Optimization of Interference Coordination Schemes in Device-to-Device(D2D) Communication

Transcription

1 2012 7th International ICST Conference on Communications and Networking in China (CHINACOM) Optimization of Interference Coordination Schemes in Device-to-Device(D2D) Communication Si WEN, Xiaoyue ZHU, Zhesheng LIN, Xin ZHANG, Dacheng Yang Wireless Theories and Technologies Lab (WT&T) Beijing University of Posts and Telecommunications, Beijing , China Abstract Device-to-Device (D2D) communication is considered to be a promising resource reuse technology for local services, to meet the system demands of higher data rates and spectrum efficiency in future networks. In order to limit severe interference between cellular links and local D2D links, previous studies mainly focus on one or two functional blocks, such as power control and resource allocation, which seems inadequate for further exploiting D2D s potential. In this paper, several interference coordination schemes, including centralized power control, resource allocation and mode selection strategy and distributed self-optimization mechanism, are proposed. Based on the proposed optimization schemes, we provide the system performance of the D2D-enabled network and analyze the impact of local service ratio and user density. Numerical results show that D2D underlay communication can increase spectrum efficiency and provide more local service opportunities compared to pure cellular communication. Index Terms device-to-device communication; spectrum sharing; mode selection; interference coordination; capacity gain I. INTRODUCTION Recently, major effort has been spent on the development of next-generation wireless communication system such as 3GPP Long Term Evolution (LTE) and WiMAX. In addition to traditional performance targets for high data rates and better coverage, these systems will enable various new services, such as file sharing services and multimedia game services. In order to handle such local services, much research work has been done to enhance the spectrum efficiency, such as cognitive radio (CR) technology[1] and device-to-device (D2D) communication[2]. A direct device-to-device radio usually operates as an underlay to a cellular network. Instead of transmitting data via the evolved Node B(eNB), D2D terminals are allowed to communicate directly with each other or through multi-hop. Therefore, it is regarded as a promising add-on component to traditional networks for the attractive spectrum sharing gain and perfect solution to tremendous local services in future networks[2]. In addition, studies in [3, 4] demonstrate several other benefits of D2D communication, such as reliability, flexibility, power-saving and plug-and-play convenience. However, enabling D2D links in a cellular network presents a challenge in radio resource management due to the severe The work in this paper is supported by the Fundamental Research Funds for the Central Universities. interference between cellular networks and local communication communities. During the cellular uplink transmission, the enb is the victim receiver of interference from all the D2D transmitters. On the other hand, D2D receivers may also suffer from the transmission of nearby cellular users. Such interference is harmful to both cellular user and D2D user and will result in high packet loss rate and low Quality of Service (QoS) satisfactory degree. In order to utilize D2D technology in handling local services without causing much damage to cellular networks, a lot of interference management schemes have been proposed. These studies are mainly based on the assumption that D2D communication underlaying cellular network being controlled by the enb and all the evolved channel state information is available to the enb. Studies in [5] and [6] propose practical and efficient resource allocation schemes for generating local awareness of the interference between cellular and D2D terminals at the base station, which then exploits the multi-user diversity to minimize the mutual interference. Besides, centralized power control mechanisms are investigated in [7] and [8], mainly by restrict D2D transmit power to ensure cellular link quality. Therefore, in case that D2D terminals are close to enb, underlay D2D link can not set up due to interference power limit. In order to find a proper mode for local area communication, mode selection schemes are discussed in [9] and [10], mainly based on optimizing system capacity. In this paper, we introduce the concept of D2D communication and set up