Prorty-based Resource Allocaton to Guarantee Handover and Mtgate Interference for OFDMA System Lex Xu, Yue Chen School of Electronc Engneerng and Computer Scence Queen Mary, Unversty of London London, UK Abstract Mtgatng Inter-cell Interference (ICI) and ensurng seamless hgh-qualty communcaton are two challengng ssues for OFDMA systems. Cell-level coordnated resource allocaton and handover (HO) are the two ey technologes for achevng these goals. They have been nvestgated ntensvely, however, manly separately. In ths paper, a novel combned Handover Guarantee and Interference Mtgaton (HGIM) cell-level resource allocaton scheme s proposed. HGIM defnes the Handover User Set (HUS) and grants hgher allocaton prorty to handover users. Other actve users are prortzed based on a unfed cell dvson model whch dvdes a cell nto dfferent ICI senstve areas. Meanwhle, HGIM defnes a Sub-carrer Preferred Lst (SPL) to optmze allocaton. Smulaton results show that HGIM acheves greater ICI mtgaton and mproved handover performance compared wth the conventonal soft frequency reuse scheme. Keywords OFDMA; Inter-cell nterference (ICI); Handover; Interference mtgaton; I. INTRODUCTION OFDMA (Orthogonal Frequency Dvson Multple Access) s consdered as the leadng multple access technque for future wreless systems, such as LTE, IMT-Advanced, as t provdes hgh spectral effcency and robust performance n a fadng envronment []. However, mult-cell OFDMA systems are senstve to Inter-cell Interference (ICI). ICI has a negatve mpact on the communcaton qualty especally for cell edge users. Inevtably, ICI affects the Handover (HO) performance as well because handover users are cell edge users. Ths mposes a challenge for future systems on coordnatng celllevel resource allocaton n order to acheve ICI mtgaton and better handover performance under the bandwdth restrcton. In recent years, cell-level resource allocaton has become a hot research topc. Most wor s based on the Soft Frequency Reuse (SFR) concept proposed n [], whch assgns a thrd of all sub-carrers to the cell edge, wth a dfferent thrd for neghbourng cells, so reducng ICI because of the frequency reuse factor of 3 for the cell edge. Sem-dynamc coordnaton has been proposed and nvestgated n [3]-[5]: ths s more flexble than SFR and has better cell edge performance. Some other schemes, such as soft fractonal reuse and other partal reuse schemes [6][7] based on the SFR concept have also been proposed, but the complex cell edge separaton causes hgh user area-selecton mstaes. For example, n [6], the cell edge part s separated nto complex areas wth dfferent resource unts. Most exstng cell-level resource allocaton schemes do not dfferentate handover users from other actve users wth the performance evaluaton focusng on the cell edge n terms of throughput and farness mprovement. Meanwhle, there has been much research on handover for OFDMA systems, such as the Receved Sgnal Strength (RSS) based hard handover for LTE [8], the fast handover for realtme servce n [9], the soft handover [], and the sem-soft handover proposed n []. Most of the wor focuses on the handover performance wthout consderng the ICI mtgaton. In ths paper, we propose a novel combned Handover Guarantee and Interference Mtgaton (HGIM) cell-level resource allocaton scheme for OFDMA systems. Handover users are dentfed and regstered n the Handover User Set (HUS): they are granted wth hgher resource allocaton prortes by mappng to the Sub-carrer Prefer Lst (SPL) before other actve users. The conventonal SFR scheme for cell-level nterference coordnaton s also modfed n ths paper. A unfed cell dvson prncple s proposed that dvdes a cell nto dfferent ICI senstvty areas assocated wth dfferent Area Interference Mtgaton Prorty (AIMP). HGIM ensures handover users n HUS tae preference n the allocaton wth good condton sub-carrers. For other