is used as a metric ad scearios where differet frequecy reuse schemes are applied. A iterestig suggestio for future work could be the import of more variables i the proposed model, aimig to provide more realistic scearios. For example, we could take care of the case that the user is served by a femtocell that belogs to aother apartmet, addig more factors i the path loss equatios. Additioally we could geeralize the proposed mechaism i order to support users mobility. However this implies the reductio of complexity i the computatios i order to become feasible the extractio of results from such a complicated model. Figure 7. Per user throughput based o User Satisfactio ad Total Cell Throughput. REFERENCES [1] T. O. Boudreau et al., Iterferece Coordiatio ad Cacellatio for 4G Networks, IEEE Commu. Mag., vol. 47, o. 4, pp. 74-81, Apr. 2009. [2] V. Chadrasekhar ad J. G. Adrews, Femtocell Networks: A Survey, IEEE Commu. Mag., vol. 46, o. 9, pp. 59-67, Sept. 2008. [3] 3GPP TR 36.921 V10.0.0, Evolved Uiversal Terrestrial Radio Access (E-UTRA); FDD Home enode B (HeNB) Radio Frequecy (RF) requiremets aalysis (Release 10), 2011. [4] Iterferece Maagemet I OFDMA Femtocells, http://femtoforum.org. [5] D. Lopez-Perez et al., Iterferece Avoidace ad Dyamic Frequecy Plaig for WiMAX Femtocells Networks, IEEE Iteratioal Coferece o Commuicatio Systems (ICCS), Nov. 2008. [6] P. Godlewski, M. Maqbool, M. Coupechoux, J-M Kélif, Aalytical Evaluatio of Various Frequecy Reuse Schemes i Cellular OFDMA Networks, ValueTools, 2008.
[7] D. Billios, C. Bouras, V. Kokkios, A. Papazois, G. Tseliou, Optimizatio of Fractioal Frequecy Reuse i Log Term Evolutio Networks, WCNC, 2012 (to appear). [8] P. Lee, T. Lee, J. Jeog, ad J. Shi, Iterferece Maagemet i Lte Femtocell Systems Usig Fractioal Frequecy Reuse, i Advaced Commuicatio Techology (ICACT), 2010 The 12th Iteratioal Coferece o, vol. 2, feb. 2010, pp. 1047 1051. [9] 3GPP TR 36.814 V9.0.0, Evolved Uiversal Terrestrial Radio Access (E-UTRA); Further advacemets for E-UTRA physical layer aspects (Release 9), 3rd Geeratio Partership Project, Tech. Rep., 2010.