a general model for D2D interference management. The principle of interference management is to maximize the system throughput while guaranteeing the QoS requirements of cellular and D2D users. Based on the system equations, we investigate schemes of several important functional blocks, including mode selection, resource allocation, power control, to guarantee high capacity of local connectivity and meet QoS requirement of the cellular network. However, the overhead of signaling for channel state feedback and centralized control informatioroadcasting is always a big problem. Therefore, we also discuss distributed algorithms of self-optimization. Simulation results show that D2D communication can provide a significant capacity gain in comparison to traditional communication strategies. Finally, we also find out that several key factors make great influence on local area networks, such as probability of local service and density of D2D pairs /12/$ IEEE

2 will suffer great interference from UE1 if the isolatioetween them is not enough. Assume that there are N users in the system, and some of them have a probability to be evolved in a D2D link. Some useful symbols are defined as: Fig. 1. Illustration of the links of different communication modes and interference between devices. The remainder of this paper is organized as follows. In Section II, we present the interference model of the hybrid network and analyze the system equations. In Section III, various centralized interference coordination schemes, such as mode selection, resource allocation, power control and a distributed self-optimization mechanism are proposed. Further, we evaluate the spectrum efficiency gain of D2D underlay network by single-cell simulation and analyze the performance of the proposed schemes in multi-cell scenario in Section IV. Finally, we summarize our results and the conclusions are given. II. FRAMEWORK A. System Model In a D2D-enabled cellular network, users can communicate directly with each other as a D2D user or via enb as a cellular user. As a D2D user, there are two modes of direct transmission, i.e., overlay D2D and underlay D2D. Overlay D2D communication occupies dedicated resources, while underlay D2D links share spectrum resources with cellular links. When the D2D communication takes place as an underlay communication to the cellular network, both cellular and D2D users suffer from mutual interference due to non-orthogonal resource sharing. Fig. 1 depicts the transmission mode and interference between different links. In the analysis of interference management mechanism, we assume that the enb has knowledge of current channel state and QoS requirements of all users. This cae implemented by channel estimation and feedback. With such information, the enb can make the best decisions on system optimization. B. Optimization Analysis In this part, we derive the objective function for system optimization through the calculation of Signal to Interference plus Noise Ratio (SINR) and throughput for both cellular and D2D users. As cae seen in Fig. 1, UE1 and UE2 are both in cellular mode and UE 3,4 and 5,6 communicate with a direct link. When D2D UE3 chooses underlay mode and share the same resources with UE1, its transmit power will be an interference to the cellular link. At the same time, D2D user L Number of Resource Blocks in the System i, j User device index with values 1 i, j N k RB index with values 1 k L PL Total path loss between Tx. and Rx. β i SINR for user i m i Link mode of user i (m i =1for D2D link and m i =0for cellular link) n Noise power at the receiver IoT Interference from adjacent cells R t,i Target Rate for user i ε t,i Threshold of Packet Error Ratio for user i According to Fig. 1, when user 1 chooses the cellular mode, its SINR cae expressed by: β c,ik = p ik /P L b,ik (1) N p jk /P L b,jk + IoT b + j=1 Similarly, SINR of user 3 with underlay D2D mode is given as follows (PL u,iik means the inner path loss of direct link): β d,ik = p ik /P L u,iik n (2) p jk /P L u,ijk + IoT d + n u j=1 To give a general expression, m i is introduced to denote the communication mode of user i, then the SINR of user i is: β ik =(1 m i ) β c,ij + m i β d,ij (3) Since β ik reflects current channel state and decides Modulation and Coding Scheme (MCS), it affects the transmit rate directly. We define F (β ik,ε)=r ik to express the mapping from β ik to transmit rate r ik. Here ε denotes the packet error ratio (PER) which is limited by QoS requirements. With these equations and constraints, we can find the optimal mode selection vector m = {m 1,m 2,...,m N } and resource allocation matrix P = {p ik, 1 i N and 0 k L}, in which p ik denotes the transmitting power of user i on RB k (if RB k is not allocated to user i, p ik =0). subject to arg max N i=1 k=1 L r ik (4) F (β ik,ε i,k )=r ik (5) ε i,k ε (6) L r ik R t,i (7) k=1 543