actve users, HGIM allocates preferred sub-carrers from the SPL to each area accordng to AIMP. Ths leads to a greater ICI mtgaton compared to the conventonal two-part, edge and nner, cell dvson approach. The rest of the paper s organzed as follows. The HUS regstraton and the unfed cell dvson model for nterference coordnaton are presented n secton. The novel HGIM scheme s proposed and analyzed n secton 3 followed by the smulaton results n secton 4 and fnally, the paper s concluded n secton 5. II. HUS REGISTRATION & UNIFIED CELL DIVISION MODEL A. Handover Algorthm and HUS Regstraton As part of the moblty management functonalty, handover s crucal for ensurng the seamless communcaton n cellular networs. In LTE, the Reference Sgnal Receved Power (RSRP) based hard handover s chosen because of ts advantages, such as reduced sgnallng load and less handover delay [8]. The handover process conssts of 4 stages: measurements, processng, decson and executon, as shown n Fg.. T m s handover measurement perod and T u s handover decson update perod. 978--444-53-4/9/ $6. 9 IEEE 783
Fg. DL ICI analyss Fg. RSRP based handover process In the measurement stage, a User Equpment (UE) measures the RSRP from ts servng and neghbourng Base Statons (BS). The functon of the layer 3 flterng s to flter out the effect of the fast fadng and any mperfectons n the layer measurements. The handover crteron s checed n the decson stage. If the condton n () s satsfed for a preset tme wndow, the tme-to-trgger (TTT) wndow, the UE sends the measurement report to ts servng BS. RSRP TC RSRP SC + H m () RSRP TC and RSRP SC represent the RSRP from the target cell and the servng cell respectvely. H m s the handover margn, or handover hysteress. In the executon stage, the UE s allocated wth a sub-carrer from ts new servng BS. The paper assumes deal layer 3 flterng whch means fast fadng and measurement mperfectons can be totally fltered out. Hence, the handover crteron can be expressed as () APG TC APG SC + H m () APG TC and APG SC represent the Average Path Gan (APG) from the target cell and the servng cell respectvely. Assumng UEs have the same average shadowng and antenna gan, the APG can be obtaned n (3) by applyng the UMTS3.3 path loss model APG SC (db)=4lg(r SC )+3lg (f)+49 (3) Where r SC s the dstance between the UE to ts servng BS and f s the carrer frequency. All users satsfyng () wll be regstered n the HUS and wll be granted hgher resource allocaton prortes than other actve users for two man reasons: ) Users n the HUS are cell edge users and requre hgher transmsson power to meet the sgnal-to-nterference rato (SINR), ths ntroducng more ICI to other co-channel users. ) A handover falure affects user satsfacton more than a new call beng bloced. In the smulaton n ths paper we assume a unform dstrbuton of users wth 3 dfferent speeds of movement and chooses a typcal 3 db handover margn[8] as the threshold for HUS regstraton. B. Interference Analyss & Unfed Cell Dvson Model Ths paper focuses on the Downln (DL) nterference mtgaton. For OFDMA systems, ICI s the man source of the nterference as the ntra-cell nterference can be elmnated by fllng the Guard Interval wth the Cyclc Prefx of the OFDM symbol []. ICI can be analyzed usng a generc two-cell scenaro shown n Fg.. Interference analyss for mult-cell scenaros can be derved followng the same prncple. In Fg., user s served by cell A and users and are served by cell B. r represents the dstance between the user and ts servng BS and d s the dstance between the user and the adacent BS. In ths partcular scenaro, r >r and d <d.. Scene: User, are assgned non-cochannel sub-carrer wth user. In ths case, ICI =, and the SINR of user, SINR can be expressed as SINR = P ( N + ICI) [,, ] (4) Where P s the receved sgnal power of user and N s the addtve whle Gaussan nose. Assumng each user s assgned one sub-carrer