ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 10 ENHANCING SPECTRAL EFFICIENCY FOR LTE-A D VA N C E D A N D B E Y O N D C E L L U L A R N E T W O R K S ENHANCING SPECTRAL-ENERGY EFFICIENCY FOR LTE-ADVANCED HETEROGENEOUS NETWORKS: A USERS SOCIAL PATTERN PERSPECTIVE XING ZHANG, YAN ZHANG, RONG YU, WENBO WANG, AND MOHSEN GUIZANI ABSTRACT The development of LTE-Advanced and beyond cellular networks is expected to offer considerably higher data rates than the existing 3G networks. Among the many potential technologies in LTE-Advanced systems, users’ characteristics and social behavior have been studied to improve the networks’ performance. In this article we present the concept of user social pattern (USP), which characterizes the general user behavior, pattern, and rules of a group of users in a social manner, and utilize USP as an optimization basis for network performance enhancement. From large-scale traffic traces collected from current mobile cellular networks, the USP model is evaluated and verified. Furthermore, to evaluate the potential of spectral efficiency and energy efficiency enhancement based on USP in LTE-A HetNets, we establish a complete system and link-level HetNet simulation platform according to 3GPP LTE-A standards. Then, based on the platform, simulations are performed to evaluate the impact of USP on spectral and energy efficiency in an LTE-A network, and a USP-based spectral efficiency and energy efficiency enhancement scheme is proposed for a HetNet of the LTE-A system. Simulation results validate that USP can be used as an effective concept for network performance optimization in an LTE-A system. Xing Zhang and Wenbo Wang are with Beijing University of Posts and Telecommunications. Yan Zhang is with Simula Research Laboratory. Rong Yu is with Guangdong University of Technology. Mohsen Guizani is with Qatar University. 10 INTRODUCTION During recent years and with the development of smart phones such as iPhone/Android, wireless data and multimedia traffic have been increasing exponentially, particularly at hotspot/ indoor areas. To meet the rapid growth and demands for high-data-rate wireless broadband services, significant effort has been made toward the development of Long Term EvolutionAdvanced (LTE-A) networks, which are expected to offer considerably higher data rates than the existing third generation (3G) networks. Meanwhile, as services migrate from voice cen- 1536-1284/14/25.00 © 2014 IEEE tric to data and multimedia centric, which requires increased link budget and coverage extension to provide uniform user experience, traditional networks optimized for homogeneous traffic face unprecedented challenges to meet the various demands effectively. Most recently, the Third Generation Partnership Project (3GPP) LTE-A has started a new study to investigate heterogeneous network (HetNet) [1, 2] deployments as an efficient way to improve the system spectral efficiency as well as effectively enhance network coverage. Under this architecture, a number of wireless technologies such as small cell enhancement [3], device-to-device (D2D) [4], and multicell cooperation have been proposed and/or developed in order to enhance spectral and energy efficiency (SE and EE). Especially in the architecture of an LTE-A HetNet, the deployment and configuration of small cells is a key technology to provide high capacity, good coverage, and high SE. In the last several years, social networks [5] have attracted billions of active users, and the number of users are increasing exponentially. Social networking applications such as Facebook, Twitter, and Micro-blog are providing valuable social information on contacts and their relationships. For example, some users exhibit a kind of social pattern/behavior and tend to operate with similar characteristics. More and more research interest has focused on users and traffic characteristics to optimize cellular networks. Different from the study of connections among users for traditional social networking [6], in this article we consider the spatial and temporal characteristics of users in cellular networks by taking into account their social characteristics. Due to the social nature of human habits, the probability that users close in vicinity (e.g., in the coverage of one or several base stations) have similar habits, pattern/behavior, and mobility rules will be high. On the other hand, because of the convergent characteristics of LTE-A HetNets with small cell enhancement, it is very suitable for the optimization of such a system to exploit rules of user behavior. For instance, in IEEE Wireless Communications • April 2014 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 11 