3 0 b p i,k P i (1 a b L and p i,k 0) (8) k=a Formula (6) reflects the PER limit of service while formula (7) means the total rate of user i is targeted to be greater than the Guaranteed Bit Rate (GBR). Formula (8) is the power and resource allocation constraints of LTE uplink, which must ensure the allocated RBs are continuous and total power not beyond the maximum power of User Equipment (UE). Based on the above system functions, various interference coordination schemes are proposed in the next section. III. UPLINK D2D COMMUNICATION In this part, centralized interference coordination schemes such as mode selection of devices, user pairing mechanism, power control method, will be described in detail. In addition, we propose a new distributed self-optimization scheme with less signaling overhead. All of the proposed schemes aiming at managing mutual interference cae used separately or together to enable stable and harmless D2D communication. A. Mode Selection As presented in Section II, there are three communication modes for users in a D2D-enabled network, i.e., Underlay D2D Mode, Overlay D2D Mode and Cellular Mode. The performance of three modes vary with different services and topology relationship. For instance, when D2D terminals are close to enb, Underlay D2D Mode will cause great interference to other cellular links. On the other hand, when D2D terminals are far from enb and isolated from other mobile stations, Underlay D2D Mode can increase system capacity, and guarantee the quality of other connections. Thus, mode selection optimizatioecomes a critical issue to the hybrid network. In the following, we will present a mode selection metric based on throughput performance and QoS requirements. In the scheduling phase, device communication mode is decided by the estimation of system capacity. For example, there are one cellular user and a D2D pair, and the total rate of network cae given as follows: R u = log(1 + p 1/P L b,1 )+log(1 + p 4/P L 3,4 ), + p 4 /P L b,4 + p 1 /P L 1,3 (9) β 1 = p 1/P L b,1 β c,min + p 4 /P L b,4 R o = 1 2 log(1 + p 1/P L b,1 )+ 1 2 log(1 + p 4/P L 3,4 ), (10) β 1 = p 1/P L b,1 β c,min R cellular = 1 2 log(1 + p 1/P L b,1 )+ 1 2 log(1 + p 4/P L b,4 ), (11) β 1 = p 1/P L b,1 β c,min By estimating system capacity, the mode selection scheme selects the mode with the highest sum rate that fulfills the SINR constraint for each user. Note that once the service QoS requirements change, the mode selection scheme can follow the tracks and conduct the mode transformation for each user. B. Power Control Power control is crucial in uplink due to co-channel interference and near-far effect, especially in the case that cellular links suffer from severe intra-cell interference generated by spectrum sharing. The easiest way to restrict D2D interference would be to mandate a predefined maximum power level to each D2D transmitter, and this level could be chosen such that the expected degradation of a cellular link stays at a tolerable level. However, such an approach would have to be designed for the worst case scenario and would lead to inefficient use of resources. In general, interference is controlled by limiting the total power output from all D2D users that are multiplexed onto the same resource blocks. In the following we will optimize D2D transmit power to maximum the capacity gain of the network. The old SINR of cellular user without the scheduled D2D user is β ci = p ci/p L b,ci (12) + IoT b The new SINR of cellular user sharing resource with the underlay D2D user is β ci = p ci /P L b,ci + IoT b + p dj /P L b,dj (13) SINR of the target D2D user cae expressed by β di = p di /P L dj,dj + IoT d + p ci /P L ci,dj (14) The rate gain on RB k cae calculated using Shannon s well-known formula: ΔR i = H(p di )=B i [log(1+β ci)+log(1+β di) log(1+β ci )] (15) Note that p ci, PL and IoT are independent values with the scheduled D2D user and H (p di ) 0, which means ΔR will reach the peak level when p di is at the end points of its feasible zone. Considering the rate and PER limits of QoS requirement, we can obtain a target SINR β t,i for a certain user. Since we must ensure β ci β t,i, the transmitting power which leads to the maximum capacity gain is limited to only 3 elements in the set 0 P P max : P dj { 0,min(PL b,dj p } ci,p max ) PL b,ci β t,i (16) The conclusion indicates that when a certain underlay D2D pair shares spectrum with one cellular user and other D2D pairs, the optimal transmit power should be selected from the feasible set (see formula (16)) to limit total interference to cellular links. We will also utilize the conclusion to estimate the capacity gain when allocating underlay resources. 544