and all users have the same SINR requrements, the expected receved power of each user meets P =P =P. So the transmttng power of servng BS to user can be obtaned by (5) usng the same path loss model as (3) PBS = P +4lg r + 3lg f + 49 (5) Scene : User, s assgned co-channel sub-carrer as user respectvely, they wll suffer the co-channel nterference. The ICI of user receved from BS-A can be expressed as: ICI = PBS 4lg d 3lg f 49 [, ] (6) Where P BS- s the transmttng power of BS-A to user. Because d >d, from (4)(5)(6), t can be obtan that: ICI >ICI (7) SINR < SINR (8) After sufferng severe nterference, SINR of user, could be less than the requred threshold, SINR thr. To meet SINR= SINR = SINR, the expected receved power of thr user meets P = SINRthr( ICI + N) > P. Meanwhle, the transmttng power from BS-B to user s PBS = P+4lgr+ 3lg f + 49 > P. So user wll suffer BS heaver ICI. From Fg. (b), suffx - denotes parameters of user after BS-B ncreases the transmttng power to user. The ICI of user meets: ICI = P 4lg d 3 lg f 49 [, ] (9) BS From (4), decreasng the SINR of user meets SINR < SINR. Once SINR s less than SINR thr, BS-A wll ncrease the transmttng power, then the ICI to the cell B wll contnue to ncrease. From (7) and (8), the expected receved power has the relatonshp P > P, whch ndcates the expected receved power of user s hgher than user. In 784
addton, due to r >r and P > P, The BS transmttng power comparson meets: P > P () BS BS The nterference ncrement of user from BS-B Δ ICI can be calculated as: where ICI = obtan: Δ ICI = ICI ICI [, ] () (See Scene ). From (7) (), we can Δ ICI >Δ ICI () () shows that user nduces severe ICI to user. From the above analyss, when users are assgned cochannel sub-carrers used n the adacent cells, those who are located further from ther servng BS wll receve more ICI as shown n (7). Hence, greater transmttng power s needed from ther servng BS than those users who are located nearer the cell centre (). Meanwhle, t can be seen from () that users who are located further from ther servng BS nduce more ICI to the adacent cell. Therefore, the area whch s further from the BS should be preferably allocated wth good channel condton sub-carrers for better nterference mtgaton performance. Based on the above fndngs, a unfed cell dvson model s proposed whch dvdes a cell nto dfferent nterference senstve areas. A Cell Interference Dstncton Threshold (CIDT) s ntroduced, whch specfes the mnmum nterference dfference between adacent dvded areas of each cell. In ths paper, CIDT s set to be 3dB. From (6), the ICI dfference between the two adacent areas can be expressed as: ICI d 3R r ICI( ) = 4lg = 4lg d 3R r CIDT (3) Compared to the conventonal two-part, edge and nner, cell dvson approach, the unfed cell dvson model usng CIDT has fner dvson granularty whch leads to better ICI mtgaton. On the other hand, ths unfed cell dvson model s much smpler than some of the soft fractonal frequency reuse scheme as proposed n [6][7]. Fg. 3 shows the unfed cell dvson model used n ths paper wth a cell radus R=m. Accordng to ICI dstncton n (3), each cell s dvded nto 5 areas. Areas to 5 are granted dfferent resource allocaton prortes. III. THE NOVEL HGIM SCHEME Based on the cell structure shown n Fg. 3, a novel combned Handover Guarantee and Interference Mtgaton (HGIM) cell-level resource allocaton scheme s proposed. Fg. 4 shows the flow chart of the HGIM scheme. Fg.4 HGIM cell-level resource allocaton scheme The resource allocaton can be summarzed nto three steps: A. Sub-carrer Prorty Intalzaton of Each Cell Step : HGIM ntalzes the sub-carrer prortzaton to guarantee the preferred resource of a partcular cell s less lely to be assgned n the neghbourng cells. Assume there are L sub-carrers avalable n total. They are dvded nto three orthogonal sets, Q, Q, Q 3. Q Q = ( ;,=,,3) (4) Each set s composed