Core network aGW S1 Macro UE SeNB Small UE SeNB Macro UE SeNB Small UE X2 MeNB SeNB MeNB Small UE Macro UE Macro UE Macro UE Macro UE X2 Macro UE X2 Macro UE SeNB Small UE User group A To meet rapid growing demands for high data-rate wireless broadband services, a significant effort has been made toward the development of LTE-advanced networks, which are expected to offer considerably higher data rates than the existing 3G networks. MeNB: Macrocell eNB SeNB: Small-cell eNB MeNB SeNB User group B Macro UE: Macrocell UE Small UE: Small-cell UE Figure 1. Architecture of heterogeneous networks for an LTE-Advanced system. [8, 9], two energy-efficient transmission control schemes in cellular networks for real-time and best-effort services are proposed, exploiting the user pattern/behavior. In this way, understanding and modeling such user patterns is crucial for the design of LTE-A HetNet systems and services, especially for the deployment and configuration of small cells. User patterns can be used as the basis for the performance optimization of the LTE-A system. Motivated from social group users and the Gini coefficient [10], we present the concept of the user social pattern (USP), which characterizes the general user behavior, pattern, and rules of a group of users as a social entity in cellular networks, such as user requirements (services/ applications), users’ social characteristics, user traffic fluctuations (in both temporal and spatial domain), and users’ mobility. Furthermore, different from the current technologies to enhance SE, we study a new solution for SE and EE enhancements from a USP perspective, specifically for the LTE-A HetNet architecture. The scope of this article is hence to study SE and EE enhancement in an LTE-A HetNet system from the USP point of view, and to explain how user social patterns can be exploited to optimize and analyze LTE-A networks. First, we introduce the architecture of an LTE-A system and HetNets with small cell enhancement. Then the concept of the USP is presented for LTE-A HetNet enhancement, which is further demonstrated from real traffic traces collected from cellular networks. We analyze and evaluate the SE and EE for LTE-A HetNets, exploiting USPs extensively through simulations. IEEE Wireless Communications • April 2014 HETEROGENEOUS NETWORKS FOR THE LTE-ADVANCED SYSTEM To meet rapidly growing demands for high-datarate wireless broadband services, significant effort has been made toward the development of LTE-A networks, which are expected to offer considerably higher data rates than the existing 3G networks. In order to meet the requirements of LTE-A (e.g., peak data rates of up to 1 Gb/s), more spectrum bands are needed. Besides the existing carriers for 3G networks, spectrum bands located at 450–470 MHz, 698–790 MHz, 2.3–2.4 GHz, and 3.4–3.6 GHz can be used for the deployment of LTE and LTE-A networks. Moreover, LTE-A has been defined to support scalable carrier bandwidth exceeding 20 MHz, potentially up to 100 MHz, in a variety of carriers for deployments. The current radio access network (RAN) for LTE-A consists of a single node, that is, the eNodeB (eNB) that provides the user plane and control plane protocol terminations toward the user equipment (UE). It is a fully distributed RAN architecture, where eNBs may be interconnected with each other by means of the X2 interface. Meanwhile, eNBs are connected through the S1 interface to the core network. In each eNB, there exist the physical (PHY), medium access control (MAC), radio link control (RLC), and Packet Data Control Protocol (PDCP) layers that implement the functionality of user plane header compression and encryption. Network coverage and highdata-rate requirements in hotspot and indoor environments have brought new challenges to LTE systems. 11 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 12 Many network configurations can be determined based on the users’ characteristics, e.g., the optimal density of small cells, the sleeping mode control for small cells, homogeneous/ heterogeneous CoMP, and the coordination of macro and small cells. In recent studies and in the standard specifications (e.g., 3GPP R11 and R12), in order to further improve the system capacity and EE, especially for hotspot and indoor environments, HetNets and their enhancements have been proposed and studied. In Fig. 1 the architecture of a HetNet for an LTE-Advanced system is illustrated. In order to meet the traffic and QoS requirements, the SE and EE of HetNets should be improved considerably compared to current 3G and LTE systems. Among the different technologies that have been proposed and studied in 3GPP HetNet, small cell enhancement is one of