4 C. Scheduling and Resource Allocation Scheduling and resource allocation play a crucial role in system performance. Since there are three kinds of users, i.e., cellular user, underlay D2D user and overlay D2D user, resource allocation scheme needs to take communication modes into account. Therefore, the general scheduling process cae expressed as follows: 1) scheduling and allocating resource for overlay users, including cellular user and overlay D2D user; 2) utilize UE pairing algorithm to optimize resource sharing process of underlay D2D users. In this paper, the proportional fair (PF) scheduling scheme [11] is adopted for overlay users. At the n-th subframe, a proportional fair (PF) metric for user i is calculated for each user as givey Q i (n) =w i (n) R i(n) α T i (n) β (17) Where R i is the achievable data rate, T i is the average throughput, w i is the priority weight, and α and β are fairness parameters. At each scheduling instance, the user with highest priority is selected for scheduling and assigned uplink resource. Overlay resource allocation process will stop if no RB is available or all the overlay users have been scheduled. Underlay D2D resource allocation is based on the resource sharing optimization process. Since D2D pairs may suffer from intra-cell interference of nearby mobile station, underlay D2D pairs prefer to share spectrum with isolated cellular UEs. In order to choose the best isolated user, we develop a pairing priority, which cae calculated by the PF priority combination. Algorithm 1 Underlay Resource Allocation Algorithm 1: Initialization 2: S c set of Overlay MS i waiting to transmit data 3: S u set of Underlay D2D i waiting to transmit data 4: for all D2D i S u do 5: for all MS j S c do 6: Calculate maximum transmit power P i,j for D2D i with formula (16) 7: Estimate R i (n) and R j (n) for D2D i and MS j 8: Calculate combined PF priority of paired UEs: Q i,j (n) =w i (n) Ri(n)α T i(n) + w β j (n) Rj(n)α T j(n) β 9: end for 10: end for 11: while S u φ do 12: Select the highest combined PF priority Q i,j (n) 13: Allow D2D i to reuse resources of MS j 14: S u S u D2D i 15: end while As demonstrated in Algorithm 1, for a certain underlay D2D user, we can utilize formula (13) and (14) to calculate the combined PF Priority on all feasible RBs and select the best overlay user which could benefit system capacity and fairness. Note that power control is considered in UE rate estimation, Fig. 2. Distributed Resource Management Procedure. so D2D interference to cellular link will be limited. On the other hand, cellular interference to D2D terminals can also be reduced due to UE pairing optimization. D. Distributed Mechanism of Self Optimization Distributed algorithms are widely studied in wireless communications, which can effectively reduce calculation complexity, the signaling overhead and result in self optimization process. In this paper, we propose a simple Distributed Resource Management (DRM) algorithm for D2D communication. Considering the D2D subsystem is like an ad-hoc system, Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) type MAC protocol cae applied for resource access. The details of a practical DRM algorithm is presented as follows: Step 1: At the beginning of a DRM period, D2D receivers with low SINR level should estimate interference level on the whole band, which cae implemented by detecting UE-specific Demodulation Reference Signal(DMRS) or Sounding Reference Signal(SRS). Based on the estimation, D2D pairs can pick out a candidate RB set for spectrum sharing. Step 2: After that, enb broadcasts resource information and D2D property messages, which can help D2D receivers to find out nearby D2D pairs and victim cellular users. The broadcasting messages should indicate UE locations, priorities, etc. Step 3: On receipt of the broadcasting messages in Step 2, D2D pairs autonomously perform DRM process to decide the best transmit RB group. Based on D2D property information, D2D receivers can detect nearby D2D pairs, and the Collision Avoidance(CA) strategy cae expressed in the following example: in case that several D2D pairs locate in a small cluster, their candidate channel sets may be the same due to similar interference relation with other mobile stations. If they all choose the first candidate to share spectrum, mutual 545