of N sub-carrer denoted as they meet: Q : A, A A N Q : B, B B N Q 3 : C, C C N N*3 L Where B and C wth the same suffx have the same prorty n P. Each cell assgns sub-carrers n ts own orthogonal set, (Q, Q and Q 3 ). Tae SPL- P as an example, cell- and cell-3 prefers to use Q and Q 3 respectvely, so are less lely to use Q, namely these hghest prorty sub-carrers have best channel condton. Meanwhle, after A set s allocated, P prefers to assgn low-prorty sub-carrers of adacent cell sub-carrers sets Q and Q 3 wth descendng order of B and C. Ths SPL structure ensures that low prorty sub-carrers are less lely to be used synchronously n adacent cells, hence the ICI s mtgated. Fg.3 Unfed cell dvson model for OFDMA systems 785
Fg.5 Sub-carrer Preferred Lst (SPL) structure B. HUS and Area Interference Mtgaton Prortsaton Step: The HUS of each cell s defned as Hd, Hd and Hd 3 respectvely. All users satsfy () are regstered to the relevant HUS and the allocaton prorty of HUS s set to be the hghest: PrHUS, L (5) Applyng the unfed cell dvson model dscussed n secton, snce the exteror crcular area prefers to allocate resource wth low ICI, the AIMP value of area s the lowest and area 5 has the optmal prorty shown as Table. Table Detaled Area Interference Mtgaton Prorty (AIMP) Area N.O.() area area 3 area 4 area 5 area 7-53- 698- Area Range -7 >86 53 698 86 AIMP (g ) -lowest 3 4 5(optmal) C. Resource Allocaton Step3: The actual sub-carrer allocaton starts wth users n HUS. Tang cell for example, users n Hd are allocated wth the sub-carrers n P frst. After allocaton, these subcarrers wll be deleted from the cell s SPL. When resource allocaton of HUS s completed, HGIM chooses the optmum match between the area ICI senstvty and the SPL for other actve users amng at maxmzng the ICI mtgaton. To guarantee hgh nterference senstve areas are assgned wth good condton sub-carrers, preferred subcarrers should be allocated to the wth large AIMP value. So the allocaton prorty of area on sub-carrer of ts cell SPL can be expressed as: Pr = g 5, L (6) Where sub-carrer of ts cell SPL does not contan the 5 allocated sub-carrers, t meets: r, r [,], and r = denotes area allocates sub-carrer. The allocaton order for users except HUS can be wrtten as: W = arg max Pr 5, L (7) ( ) = Sub-carrer s assgned to the maxmum then the Pr value area, and Pr value s deleted. The allocaton process wll recycle untl all areas have been assgned wth requred resources or all sub-carrers have been allocated. Fg.6 HGIM resource allocaton dagram After step 3, HUS taes the user-level handover process, and each area taes the user-level sub-carrer, power, bt allocaton wthn the allocated sub-carrers. Ths paper only consders the cell-level sub-carrer allocaton. IV. SIMULATION ANALYSIS A. Smulaton Platform Confguraton To verfy the performance of the proposed HGIM scheme, a OFDMA system-level smulaton platform was set up contanng 36 cells wth three-cell clusters. Users are randomly dstrbuted and the traffc arrval rate follows the Posson dstrbuton. Some other smulaton parameters are shown n Table. Table Smulaton parameter Parameter Assumpton Total Sub-carrer Number 9 Carrer Frequency GHz Cell Radus m Cell Dvson and AIMP 5 areas (See Table ) User Speed,5,5 (m/s) Path Loss Model 4 lg (r) + 3 lg (f) + 49, r-m Fadng Model Lognormal fadng The SFR scheme proposed n [] s used as the reference scheme n ths paper and ts parameters are: cell edge covers % of cell radus wth frequency reuse factor 3, and cell nner wth frequency reuse factor. Other parameters are the same as HGIM. B. Smulaton Results Fg.7 and Fg.8 compare the performance of the HGIM and the conventonal SFR n terms of the handover droppng probablty and the new call blocng probablty respectvely. Handover Dropng Probablty.8.7.6.5.4.3....5..5.3.35 Posson Arrval Rate(lamda) User/S Fg.7 Handover droppng probablty VS arrval rate 786