the major techniques. The benefits of using small cell enhancement include: • Flexible small cell deployment according to users and traffic distributions • Optimized small cell mobility by reducing RAN to core network signaling • Increased data rates by using macro and small cells together • Energy saving by using dynamic small cell sleeping In Fig. 1, under the coverage of one macrocell eNB (MeNB) there are several hotspot areas (e.g., offices, shops, and classrooms). In this situation, several small cells (SeNBs) are deployed dynamically according to certain criteria (e.g., SE, EE, and quality of experience [QoE]). In the coverage of one MeNB, due to the social habits of humans, the users (both macro UE and small UE) and their traffic volume will be heterogenously distributed and exhibit a convergent pattern in both the temporal and spatial domains. Under this condition, these small cells can be managed smartly. Many network configurations can be determined based on users’ characteristics (e.g., the optimal density of small cells, sleeping mode control for small cells, homogeneous/heterogeneous CoMP, and the coordination of macro and small cells). Thus, understanding such user behavior is very crucial and beneficial for the design of LTE-A HetNet systems. In the next section, the USP, describing user distributions and characteristics, is presented and studied. USER SOCIAL PATTERN: DEFINITION AND MODELING MOTIVATION AND CONCEPT OF USP In current wireless cellular networks, the traffic requested by different users is diverse and dynamically changing in both the temporal and spatial domains. Large amounts of traffic are requested by users distributed in relatively small hotspot regions, while much less traffic is requested by users in other regions. We define the hotspot density as the ratio of the hotspot area over the whole coverage area. It denotes the degree of how much hotspot area there is. In Fig. 1, user group A and user group B are two kinds of social groups in office and home areas; these two groups can be regarded as two hotspots covered by small cells. These two groups will probably contribute most of the data amount in the macro-eNB (MeNB), while the other area in the MeNB will contribute less traf- 12 fic. Thus, the distributions and characteristics of these groups will be important for the system design. On the other hand, in the time dimension, a large number of users may request traffic during peak hours (e.g., 10 a.m.–12 a.m., 20 p.m.–22 p.m.). Besides this, in the service (application) domain, various services (applications) may gather in various regions for a certain (short) time period. This user pattern exhibits strong social characteristics, and we want to find a simple yet effective parameter or metric to describe this inhomogeneous phenomenon. Here, we denote this as USP, which characterizes the general user behavior, and the patterns and rules of a group of users in a social manner. To mathematically describe the USP, we take the Gini coefficient [10] in statistics and economics as a reference. The Gini coefficient (also known as the Gini index or Gini ratio) is a measure of statistical dispersion intended to represent the income distribution of a nation’s residents. Similar to incomes, users and traffic in HetNets can also be described and modeled. In the following sections, two key parameters for the USP model, curve r(x) and coefficient h, are described. MODELING OF A USER SOCIAL PATTERN Figure 2 shows the illustration of user social pattern. Figures 2a and 2b describe the temporal and spatial characteristics, respectively. To quantitatively model the proposed USP in heterogeneous networks, we take the spatial domain as an example. The modeling of the temporal and/or service/application domains is similar. In a cellular network with users distributed across X km × Y km area, as shown in Fig. 2b. The calculation for the curve r(x) is written as follows: First, we subdivide the X × Y area into N small regions (these regions can be the same size of small cells) and obtain the number of users or traffic volume in each region, that is, {ui, i = 1, 2, …, n}. Second, according to u i , we sort all the regions in ascending order as u (i) : u (1) < u (2) <…< u(n). Finally, the curve is then obtained as the ratio of cumulative amount of u (i) and the total amount, which can be expressed as ρ (x) = Σ i ≤ nx u(i ) Σ i ≤ n u(i ) , 0 ≤ x ≤1 Under this process, it is seen that if the users are more equally distributed, the curve is closer to the 45˚ line. But when the users are more convergent (centralized), the curve will be closer to the lower right corner. Meanwhile, in order to characterize the degree of USP based