5 Single Cell Spectrum Efficiency(bps/Hz) % +92% +36% SISO SIMO MRC MU MIMO Underlay D2D TABLE I SIMULATION PARAMETERS Parameter Setting System Bandwidth 10MHz, 50 RB Cellular Layout 19 cells, 3 sector per cell Inter-site Distance 500m Number of cellular UEs per cell 10 Number of D2D pairs per cell 10,20,30 Number of Clusters per cell 1,2,3,4 Average Radius of Clusters 30m Average direct link distance 10-20m Noise Figure 5dBatBS/9dBatUE Max Transmit Power 24 dbm Path loss for cellular link Urban Macro-cell Model[13] Path loss for direct link Indoor Hotspot Model[13] Fig. 3. Spectrum efficiency comparison in single cell scenario. interference will draw back the underlay performance. However, they can avoid such collision with the help of priority information. D2D pair with highest priority choose the first candidate, while D2D pair with second-highest priority choose the second one. In this way, D2D pairs in one cluster will not transmit on the same channel and severe mutual interference cae avoided. Step 4: D2D transmitter should communicate on new RB group according to the report from D2D receiver. After that, D2D receiver can detect SINR level and decide whether to send a power increase request to enb. Step 5: The enb should reply to specific D2D transmitter according to the interference level of the cellular link and the request from D2D receiver. When cellular link suffers from underlay interference, a power decrease signal should be sent to D2D transmitter to limit D2D power. On the other hand, when D2D pairs caenefit from increasing transmit power without damage the cellular link, enb should allow its power increase request. The complete distributed resource management procedure mechanism is illustrated in Fig 2. IV. PERFORMANCE ANALYSIS In this section, we first evaluate the capacity gain for D2D communication underlaying to a LTE cellular network by the simulation of a single-cell scenario. Second, system level simulations are carried out to analyze the proposed schemes in multicell scenario. In addition, we also pay attention to several key factors that influence D2D communication performance, including the probability of local services and D2D user density. A. Single Cell Evaluation We consider a LTE cellular network operating on a 10 MHz band using FDD. There are 5 cellular users and 5 D2D pairs located randomly in a single cell. The SINR distribution at enb and D2D receivers are generated with SINR cumulative distribution function (CDF) provided by [12]. Scheduling scheme and interference cancellation algorithm proposed in Section III are used in simulation. Fig. 3 shows the spectrum efficiency of different communication mode. It is seen that D2D underlay communication can provide an impressive 90% gain over SIMO MRC receiver, and 36% gain over MU-MIMO with MMSE receiver. The impressive gain comes from efficient spectrum sharing, utilization of proper interference cancellation algorithm, and adaptability to local services. B. Multi-cell Simulation 1) Scenario, Channel Model and Simulation Parameters: In this part, a LTE-based system-level simulation is conducted to evaluate the performance of the whole system. The systemlevel simulation methodology and platform are studied and developed by ourselves, and part of the achievements have been published on [14]. So here we just give a brief introduction of the platform and list some simulation parameters different from [14]. The multi-cell scenario is of a traditional 19-cell system setup with wrap-around. Cellular users are randomly dropped in each cell, while D2D users are dropped in several clusters in each cell. The cluster number and radius can reflect local user density. Other system parameters are listed in Table I. 2) Numerical Results: Fig. 4 shows the system capacity with or without D2D communication. An observation of 220%-430% gain cae found when D2D communication is enabled. Besides, as cae seen from Fig. 4, system capacity increases almost linearly with increasing D2D service probability. When the density of D2D pairs is low and mutual interference cae limited, such conclusion is reasonable for full space multiplexing. However, when the user density becomes too high, increasing local service probability will not always result in satisfying system capacity. Fig. 5 demonstrates the harm of high density of local users. When too many D2D pairs share resources in a small local area, severe interference will decrease transmission efficiency and quality. In this case, instead of random scheduling schemes, collision avoidance algorithm should be utilized to schedule D2D pairs properly. A comparisoetween CSMA/CA DRM algorithm and random DRM algorithm is carried out in multi-cell scenario. Numerical results show that CSMA/CA algorithm can obtain an average 23% capacity gain, especially in high D2D density scenario(about 103% gain). 546

