Resource Allocaton Optmzaton for Devce-to- Devce Communcaton Underlayng Cellular Networks Bn Wang, L Chen, Xaohang Chen, Xn Zhang, and Dacheng Yang Wreless Theores and Technologes (WT&T) Bejng Unversty of Posts and Telecommuncatons (BUPT) Bejng, 1876, Chna Emal: wangbnfgb@gmal.com, albupt@gmal.com Abstract Devce-to-Devce (D2D) communcaton wll become an mportant technology n future networks wth the ncrease of the requrements of local communcaton servces. The nterference between cellular communcaton and D2D communcaton can be coordnated by proper power control and resource allocaton. In ths paper, we analyze the resource allocaton methods for D2D communcaton underlayng cellular networks. A novel resource allocaton method that D2D can reuse the resources of more than one cellular user s proposed. After that, we dscuss the selecton of the optmal resource allocaton method from the proposed method and the conventonal methods. Fnally, the performance of dfferent methods s evaluated through numercal smulaton. The smulaton results show that the proposed method s the optmal resource allocaton method when the D2D par locates at the most part of the cell area n both uplnk and downlnk. The proposed method can mprove the sum throughput of cellular communcaton and D2D communcaton sgnfcantly. Keywords-Devce-to-Devce (D2D), resource allocaton, cellular networks, spectral effcency, energy effcency. I. INTRODUCTION Wth the applcaton of dfferent knds of multmeda servces, such as moble TV, vdeo phone, etc. there s an ncreasng requrement for hgher data rate transmsson. But the spectrum below 5 GHz s already congested. The spectrum whch s allocated to moble communcaton systems must be used effcently to meet the requrement of hgh data rate. 3GPP [1] proposes that the peak data rate wll acheve downlnk (DL) 1Mbps and uplnk (UL) 5Mbps n Long Term Evoluton (LTE) system. In order to fulfll the requrement of data rate, LTE adopts some advanced technology, e.g. MIMO, COMP, and relay, whch can ncrease the spectral effcency. There s also an ncreasng requrement for local communcaton, such as games between moble phones, vdeo sharng, and nformaton broadcast n hot pont area lke market. Machne to Machne [2] communcaton n small dstance n the Internet of Thngs also becomes a hot topc, whch s a local communcaton technology. The Devce-to-Devce (D2D) communcaton descrbed n [3] can ncrease system throughput by reusng the resource of the cellular user. The nterference between cellular communcaton and D2D communcaton can be coordnated by proper power control and resource allocaton. The D2D communcaton reduces the transmt power of termnal, whch ncreases the workng tme of termnal and mproves the energy effcency. The D2D communcaton can also decrease the load of base staton (BS) va drect transportaton. Besdes, the D2D communcaton has the advantage that lcensed spectrum s allocated to local communcaton. Lcensed spectrum can guarantee a planned envronment nstead of an uncoordnated one. D2D needs new communcaton mechansms, because of the dfference wth the conventonal cellular communcaton. In [4], the mechansms for D2D communcaton sesson setup and management nvolvng procedures n the LTE System Archtecture Evoluton are proposed. The BS s capable of coordnatng the nterference between cellular communcaton and D2D communcaton by proper power control and resource allocaton wth the channel state nformaton (CSI) of all nvolved lnks. In [5], a resource allocaton method that can mnmze the system nterference has been proposed, there were N cellular users and N D2D pars, and one D2D par reused the resource of only one cellular user n the system model. In [6] they dscussed three D2D resource allocaton methods and the power optmzaton when there was only one cellular user and one D2D par. In [7] they dscussed the mode selecton method on the bass of [6]. Two mechansms were proposed to coordnate the mutual nterference between cellular communcaton and D2D subsystems n [8]. In [9] tme hoppng (TH) based rado resource allocaton schemes amng to mprove the robustness of the hybrd network were proposed. However, n all of these references, one D2D par reused the resource of only one cellular user whenever there was one or more cellular user n the system. In actual networks, cellular users are more than D2D pars generally, so the spectrum s not used effcently when one D2D par reuses the resource of only one cellular user. In ths paper, we propose a resource allocaton method when there are more than one cellular user n the system. One D2D par can reuse the resources of more than one cellular user. The proposed method can ncrease the data rate of D2D communcaton and the spectrum effcency wth gvng prorty to the cellular user. After that, we apply the proposed 978-1-4244-8331-/11/$26. 211 IEEE
