2012 IEEE Wireless Commuicatios ad Networkig Coferece: Mobile ad Wireless Networks Optimizatio of Fractioal Frequecy Reuse i Log Term Evolutio Networks Dimitrios Bilios 1,2, Christos Bouras 1,2, Vasileios Kokkios 1,2, Adreas Papazois 1,2, Georgia Tseliou 1,2 1 Computer Techology Istitute & Press Diophatus, Patras, Greece 2 Computer Egieerig & Iformatics Dept., Uiv. of Patras, Greece bilios@ceid.upatras.gr, bouras@cti.gr, kokkios@cti.gr, papazois@ceid.upatras.gr, tseliou@cti.gr Abstract I cellular systems, Fractioal Frequecy Reuse (FFR) partitios each cell ito two regios; ier regio ad outer regio ad allocates differet frequecy bad to each regio. Sice the users at the ier regio are less exposed to iter-cell iterferece, the frequecy resources i each ier regio ca be uiversally used. Based o this frequecy bad allocatio, FFR may reduce chael iterferece ad offer large system capacity. This paper proposes a mechaism that selects the optimal FFR scheme based o the user throughput ad user satisfactio. I detail, the mechaism selects the optimal size of the ier ad outer regio for each cell as well as the optimal frequecy allocatio betwee these regios that either maximizes the mea user throughput or the user satisfactio. The mechaism is evaluated through several simulatio scearios. Keywords-log term evolutio; fractioal frequecy reuse; orthogoal frequecy divisio multiple access; I. INTRODUCTION Orthogoal Frequecy Divisio Multiple Access (OFDMA) has become a attractive techology i recet years ad it is part of various system stadards for mobile commuicatios. This happes because each termial occupies a subset of subcarriers (called OFDMA traffic chael) ad each traffic chael is assiged exclusively to oe user at ay time [1]. Several techiques with differet degrees of complexity ca be cosidered for out-of-cell iterferece mitigatio i OFDMA systems. OFDMA provides a degree of freedom by allowig dyamic assigmet of chaels/subcarriers to differet users at differet time istaces, to take advatage of the chael respose variatios amog differet users o differet chaels [1]. Sub-chaelizatio implies that a sigificat fractio of the power is used o oly a portio of the badwidth used to serve the weak user eve though uiversal reuse. Nevertheless, eighborig sectors should assig orthogoal subcarriers to cell edge users ad it is importat to cosider iterferece whe assigig subcarriers to users. Oe of the key characteristics of a cellular etwork is the ability to reuse frequecies i order to icrease both capacity ad coverage. Fractioal Frequecy Reuse (FFR) is discussed i the OFDMA-based etwork, such as the Log Term Evolutio (LTE), to overcome the Co-Chael Iterferece (CCI) problems [2]. I FFR the cell space is divided ito two regios: ier, which is close to the Base Statio (BS) ad outer, which is situated to the borders of the cell. The whole frequecy bad is divided ito several sub-bads, ad each subbad is differetly assiged to ier ad outer regio of the cell respectively. As a result of FFR, itra-cell iterferece is elimiated, ad iter-cell iterferece is substatially reduced [3]. At the same time the system throughput is ehaced. Various reuse factors ad iterferece mitigatio levels ca be achieved by adjustig either the badwidth proportio assiged to each regio or the trasmissio power of each bad. It should be oted that the majority of the related work regardig FFR i the cotext of OFDMA systems has maily bee discussed i cellular etwork stadardizatio for 3GPP ad 3GPP2 [4]. This idea was first proposed for GSM etworks, which is aalyzed by the authors of [5] ad has bee adopted i the WiMAX forum [2]. FFR is also proposed i the course of the LTE stadardizatio. I the research works [6] ad [7], the focus lies o practically implemetable algorithms. The authors use the full set of available resources i the reuse 1 areas ad oe third of the same resources i the reuse 3 areas. I [6] the reuse 1 areas are covered with a reduced power level, while i [7] the trasmit power of iterferig base statios is