on the definition of Gini index [1], we further present the coefficient h of USP based on the curve r(x), whose range is from 0 to 1. In Fig. 2, the coefficient of USP is calculated as the ratio of areas A and B [10] (shown in Fig. 2c), h = 1− 1 A = 1 − 2 ∫ ρ ( x ) dx. 0 A+ B Then the value of coefficient h (0 h 1) IEEE Wireless Communications • April 2014 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 13 can be used as a metric of the USP degree. A larger h shows more user convergence in the spatial or temporal domain, while a smaller h shows more equally distributed users. Temporal domain Traffic volume for a typical day (24 hours) 80 USP FOR DIFFERENT TRAFFIC DISTRIBUTIONS Sn-1 70 A smaller coefficient h indicates a more equal distribution, with zero corresponding to complete equality. A higher coefficient indicates that users are more centralized, with one corresponding to complete convergence. In this subsection, USP coefficient h for various probability distributions are presented. For a distribution with mean m and standard deviation s, for constant distribution, h = 0. For exponential distribution, h = 0.5. For lognormal distribution, Number of users 60 S3 50 S2 40 30 Sn 20 S1 10 0 h = erf ⎛⎜ 0.5 ln ⎡⎣1 + (σ / μ )2 ⎤⎦ ⎞⎟ . ⎝ ⎠ 0 × 2 F1 (1, 2(σ / μ ) + 1;(σ / μ ) 9 12 16 Time (hour) (a) 18 21 24 A cellular coverage area of X km * Y km (σ / μ )−2 Γ ((σ / μ )−2 + 1 / 2) π Γ ((σ / μ )−2 + 2) −2 6 Spatial domain For Gamma distribution, h = 1− 3 −2 U1 U2 U3 Non-hotspot + 2;1 / 2) where () is the Gamma function, and 2F1(a, b; c; z) is the Gaussian hypergeometric function. In the next subsection, a large amount of traffic data is collected from cellular networks, and the USP model is studied and verified by real traffic traces. Hotspot Un-1 USP EVALUATION FROM LARGE-SCALE DATA SETS OF CELLULAR NETWORKS To evaluate the USP model, massive real traffic traces of three weeks were collected from CDMA2000 cellular networks of China Telecom (one of the three operators in China with more than 186 million mobile subscribers, 103 million of which are 3G mobile subscribers until December 2013). Two metropolitan cities in China, Nanjing and Chengdu, are selected for the USP evaluation; the former is located in the eastern region, while the latter is in the far west. Through analysis we find that the users are distributed according to a Gamma distribution in the spatial domain. In Fig. 3, the coefficient h of the USP for theoretical analysis and real data analysis are compared. The theoretical curve is calculated based on the h equation for the Gamma distribution earlier, while each dot is calculated based on one snapshot’s traffic data. From the comparison, we can see that the coefficient of USP h can fit the real traffic data of a cellular network well. Furthermore, for a city, the users’ social characteristics of different suburban, urban, and dense urban regions may be very different. To describe the USP for different regions, we compare the USP curves and calculate the h coefficient for Chengdu city regions. Figures 4a and 4b show the 2 km × 2 km urban and dense urban areas, respectively. The blue (+) dots represent the locations of base stations; Voronoi cells [11] are used to decide the coverage of cells based on IEEE Wireless Communications • April 2014 Hotspot Un (b) 2 p(x) A 0.5 USB curve 0.28 B 0.14 0.06 0 0.2 0.3 0.5 0.8 1 x (c) Figure 2. Illustration of user social pattern modeling: a) the temporal domain: traffic volumes in one typical day; s1, s2 … denotes the user traffic amount in different time intervals; b) the spatial domain: a cellular coverage area, u1, u2 ... denote the user traffic amount in small regions; c) the curve for USP r(x). 13 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 14 1 1 Nanjing Chengdu 0.9 0.9 0.85 0.85 0.8 0.8 h 0.95 h 0.95 0.75 0.75 0.7 0.7 0.65 0.65 Theory Real data Theory Real data 1 1.5 2 2.5 3 3.5 σ/μ 4 4.5 5 5.5 6 1 2 3 4 5 σ/μ 6 7 8 9 Figure 3. User social pattern coefficient h for spatial evaluation of cellular networks in two cities of China: Nanjing and Chengdu. The x-axis shows the CV (coefficient of variation, s/m) for a given area of 3 × 6 km, three weeks traffic (Sept. 2012) are collected. locations of the base stations. The boundaries of cells are depicted using blue lines. In Figs. 4a and 4b, the red dots show that users are distributed among Voronoi cells. In Fig. 4c, the three USP curves are illustrated and compared, showing that the social pattern for the dense urban area is more intense (coefficient h = 0.83), while users in the suburban area are more equally scattered (coefficient h = 0.4). Through our analysis, the advantages of using the USP can be summarized