method to resource allocaton optmzaton for D2D communcaton underlayng cellular networks. We present the mprovement of sum throughput of cellular communcaton and D2D communcaton by usng the proposed method compared wth the method that one D2D par reuses the resource of only one cellular user. The rest of ths paper s organzed as follows. Secton II descrbes the system model of D2D communcaton underlayng cellular networks. The algorthm used n the proposed resource allocaton method s dscussed n Secton III. The algorthm for resource allocaton optmzaton s also presented n ths secton. Secton IV presents and dscusses the numercal results. Fnally, concluson s drawn n Secton V. II. SYSTEM MODEL The system model ncludes a sngle cell envronment as llustrated n Fg. 1, where N cellular users are R far away from the BS. We assume there s one D2D par shares the avalable rado resources wth those N cellular users. Smlar analyss can be used for the scenaro wth more than one D2D par. The dstances between the BS and both of the D2D users are D. The dstance between the two D2D users s L. UE 3 R BS UE 1 D R D L UE n UE 2 Fgure 1. System model of D2D communcaton underlayng cellular networks. We assume that the BS has the CSI of all nvolved lnks. The avalable resources are allocated to the cellular users and the D2D par based on the CSI. We propose a novel resource allocaton method that allows the D2D par reuses the resources of more than one cellular user, whch can ncrease the sum throughput of cellular communcaton and D2D communcaton. Thus, there are four resource allocaton methods n D2D communcaton underlayng cellular networks, the frst three methods are descrbed n [6], and there are only one cellular user and one D2D par n ther system model. The last one s proposed by us. Cellular mode: The D2D users could not communcate drectly. They communcate wth each other through the BS lke the cellular users. The resources whch are shared by the D2D par and cellular users are dvded nto N+2 parts unformly, and every user uses one part of them n ths mode. Dedcated resource mode: The D2D users communcate wth each other wth dedcated resource. There s no nterference between cellular communcaton and D2D communcaton snce ther resources are orthogonal. There are N+1 orthogonal transmsson lnks at ths tme. The resources whch are shared by the D2D par and cellular users are dvded nto N+1 parts unformly, and the D2D par uses one part of them n ths mode. Reusng the resource of only one cellular user: The D2D par reuses the resource of only one cellular user who ncurs the smallest nterference between them. The resources whch are shared by the D2D par and cellular users are dvded nto N parts unformly, and every cellular user uses one part of them n ths method. The D2D par reuses the resource of one of the users. Reusng the resources of more than one cellular user: Wth guaranteeng a mnmum transmsson data rate of the cellular users, the D2D par can reuse the resources of more than one cellular user. The resources whch are shared by the D2D par and cellular users are dvded nto N parts unformly, and every cellular user uses one part of them n ths method. The D2D par reuses the resources of some of the users. In the cellular and the dedcated mode, both the cellular users and the D2D par can use the maxmum transmt power to get the best system performance, because they do not nterfere wth each other wth orthogonal resource. However, the BS should control the nterference between cellular communcaton and D2D communcaton by proper power control and resource allocaton when D2D reuses the resources of one or some cellular users. When D2D reuses the resources of cellular user, we gve prorty to the cellular user by guaranteeng a mnmum transmsson data rate, whch can be acheved at an SINR of γ cl. We