reduced. Fially i [8] a ovel FFR scheme for multi-cell OFDMA systems that improves chael capacity is proposed. Mai goal of this paper is to propose ad evaluate a iterferece maagemet FFR mechaism for OFDMA macrocell etworks. The mechaism calculates the optimal FFR scheme based o two parameters: user throughput ad user satisfactio. The proposed mechaism successively checks the ier cell radius ad the ier cell frequecy ad calculates the per-user Sigal to Iterferece plus Noise Ratio (SINR), capacity ad throughput. These values are the used i order to calculate the cell mea throughput ad the user satisfactio. Fially, the mechaism selects the optimal FFR scheme that either maximizes the cell mea throughput or the user satisfactio. The paper also presets several simulatio scearios i order to evaluate the proposed FFR mechaism. The rest of this paper is orgaized as follows. Basic theoretical backgroud regardig FFR is explaied i Sectio II. Sectio III describes the procedure used for the calculatio of SINR, throughput ad user satisfactio ad itroduces the proposed mechaism. The evaluatio of the mechaism ad the simulatio results are described i Sectio IV, while the coclusios ad ideas for future work are show i Sectio V. 978-1-4673-0437-5/12/$31.00 2012 IEEE 1875
II. FRACTIONAL FREQUENCY REUSE I FFR, i order to esure that the mutual iterferece betwee users ad base statios remais below a harmful level, adjacet cells use differet frequecies. I fact, a set of differet frequecies is used for each cluster of adjacet cells. Cluster patters ad the correspodig frequecies are reused i a regular patter over the etire service area. The closest distace betwee the ceters of two cells usig the same frequecy (i differet clusters) is determied by the choice of the cluster size ad the lay-out of the cell cluster. This distace is called the frequecy reuse distace. Oe of the mai objectives of LTE is to achieve high spectral efficiecy, meaig the use of the whole of the system s badwidth i all cells. This approach is called Frequecy Reuse 1, ad it is cosidered the simplest scheme; all sub-bads of the available badwidth are allocated to each cell. I Frequecy Reuse 3, the system badwidth is divided ito 3 equal sub-bads; each oe of these is allocated to cells i a maer that o other surroudig cell is usig the same subbad. Full frequecy reuse i each cell ca exempt the ecessity of advace frequecy plaig amog differet cells, ad the frequecy reuse patters ca be dyamically adapted o a frame-by-frame basis i each cell. I this work we study a sub-case of these approaches, which we aalyze below. Firstly we defie a LTE OFDMA multi-cellular etwork. Our mai objective is to apply FFR i order to improve the SINR ad throughput ad simultaeously reduce CCI. A idicative architecture ad frequecy bad allocatio are depicted i Figure 1. I detail, each macrocell of the architecture is divided ito two regios; ier ad outer regio. The total available badwidth of the system is split ito four ueve spectrums deoted by A (yellow), B (red), C (gree) ad D (blue). Spectrums B, C ad D have equal badwidth ad are allocated i outer regios with Frequecy Reuse 3. O the other had, spectrum A is allocated i all ier regios with Frequecy Reuse 1 (Figure 1). The frequecy resources i all ier regios are uiversally used, sice the ier regio users are less exposed to iter-cell iterferece. This frequecy allocatio ca elimiate the itra-cell iterferece ad greatly reduce the iter-cell iterferece. The users are assumed to be distributed uiformly i the cells of the topology. III. SYSTEM MODEL AND MECHANISM DESCRIPTION A. Throughput ad User Satisfactio Calculatio I this sectio we describe the theoretical approach to calculate the SINR, throughput ad user satisfactio factor. We assume that the overall etwork is composed of N adjacet cells. Each cell cotais a umber of users seekig to share a group of subcarriers. We distiguish the case where a user is foud i the ier regio of the cell or i the outer. I a typical OFDMA cellular etwork, for