6 Fig. 4. System Capacity(Mbps) MS 10MS+10D2D 10MS+20D2D 10MS+30D2D Local Area Service Probability System capacity trend with increasing local service probability System Capacity(Mbps) % High Density +23% +9.6% Low Density CSMA/CA DRM Random DRM D2D Cluster Number Fig. 5. System capacity with different densities of D2D pairs. The 30 D2D pairs randomly distributed in 1,2,3,4 clusters with radius of 30 meters. V. CONCLUSION In this paper, we introduce the concept of D2D communication and set up a general model for D2D interference management. Centralized interference management schemes, such as mode selection, power control, scheduling and resource allocation algorithm and distributed interference coordination mechanism, are proposed to guarantee the high capacity of local connectivity and meet the QoS requirement of users in the hybrid network. Numerical analysis shows that D2D communication can significantly increase system capacity by operating as an underlay to the cellular network. The proposed schemes are also proved to be efficient in interference management, especially when the local service load is heavy. Further studies will focus on efficient distributed algorithms for a more stable D2D communication. REFERENCES [1] S. Haykin, Cognitive radio: brain-empowered wireless communications, Selected Areas in Communications, IEEE Journal on, vol. 23, no. 2, pp , [2] P. Janis, C.-H. Yu, K. Doppler, C. Ribeiro, C. Wijting, K. Hugl, O. Tirkkonen, and V. Koivunen, Device-to- Device communication underlaying cellular communication systems, International Journal on Communications, Networking and System Science, vol. 2, no. 3, pp , [3] K. Doppler, M.P. Rinne, P. Janis, C. Ribeiro, and K. Hugl, Device-to-Device Communications; Functional Prospects for LTE-Advanced Networks, IEEE International Conference on Communications Workshops, Jun. 2009, pp.1-6. [4] K. Doppler, M. Rinne, C. Wijting, C. Ribeiro, and K. Hugl, Device-todevice communication as an underlay to LTE-advanced networks, IEEE Communications Magazine, vol. 47, no. 12, pp , [5] Shaoyi Xu, Haiming Wang, Tao Chen, Qing Huang, Tao Peng, Effective Interference Cancellation Scheme for Device-to-Device Communication Underlaying Cellular Networks, Vehicular Technology Conference Fall (VTC Fall 10), Oct. 2010, pp.1-5. [6] P. Janis, V. Koivnuen, C. Ribeiro, J. Korhonen, K. Doppler, and K. Hugl, Interference-Aware Resource Allocation for Device-to-Device Radio Underlaying Cellular Networks, in Vehicular Technology Conference Spring(VTC Spring 09), Apr. 2009, pp.1-5. [7] C.H. Yu, O. Tirkkonen, K. Doppler, and C. Ribeiro, On the performance of Device-to-Device underlay communication with simple power control, Vehicular Techonlogy Conference Spring (VTC Spring 09), Apr. 2009, pp.1-5. [8] C.H. Yu, O. Tirkkonen, K. Doppler, and C. Ribeiro, Power Optimization of Device-to-Device Communication Underlaying Cellular Communication, Communications, ICC 09. IEEE International Conference on, pp1-5. [9] S. Hakola, Tao Chen, J. Lehtomaki, and T. Koskela, Device-To-Device (D2D) Communication in Cellular Network - Performance Analysis of Optimum and Practical Communication Mode Selection, Wireless Communications and Networking Conference (WCNC), Apr. 2010, pp1-6. [10] K. Doppler, C.-H. Yu, C. Ribeiro, P. Janis, Mode Selection for Device-To-Device Communication Underlaying an LTE-Advanced Network, in Wireless Communications and Networking Conference (WCNC), Apr. 2010, pp [11] H. Lei, C. Fan, X. Zhang, and D. Yang, QoS Aware Packet Scheduling Algorithm for OFDMA Systems, IEEE 64th Vehicular Technology Conference (VTC Fall 07), Oct. 2007, pp [12] CELTIC CP5-026, Calibration for IMT-Advanced Evaluations, available: [13] 3GPP TR V9.0.0, Further advancements for E-UTRA physical layer aspects, available: [14] Li Chen, Wenwen Chen, Bin Wang, Xin Zhang, Hongyang Chen, and Dacheng Yang; System-Level Simulation Methodology and Platform for Mobile Cellular Systems ; in IEEE Communication Magazine, Vol 49, NO. 7; Pages: ;