also assume a mnmum transmsson data rate of D2D whch s acheved at an SINR of γ dl when D2D reuses the resource of one cellular user, otherwse D2D do not reuse the resource of ths user. We assume an upper lmt on the data rate based on the hghest modulaton and codng scheme (MCS), whch can be acheved at an SINR hgher than the target SINR γ h. III. RESOURCE ALLOCATION ALGORITHM In ths secton, we study the sum throughput of cellular communcaton and D2D communcaton under dfferent resource allocaton methods, based on whch, the optmal resource allocaton method s analyzed. Accordng to the system model and resource allocaton methods descrbed n Secton II, by applyng the Shannon capacty formula [1], the sum throughput of cellular communcaton and D2D communcaton of the four methods can be expressed as:
1 N S 1 mn( Sm, Sn) Rcell = log 1 2(1 + ) + log 2(1 + ) (1) = N + 2 N N + 2 N 1 N S 1 Smn Rded= log 1 2(1 + ) + log 2(1 + ) (2) = N + 1 N N + 1 N 1 N S 1 Sk Rre _1 = log 1, 2(1 + ) + log 2(1 + ) = k N N N Idck + N (3) 1 S log 2 (1 mn + + ) N Icdk + N 1 N S Rre _ n = log 1,,..., 2(1 + ) = j k N N 1 S j S log 2 ((1 )(1 mn + + + )) N Idcj + N Icdj + N (4) 1 Sk Smn +... + log 2 ((1 + )(1 + )) N I + N I + N dck cdk where S denotes the receved power of the lnk between BS and UE, the recever s BS n UL or UE n DL. S mn denotes the receved power of the lnk between UE m and UE n. I dcj denotes the nterference receved by BS n UL or UE j n DL from the D2D communcaton when D2D reuses the resource of UE j. I cdj denotes the nterference receved by the D2D par from cellular communcaton when D2D reuses the resource of UE j. N s the thermal nose power. The formulas are adapted for both UL and DL resource reusng. A. Reusng the Resource of Only One Cellular User By applyng the Shannon capacty formula [1], the data rate of UE wthout resource reusng s depcted as follows:. Cc () = log(1 2 + ) (5) N The sum throughput of cellular communcaton and D2D communcaton when they use the same resource s calculated as follows:. Pd. gd Ccd (, ) = log 2(1 + ) + log 2(1 + ) (6) Pd. gdc + N P. gcd + N Wth gvng prorty to the cellular user, D2D wll reuse the resource of one user, whch can be determned by: k = arg max{ C( c, d) C( c)} j (7) where j { : γ η γ, γ η γ } cl c h dl d h. Pd. gd where η c =, η d =, P and g denote Pd. gdc + N. cd + N the transmt power and channel gan of the lnk between BS and UE, respectvely. P d and g d denote the transmt power and channel gan of the D2D lnk, respectvely. g cd and g dc denote the channel gan of the nterference lnk from the cellular communcaton to the D2D communcaton and that from the D2D communcaton to the cellular communcaton, respectvely. B. Reusng the Resources of More Than One Cellular User Wth guaranteeng a mnmum transmsson data rate of cellular users, the D2D par can reuse the resources of more than one cellular user to ncrease the data rate. Wth reusng the same resources, the SINR of the cellular communcaton should be hgher than γ cl, and the SINR of D2D communcaton should be hgher than γ dl. An SINR hgher than γ h does not ncrease the sum throughput because of the restrcton of the hghest MCS. D2D wll reuse the resources of some users, whch can be determned by: k {: C(, c d) C() c >, γ η γ, γ η γ } (8) cl c h dl d h We can see that the resource reusng method wll ncrease the sum throughput of cellular communcaton and D2D communcaton, because D2D reuse the resources of cellular users only f the sum throughput can be ncreased wth resource reusng as shown n (8). C. Resource Allocaton Optmzaton The BS allocates the resources to the cellular users and the D2D par wth the CSI of all nvolved lnks. Wth gvng prorty to the cellular user, the optmal resource allocaton method s the one that can lead to the hghest sum throughput of cellular communcaton and D2D communcaton, whch can be expressed as follows: R = max( R, R, R, R ) (9) max cell ded re _1 re _ n Ths s the optmal resource allocaton method n D2D communcaton underlayng cellular networks. The optmal resource allocaton method s adapted for both UL and DL resource reusng. Wth the CSI of all nvolved lnks, the BS can maxmze the sum system throughput by usng the optmal resource allocaton method whenever how many cellular users there are and how they are dstrbuted n the cell. IV. NUMERICAL