a user x who is served by a base statio b o subcarrier, the related SINR is give by the followig equatio [9]: SINRx, = σ Gbx, Pb, hbx,, k 2 + Gj, xpj, hj, x, j I (1), G b,x refers to the path loss associated with the chael betwee user x ad base statio b, P b, is the trasmit power of the base statio o subcarrier, h b,x, is the expoetially distributed chael fast fadig power ad σ 2 is the oise power of a Additive White Gaussia Noise (AWGN) chael. Symbols k ad j refer to the set of all the iterferig base statios (i.e. base statios that are usig the same sub-bad as user x). Their physical meaig is that j is the cell idex ad k the umber of co-chael cells. I our aalysis, we assume that equal trasmit power is applied, P b, =P for all base statios. The coefficiet h b,x, is replaced by its mea value (h b,x, = 1) i equatio (1). The iterferece that occurs comes from disjoit sets of dowliks i the ier ad outer regio. A trasmissio i a cell ier regio that is assiged a specific frequecy bad causes iterferece oly to ier users of other cells that are assiged the same bad. Furthermore, it is ecessary to distiguish two categories of base statios. The first cosists of all iterferig base statios trasmittig to cell ier users o the same sub-bad as user x ad the secod cosists of all iterferig base statios trasmittig to cell-edge users o the same sub-bad as user x. After the SINR estimatio, we proceed with the throughput calculatio. The capacity of user x o subcarrier ca be calculated by the followig equatio [8]: (1) C =Δf log (1 + SINR ) (2) x, 2 x, where, Δf refers to the available badwidth for each subcarrier divided by the umber of users that share the specific subcarrier. Moreover, the throughput of the user x ca be expressed as follows: Figure 1. Proposed frequecy bad allocatio. T = β C (3) x x, x, 1876
where, β x, represets the subcarrier assiged to macro user x. Whe β x, = 1, the subcarrier is assiged to macro user x. Otherwise, β x, = 0. Moreover, i order to evaluate the performace of our experimets we defie the term User Satisfactio (US) as the sum of the users throughput divided by the product of the maximum user s throughput ad the umber of users (X). This metric physically presets how close the user s throughput is to the maximum throughput i the area. Specifically: X Tx x= 1 US = max_ user _ throughput * X US rages betwee 0 ad 1. Whe US approaches 1, all users i the correspodig cell experiece similar throughput, while whe US approaches 0, there is a big differece i the throughput achieved by the users i the cell. B. Mechaism Descriptio The mechaism assumes a umber of multicast users that are uiformly distributed i the topology. I order to fid the optimal FFR scheme, the mechaism divides each cell ito two regios ad calculates the total throughput ad US for the followig 26 Frequecy Allocatios (FA), assumig Frequecy Reuse 1 ad 3 for the ier ad the outer regio respectively: FA1: All (25) subcarriers are allocated i ier regio. No subcarriers are allocated i outer regio. FA2: 24 subcarriers are allocated i ier regio. 1/3 subcarrier allocated i outer regio. FA25: 1 subcarrier allocated i ier regio. 24/3 subcarrier allocated i outer regio. FA26: No subcarriers allocated i ier regio. 25/3 subcarriers allocated i outer regio. For each FA, the mechaism calculates the per-user throughput ad the mea throughput ad US. This procedure is repeated for successive ier cell radius (0 to R, where R is the cell radius). Fially, the mechaism selects the optimal FFR schemes that maximize the mea throughput ad US. % Mechaism for the optimizatio of FFR geerate_etwork_ cell &users() for r = 0:R % ier cell radius for = 0:26 % ier cell subcarriers for x = 1:X % users calculate_sir(x) calculate_capacity(x) calculate_throughput(x) calculate_mea_throughput(r,) calculate_user_satisfactio(r,) calculate_ffr_for_max_mea_throughput() calculate_ffr_for_max_ user_satisfactio() (4) The pseudo-code of the proposed FFR mechaism is preseted above. The complexity ad the ruig time of the