as follows: • Generally describing the users’ social convergence characteristics (inhomogeneous distribution) in the temporal or spatial domain • Reflecting real users’ behavior based on real network traffic traces • Easy to use with only one parameter — coefficient of USP h ENHANCING SPECTRAL/ENERGY EFFICIENCY BY EXPLOITING USER SOCIAL PATTERNS In the current HetNet small cell enhancement (SCE) evaluation standard [3], in the coverage of one macrocell, there are several hotspots (socalled small cell cluster), the SeNBs are uniformly distributed among the area of a hotspot, and the users are uniformly distributed among the coverage of the SeNB. In this way, hotspot characteristics of users and the traffic can be modeled. In this section, we focus on the optimization of HetNet exploiting such non-uniform characteristics of users and traffic. Spectral and energy efficiency enhancement exploiting USP is studied. For a given area with non-uniform distributed users and traffic, based on the user social pattern (h), proper configuration of HetNet to enhance the spectral efficiency is determined, such as SeNB density lm, MeNB density lM, transmitting power for SeNB, and MeNB Pm and PM. 14 SPECTRAL EFFICIENCY IMPROVEMENT EXPLOITING USER SOCIAL PATTERNS In this subsection, the performance of the spectral efficiency exploiting USP in an LTEAdvanced HetNet system under different scenarios is evaluated through OPNET-based system-level and Matlab-based link-level simulation. The simulation parameters, including the channel model and system configurations, are based on 3GPP specifications [12] and the FP7 EARTH project [13]. In the simulation platform, 19 cells/3 sectors per cell are simulated with wraparound. The inter-site distance (ISD) is chosen as 500 m. In each MeNB, there are a fixed number of hotspots uniformly distributed across the coverage of the MeNB. In each hotspot there are several users uniformly distributed. The path loss model for MeNB-to-UE link is 131.12 + 42.8log(R) in dB, R in kilometers; while for SeNB-to-UE link, the line of sight (LOS) and non-LOS (NLOS) link use 103.8 + 20.9log(R) and 145.4 + 37.5log(R), respectively. Under these simulation configurations, we can calculate the USP coefficient h based on the earlier proposed method. To compare the performance of various h, since the total number of users and hotspots are fixed, we then vary the coefficient h through changing the coverage of each hotspot. In Fig. 5, we compare the spectral efficiency for different deployments of small cells under various coefficients of USP h. When h increases, the optimal number of SeNBs for maximizing spectral efficiency tends to increase. This is because the coverage of each hotspot will be smaller when h becomes larger. In this way, the hotspots will be more isolated. In order to cover these hotspots, more SeNBs should be needed to achieve the maximal spectral efficiency. On the other hand, as shown in Fig. 5, for a given USP coefficient h, when the number of SeNB N increases, the spec- IEEE Wireless Communications • April 2014 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 15 tral efficiency first increases. When N reaches a certain point, the spectral efficiency begins to decrease. This can be explained as follows. More SeNBs will cause severe interference between the links of SeNB-SeNB and SeNB-MeNB. That is, for a certain USP, there is an optimal number of small cells for maximizing spectral efficiency. The proposed USP can be used as an indicator for choosing the optimal small cells for the best spectral efficiency. In the practical deployment of an LTE-Advanced HetNet system, first, the user and traffic distributions are collected in a given coverage area; then the USP is estimated and the curve and coefficient h is calculated; and finally, based on the value of h and configurations of HetNet, the optimal number of small cells N can be determined as the deployment scheme. ENERGY EFFICIENCY IMPROVEMENT EXPLOITING USP For the energy efficiency improvement of LTE-A heterogeneous networks, in addition to the transmitting power of eNBs, static power of MeNB PMc and SeNB Pmc should be considered. For an energy-efficient HetNet with capacity C, the density and transmitting power for MeNB and SeNB should be optimized, under the condition of USP h and minimum data rate requirement for the cell boundary users, Rmin. We then formulate the EE problem as λm ,λ M , Pm , PM Subject to EE = C 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 IEEE Wireless Communications • April 2014 1 1.2 1.4 1.6 1.8 1.2 1.4 1.6 1.8 0.8 0.9 2 1.6 1.4 1.2 0.8 High spectral efficiency and energy efficiency are expected for LTE-A HetNets. Among the many