SIMULATION In ths secton, we resort to the numercal smulaton to evaluate the performance of the optmal resource allocaton method selecton and the sum throughput performance of the proposed resource allocaton method presented n Secton III. We consder a normalzed crcular cell (Radus = 1). The sngle-slope path loss model Pd ( ) = Pd ( ) d α s used n the smulaton, where Pd ( ) s the receved power at the dstance d from the transmtter, Pd ( ) s the receved power at reference dstance of d, and α s the path loss exponent. We assume α = 4 for all the lnks. Pd ( ) s normalzed to the transmt power snce we consder a normalzed cellular cell. The transmt power and N are both normalzed so that the SINR of cellular communcaton s db at the cell edge wthout resource reusng. We assume γ cl =db, γ dl =-1dB, γ h =2dB [6] n both UL and DL. The sum of transmt power and antenna gan of BS s
25dB hgher than that of the cellular user. We assume that the number of cellular users N s 6. The dstance between BS and all the cellular users are R. The dstance between BS and both D2D users are D. The D2D users are L far away from each other. D2D can reuse UL resources or DL resources of the cellular network. We wll dscuss both DL and UL separately, whch means that D2D reuses DL and UL resources, respectvely. We consder the relatonshp between the optmal resource allocaton method selecton and the locaton of the transmtter of D2D n DL or the locaton of the recever of D2D n UL n the followng smulaton. The transmt power of D2D communcaton when DL resources are reused s larger than the transmt power when UL resources are reused because the transmtter s the BS n DL, whch has hgher transmt power. A. DL Smulaton The maxmal transmt power of D2D communcaton s the same as cellular users when D2D reuses DL resources. The BS can reduce the nterference between cellular communcaton and D2D communcaton by resource allocaton and power control. We consder two scenaros: the one s that the cellular users are unformly dstrbuted n the cell; the other s that the cellular users congregate n some area of the cell. The performance of the optmal resource allocaton method selecton and the sum throughput performance of the proposed resource allocaton method wll be evaluated. We assume R=.6, L=.2. The transmtter of D2D s D far away from the BS, the recever of D2D s D far away from the BS f D>L/2 or L-D far away from the BS f D<L/2. Fg. 2(a) llustrates the selecton results of the optmal resource allocaton method when the transmtter of D2D s dstrbuted n dfferent place n the cell. The color ndcates the optmal resource allocaton method when the transmtter of D2D s n the correspondng locaton. Re_n n the fgure represents that D2D reuses the resources of n cellular users. The N(=6) cellular users are dstrbuted unformly n the cell ndcated by damonds. We can see that the cellular mode s the optmal method when the D2D par locates at the cell center, because the lnks between the BS and the D2D users are n good condton. The dedcated resource mode becomes the optmal method when the D2D par gets farther away from the BS. D2D wll reuse the resources of more than one cellular user, and the number of users whose resources are reused by D2D ncreases wth the dstance between the D2D transmtter and the BS. D2D wll reuse the resources of 6 users at the cell edge. The conventonal resource allocaton method that D2D reuses the resource of only one cellular user never becomes the optmal method wherever the D2D par locates. We can see that reusng the resources of more than one cellular user s the optmal resource allocaton method n the most part of the cell area. We can also utlze the resource reusng method by reusng the resources of other users when D2D transmtter s very close to one of the cellular user. Fg. 2(b) llustrates the sum throughput gan of cellular communcaton and D2D communcaton by usng the proposed resource allocaton method compared wth the method that D2D reuses the resource of only one cellular user. The sum throughput gan s 1 when the D2D par locates at the cell center, because D2D do not take the resource reusng mode. The sum throughput gan ncreases when the D2D transmtter becomes farther away from the BS. Moreover, the sum throughput of cellular communcaton and D2D communcaton by reusng the resources of