algorithm are proportioal to the umber of users multiplied by the umber of cells i the topology. This meas that that the complexity ca be expressed as Ο(# users # cells ). IV. PERFORMANCE STUDY A. Simulatio Parameters The simulatio parameters that are ecessary for the coductio of the experimet are preseted i Table I. I detail, we cosider a system with 10MHz of badwidth (i.e. LTE) divided ito 25 sub-carriers each havig a badwidth of 375 KHz. The sceario assumed is urba cayo macro, which exists i dese urba areas served by macro-cells. Path losses are calculated accordig to Cost-Hata Model [10] ad the correlatio distace of the shadowig is set to 40m [8]. TABLE I. SIMULATION PARAMETERS Parameter Uits Value System badwidth MHz 10 Subcarriers 25 Subcarriers badwidth KHz 375 Carrier frequecy MHz 2000 Cell Radius m 250 Correlatio distace m 40 Chael model 3GPP Typical Urba Path loss db Cost 231 Hata Model BS trasmit power dbm 46 Power Noise Desity dbm/hz -174 B. Optimal FFR based o Cell Mea Throughput The first simulatio experimet presets the operatio of the mechaism whe selectig the FFR scheme that maximizes the cell mea throughput. The sceario assumes that the users are distributed uiformly i the topology, which cosists of 16 cells (Figure 2). We will focus o oe cell of the topology, (secod row ad third colum) which is highlighted i Figure 2. The specific cell cotais 30 users. As Figure 2 presets, the FFR scheme that maximizes the mea throughput for the examied cell cosists of two regios, the ier regio (yellow) has 93.5m radius ad cotais 3 users while the rest cell area (red) costitutes the outer regio ad cotais 27 users. Figure 2. Deploymet of optimal FFR scheme based o cell mea throughput (optimal ier cell radius = 93.5m). 1877
of the mechaism whe selectig the FFR scheme that maximizes the US. The topology ad user distributio are i accordace to the previous experimet. However, as depicted i Figure 5 the optimal FFR scheme occurs whe the ier regio has 150m radius ad cotais 13 users. Figure 3. 3D represetatio of cell mea throughput vs. ier regio badwidth vs. ier regio radius. Figure 3 depicts the 3D represetatio of the cell mea throughput agaist the ier regio badwidth ad radius. Accordig to this figure, the highest values of mea throughput are observed for ier regio radius betwee 80m ad 150m ad for ier regio badwidth betwee 6 MHz ad 10 MHz. However, as metioed the FFR scheme that maximizes the cell mea throughput occurs for ier regio radius equal to 93.5 m ad ier regio badwidth equal to 10 MHz. Figure 5. Deploymet of optimal FFR scheme based o user satisfactio (optimal ier regio radius = 150m). Figure 6 presets the US as a fuctio of the ier regio badwidth ad the ier regio radius. We have to remid that a value of US metric closer to 1 is the optimal sice it leads to small differeces i the values of users throughput. Therefore, accordig to Figure 6 the optimal FFR scheme correspods to the deploymet with 150m ier regio radius ad 2.06 MHz ier regio badwidth that results i US equal to 0.756. Figure 4. Cell mea throughput vs. ier regio badwidth for the optimal ier regio radius. Figure 6. 3D represetatio of user satisfactio vs. ier regio badwidth vs. ier regio radius. Figure 4 presets the cell mea throughput agaist the ier regio badwidth for the optimal ier cell radius. We observe that the mea throughput icreases liearly as the badwidth allocated i the ier regio icreases. I additio, the maximum mea throughput occurs whe all the available badwidth is allocated to the three users of the ier regio. O the other had, the mechaism does ot allocate badwidth to the remaiig 27 users of the outer regio. C. Optimal FFR based o User Satisfactio I order to overcome the ufairess i badwidth allocatio amog the users i differet regios of the cell, i Sectio III.A we itroduced the user satisfactio (US) metric, which esures that all users i a cell will experiece similar values of throughput. The experimet that follows presets the operatio Figure 7. User satisfactio vs. ier regio badwidth for the optimal ier regio radius. 1878