New Call Blocng Probablty.6.4...8.6.4...5..5.3.35 Posson Arrval Rate(lamda) User/S Fg.8 Call blocng probablty VS arrval rate A connecton requestng handover s dropped when none of the remanng sub-carrers on the target BS SPL can meet the SINR requrement; a call s bloced when no sub-carrer on ts servng BS SPL can meet the SINR requrement. It s clear the handover performance s mproved wth the HGIM scheme because handover users are dentfed n the HUS and granted hgher resource allocaton prorty than other actve users. Low ICI sub-carrers are allocated to the HUS that leads to great ICI mtgaton The decreased call blocng probablty s because users located n the exteror area, the area wth hgh nterference senstvty, are assgned wth better condton sub-carrers than users n the nteror area, so they are more lely to get access to the servng BS. Inter-Cell Interference (ICI) System Throughput [bps] 6 5 4 3..5..5.3.35 Posson Arrval Rate(lamda) User/S 7 6 5 4 3 Fg.9 ICI comparson.5..5..5.3.35 Posson Arrval Rate(lamda) User/S Fg. System throughput comparson same networ condton, the average ICI s lower and the average system throughput s hgher wth the HGIM scheme. Ths s also due to the optmum mappng between the ICI senstvty areas and the sub-carrers on the SPL. Greater ICI mtgaton leads to less overall nterference n the system, whch n return, ncreases the system throughput. V. CONCLUSIONS In ths paper, a novel combned Handover Guarantee and Interference Mtgaton (HGIM) cell-level resource allocaton scheme s proposed. Compared wth the exstng cell-level resource allocaton and nterference coordnaton schemes, HGIM dfferentates the handover users and grants them wth hgher prortes n the resource allocaton process; HGIM adopts a unfed cell dvson model, whch acheves better tradeoff between the algorthm complexty and the effectveness of the ICI mtgaton; Meanwhle, HGIM defnes the SPL to optmze the sub-carrer allocaton. The performance of HGIM s nvestgated va a system level smulaton and compared wth the conventonal SFR scheme. Results show that the proposed scheme acheves better handover performance and greater ICI mtgaton at the same tme. Future wor wll nvestgate the HGIM scheme s performance under more complcated system scenaros, whch ncludes mxed traffc wth dfferent QoS requrements, the non-unform load dstrbuton, and the exstence of the handover measurements errors etc. REFERENCES [] T.S.Rappaport etc, Wreless communcatons:past events and a future perspectve, IEEE Commun. Mag., vol.4, no.5, pp.48-6, May [] 3GPP TSG RAN WG Meetng #4 R-557 Athens, Greece, Huawe, 9-3 May, 5 [3] TSG-RAN WG Meetng #4 R-5738 London, UK, Semens, 9 August- September, 5 [4] 3GPP TSG RAN WG, R-559, San Dego, USA, Texas Instruments, October - 4 October, 5 [5] 3GPP TSG RAN WG Meetng #43, R-534, Seoul, Korea, Samsung, 7- November, 5 [6] Fangmng Xu, Xaofeng Tao, Xaodong Xu, Soft Fractonal Frequency Reuse, ZTE Communcatons, pp.7-9,aprl 7 [7] Zhongnan L, Yafeng Wang, A Hybrd Inter-cell Interference Mtgaton Scheme for OFDMA System, ICCS8, pp. 656-66 [8] Mohmmad, Anas etc, Performance Evaluaton of Receved Sgnal Strength Based Hard handover for UTRAN LTE, VTC7-Sprng, pp.46-5,-5 Aprl 7 [9] S Cho, Gyung-Ho Hwang etc, Fast Handover Scheme for Real-Tme Downln Servces n IEEE 8.6e BWA System, IEEE VTC 5- sprng,vol. 3, pp. 8-3, June 5 [] Antt Toll, Maran Codreanu and Maru Juntt, Soft Handover n Adaptve MIMO-OFDM Cellular System wth Cooperatve Processng, IEEE PIMRC 6, pp.-5, -4 Sept. 6 [] Hyungeu Lee, Hyumn Son, Sanghoon Lee, Semsoft Handover Gan Analyss Over OFDM-Based Broadband Systems, IEEE Trans on Vehcular Tech, vol. 58, no. 3, pp. 443-453, Mar 9 [] Changchuan Yn, Tao Luo etc, Mult-carrer wreless communcaton technology. BUPT Press, July 4 Fg.9 and Fg. shows that compared wth the SFR, HGIM s more effectve n term of ICI mtgaton because under the 787