more than one cellular user s.5 tmes larger than that by reusng the resource of only one cellular user. We can see that the proposed resource allocaton method that D2D reuses the resources of more than one cellular user can use spectrum more effcently. (a) (b) Fgure 2. The N(=6) cellular users are unformly dstrbuted n the cell: (a) the optmal resource allocaton method vs. the locaton of the D2D transmtter. (b) the throughput gan vs. the locaton of the D2D transmtter. In the followng, we study the sum throughput performance n the scenaro that the cellular users congregate n some area of the cell. Fg. 3(a) llustrates the selecton results of the optmal resource allocaton method when the transmtter of D2D s dstrbuted n dfferent place n the cell when DL resources are reused. The color ndcates the optmal resource allocaton method when the transmtter of D2D s n the correspondng locaton. The N(=6) users congregate n some area of the cell ndcated by damonds. We can see that the dedcated resource mode s the optmal method when the D2D transmtter s close to the BS or the cellular users. The number of users whose resources are reused by D2D ncreases gradually when the D2D transmtter becomes farther away from the cellular users. The method that D2D reuses the
resources of 6 users s the optmal resource allocaton method n the most part n the cell. Fg. 3(b) llustrates the sum throughput gan of cellular communcaton and D2D communcaton by usng the proposed resource allocaton method compared wth the method that D2D reuses the resource of only one user when cellular users congregate n some area of the cell. The throughput gan ncreases gradually wth the ncrease of the dstance between D2D transmtter and the BS or the cellular users, and acheves almost up to 1.5 at the cell edge. (a) assume R=.6, L=.2. The recever of D2D s D far away from the BS, the transmtter of D2D s D far away from the BS f D>L/2 or L-D far away from the BS f D<L/2. Fg. 4(a) llustrates the selecton results of the optmal resource allocaton method when the recever of D2D s dstrbuted n dfferent place n the cell. The color ndcates the optmal resource allocaton method when the recever of D2D s n the correspondng locaton. The N(=6) users dstrbuted unformly n the cell ndcated by damonds. We can see that the dedcated resource mode s the optmal method when the D2D par locates at the cell center. The number of users whose resources are reused by D2D ncreases gradually when the D2D recever becomes farther away from the BS. The proposed method that D2D reuses the resources of more than one cellular user s the optmal method when the D2D par locates at the most part of the cell area. Fg. 4(b) llustrates the sum throughput gan of cellular communcaton and D2D communcaton by usng the proposed resource allocaton method compared wth the method that D2D reuses the resource of only one cellular user. The sum throughput gan s 1 when the D2D par locates at the cell center. The throughput gan ncreases gradually when the D2D recever becomes farther away from the BS, and acheves more than 1.5 at the cell edge. The sum throughput gan s sgnfcantly n the most part of the cell area. (b) Fgure 3. The N(=6) cellular users congregate n some area of the cell : (a) the optmal resource allocaton method vs. the locaton of the D2D transmtter. (b) the throughput gan vs. the locaton of the D2D transmtter (a) We can see that the resource allocaton method that we proposed s the optmal method n the most part of the cell area whenever the cellular users are dstrbuted unformly or not when DL resources are reused. The proposed method wll ncrease the sum throughput of cellular communcaton and D2D communcaton sgnfcantly. B. UL Smulaton In ths subsecton, we smulate the performance of the optmal resource allocaton method selecton and the sum throughput performance of the proposed resource allocaton method when the UL resources are reused by D2D par. The transmt power of D2D communcaton s lower than that of the cellular users by 16dB to reduce the nterference from D2D to cellular communcaton when D2D reuses UL resources. We (b) Fgure 4. The N(=6) cellular users are unformly dstrbuted n the cell: (a) the optmal resource allocaton method vs. the locaton of the D2D recever. (b) the throughput gan vs. the locaton of the D2D recever.