Fially, Figure 7 presets the way that the user satisfactio chages as the ier regio badwidth icreases. The ier regio radius equals to 150m. Accordig to this figure, as the badwidth allocated to the ier regio icreases from 0 to 2.06 MHz the US icreases ad reaches the maximum value (0.756) for badwidth equal to 2.06 MHz. Higher values of ier regio badwidth lead to US decremet, while allocatig all the available badwidth to the ier regio leads to the miimum value of US. D. Compariso of the Approaches This sectio makes a direct compariso betwee the FFR schemes that preseted i sectios IVB ad IVC. More specifically, Figure 8 presets the per-user throughput achieved by each approach. V. CONCLUSIONS & FUTURE WORK I this paper we proposed a iterferece maagemet FFR mechaism that calculates the per-user SINR, capacity ad throughput ad uses these values to calculate the cell mea throughput ad user satisfactio. Based o these values the mechaism may follow two differet approaches for selectig the optimal FFR scheme. The first approach is based o maximizatio of the cell mea throughput ad the secod is based o the maximizatio of user satisfactio. Moreover, we have coducted simulatio experimets i order to examie the operatio of the mechaism ad compare the two approaches. As a geeral coclusio, we ca say that the approach that is based o the maximizatio of user satisfactio presets a better ad fairer overall behavior. The step that follows this work could be the extesio of the mechaism i order to support the users mobility. The exted mechaism could also moitor the etwork ad users coditios ad use them as iput to decide the appropriate frequecy allocatio betwee the ier ad outer regio of the cell. Figure 8. Per-user throughput for the two approaches (based i cell mea throughput ad user satisfactio). Accordig to Figure 8, the optimal FFR scheme that is based o the maximizatio of cell mea throughput allocates the available badwidth to the three users that are located i the ier regio ad serves oly these users with very high throughput. However, this approach does ot guaratee that all the users will be served. Ideed, the 27 users that are located i the outer regio have throughput equal to 0. O the other had, the optimal FFR scheme that is based o the maximizatio of user satisfactio esures that all users will share similar values of throughput irrespectively of their locatio i the cell. Ideed, eve if the mea throughput is lower compared to the first approach, we ca coclude that the FFR scheme that is based o the maximizatio of user satisfactio presets a better ad fairer overall behavior. REFERENCES [1] IEEE stadard for local ad metropolita area etworks, Part 16: Air iterface for fixed ad mobile broadbad wireless access systems, 28 February 2006. [2] Mobile WiMAX-Part II: A comparative aalysis, Tech. Rep., WiMAX Forum, May 2006. [3] M.C. Necker, Local Iterferece Coordiatio i Cellular OFDMA Networks IEEE 66th Vehicular Techology Coferece, 2007. VTC- 2007 Fall. 2007. [4] 3GPP, Radio access etwork work group1 cotributios, http://www.3gpp.org, September 2005. [5] K. Begai, G. I. Rozsa, A. Pfeig, ad M. Telek, Performace aalysis of GSM etworks with itelliget uderlay-overlay, i Proc. 7th Iteratioal Symposium o Computers ad Commuicatios (ISCC 2002), Taormia, Italy, 2002, pp. 135141. [6] 3GPP TSG RAN WG1#42 R1-050841, Further aalysis of soft frequecy reuse scheme, Huawei, Tech. Rep., 2005. [7] 3GPP TSG RAN WG1#42 R1-050764, Iter-cell iterferece hadlig for E-UTRA, Ericsso, Tech. Rep., September 2005. [8] H. Lei, L. Zhag, X. Zhag, ad D. Yag, A Novel Multi-Cell Ofdma System Structure Usig Fractioal Frequecy Reuse, I proceedigs of the 18th Aual IEEE Iteratioal Symposium o Persoal, Idoor ad Mobile Radio Commuicatios (PIMRC'07). [9] 3GPP TSG-RAN, 3GPP TR 25.814, Physical Layer Aspects for Evolved UTRA (Release 7), v1.3.1 (2006-05). [10] S. Sauders, Ateas ad Propagatio for Wireless Commuicatio Systems, Wiley, 2000, 409 p. 1879