The results when the cellular users congregate n some area of the cell are smlar wth the results of reusng DL resources, so we do not dscuss them here. We can see that the proposed resource allocaton method that D2D reuses the resources of more than one cellular user s the optmal method n the most part of the cell area whch s the same as the DL. The proposed method can ncrease the sum throughput of cellular communcaton and D2D communcaton sgnfcantly. Throughput gan 3 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 R=.2 DL R=.4 DL R=.6 DL R=.8 DL R=.2 UL R=.4 UL R=.6 UL R=.8 UL 1.5 1 1.5 2 2.5 3 3.5 4 4.5 5 L Fgure 5. The throughput gan of the proposed method compared wth reusng the resource of only one cellular user under dfferent R and L. In the followng, we study the sum throughput gan performance when the D2D par locates at the cell edge. Fg. 5 llustrates the sum throughput gan of cellular communcaton and D2D communcaton by usng the proposed resource allocaton method compared wth the method that D2D reuses the resource of only one cellular user when D2D par locates at the cell edge under dfferent R and L. We can learn that the sum throughput gan decreases wth the ncreasng of L, because the lnk of D2D becomes worse wth larger L. D2D wll not reuse DL resources f L>.5. The sum throughput gan ncreases wth R. The nterference between D2D and the user whose resources are reused by D2D decreases wth the ncreasng of R, because the dstance between the D2D par and the user becomes larger wth R. We can see that the proposed method that D2D reuses the resources of more than one cellular user wll ncrease the sum throughput sgnfcantly whenever D2D reuses the resources of DL or UL. The proposed resource allocaton method can use spectrum more effcently. both UL and DL whenever the cellular users are dstrbuted unformly n the cell or not. The proposed method acheves better performance when the D2D par becomes closer to the cell edge. The sum throughput mprovement s also more sgnfcant wth smaller L and bgger R. REFERENCES [1] 3GPP TS 36.3, v9.1., E-UTRA and E-UTRAN; Overall descrpton, September 29. [2] R1-82813, M2M Communcatons, Alcatel-Lucent, 3GPP RAN1#54, 18-22 August 28. [3] K. Doppler, M. Rnne, P. Jäns, C. Rbero, and K. Hugl, Devce-to- Devce Communcatons; Functonal Prospects for LTE-Advanced Networks, n IEEE Internatonal Conference on Communcatons (ICC), Dresden, Germany, June 29. [4] K. Doppler, M. Rnne, C. Wjtng, C. Rbero, and K. Hugl, Devce-todevce communcaton as an underlay to LTE-advanced networks IEEE Communcatons Magazne, vol. 47, no. 12, pp. 42-49. [5] P. Jäns, V. Kovunen, C. Rbero, J. Korhonen, K. Doppler, and K. Hugl, Interference-aware resource allocaton for devce-to-devce rado underlayng cellular networks, n IEEE Vehcular Technologes Conference 29 Sprng, Barcelona, Span, Aprl 29. [6] C.-H. Yu, O. Trkkonen, K. Doppler, and C. Rbero, Power optmzaton of devce-to-devce communcaton underlayng cellular communcaton, n IEEE Internatonal Conference on Communcaton Technology, Dresden, Germany, June 29. [7] K. Doppler, C.-H. Yu, C. Rbero, and P. Jäns, Mode selecton for Devce-to-Devce Communcaton underlayng an LTE-Advanced Network, n IEEE Wreless Communcatons and Networkng Conference (WCNC), Sydney, Australa, Aprl 21. [8] T. Peng, Q. Lu, H. Wang, S. Xu, and W. Wang, Interference Avodance Mechansms n the Hybrd Cellular and Devce-to-Devce Systems, n IEEE 2 th Internatonal S ymposum on, Personal, Indoor and Moble Rado Communcatons, Tokyo, September 29. [9] T. Chen, G. Charbt, and S. Hakola, Tme Hoppng for Devce-todevce Communcaton n LTE Cellular System, n IEEE Wreless Communcatons and Networkng Conference (WCNC), Sydney, Australa, Aprl 21. [1] T. M. Cover and J. A. Thomas, Elements of nformaton theory. New York: Wley, 1991. V. CONCLUSION D2D s an advanced technology that can ncrease the spectral effcency and energy effcency by reusng the resource of cellular user. In ths paper, we studed the resource allocaton optmzaton for D2D communcaton underlayng cellular networks. We proposed a resource allocaton method that D2D can reuse the resources of more than one cellular user. We assumed the BS can select the optmal resource allocaton method wth the CSI of all nvolved lnks. The results show that the proposed method s the optmal method when D2D locates at the most part of the cell area. The sum throughput mprovement by usng the proposed method s sgnfcant n