technologies to improve the performance of LTE-A HetNet systems, enhancement from insight into users’ characteristics is a new 0.8 1.8 λ M ( PM + PMc ) CONCLUSION 0.6 Dense urban 1 We can use the classic Lagrange multiplier method to solve this optimization problem; the network configurations in an LTE-A HetNet system (i.e., lm, lM, Pm, PM) can be determined to improve the EE. For the baseline scenario without using USP in HetNet, the density of SeNB l m is chosen based on the density of the number of hotspots. The transmitting power is also fixed; in this case the EE can also be calculated based on Eq. 2. In Fig. 6, we show the energy efficiency improvement of our HetNet configuration scheme compared to that without using the USP. Three different scenarios with low, medium, and high hotspot densities are studied, the hotspot density denoting the degree of hotspot area in a cell. It is shown that the EE can be improved by an average of 80 percent when the USP coefficient h = 0.4. While users are more convergent (i.e., h = 0.6), the improvement can reach more than 200 percent. Meanwhile, it is easily seen that for a certain h, the EE improvement becomes higher when the hotspot density is large. 0.4 2 λ m ( Pm + Pmc ) + h; Rmin . 0.2 (a) 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 2 (b) Suburban: h=0.40 Urban: h=0.75 Dense urban: h=0.83 1 0.9 0.8 0.7 0.6 p(x) max Urban 2 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0 0.6 0.7 1 x (c) Figure 4. a) Urban; b) dense urban scenarios; c) user social pattern. The areas are 2 km × 2 km; the cell boundary is processed using Voronoi cells. 15 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 16 16 effective basis for network performance optimization in LTE-A systems. h=0.80 h=0.65 h=0.40 ACKNOWLEDGMENT This research is supported by the National 973 Program of China under grant 2012CB316005, by the National Science Foundation of China (NSFC) (61372114, 61370159, U1035001, U1201253), by the Beijing Higher Education Young Elite Teacher Project (No.YETP0434), by the European Commission FP7 Project EVANS (2010-269323), and by the SmartGrids ERA-Net project PRO-NET funded through the Research Council of Norway (project 217006). This work has been partially supported by the European Commission COST Actions IC0902, IC0905, and IC1004. Spectral efficiency (bps/Hz) 14 12 10 8 Optimal number of SeNBs for maximizing SE 6 REFERENCES 4 10 5 15 20 Number of SeNBs 25 30 Figure 5. Spectral efficiency for user social pattern under different number of SeNBs. 250 Energy efficiency improvement [%] Hopspot density=9% Hopspot density=15% Hopspot density=23% 200 150 100 50 0 0.4 0.5 h 0.6 Figure 6. Energy efficiency (EE) improvement exploiting user social pattern. The y-axis shows EE improvement [%] compared to the scheme without using USP. paradigm. In this article, the concept of a user social pattern (USP) is presented to characterize the general user behavior, pattern, and rules of a group of users in a social manner, which is further evaluated based on real traffic traces collected from mobile cellular networks. Our results show that the proposed USP model can effectively describe the convergence phenomenon of humans in cellular networks, which is utilized as an optimization metric for system performance improvement. Based on the model of USP, USPbased spectral efficiency and energy efficiency enhancement schemes are proposed and evaluated for LTE-A HetNet systems. Simulation results validate that the USP concept can be used as an 16 [1] R. Q. Hu et al., “Hetnets: A New Paradigm for Increasing Cellular Capacity and Coverage,” IEEE Wireless Commun., vol. 18, no. 3, June 2011, pp. 8–9. [2] A. Damnjanovic et al., “A Survey on 3GPP Heterogeneous Networks,” IEEE Wireless Commun., vol. 18, no. 3, June 2011, pp. 10–21. [3] 3GPP TR 36.932 V12.1.0, “Scenarios and Requirements for Small Cell Enhancements for E-UTRA and E-UTRAN,” Mar. 2013. [4] K. Doppler et al., “Device-to-Device Communication as an Underlay to LTE-Advanced Networks,” IEEE Commun. Mag., vol. 47, no. 12, Dec. 2009, pp. 42–49. [5] B. Wellman, “Computer Networks as Social Networks,” vol. 293, no. 5537, Science, Sept. 2001, pp. 2031–34. [6] C. Song et al., “Limits of Predictability in Human Mobility,” Science, vol. 327, no. 5968, Feb. 2010, pp. 1018–21. [7] B. Azimdoost et al., “Capacity of Wireless Networks with Social Behavior,” IEEE Trans. Wireless Commun., vol. 12, no. 1, Jan. 2013, pp. 60–69. [8] Y. Huang et al., “Analysis and Design of Energy Efficient Traffic Transmission Scheme based on User Convergence Behavior in Wireless System,” Proc. IEEE PIMRC 2012, Sydney, Australia, Sept. 2012, pp. 815–19. [9] Y. Huang et al., “An Energy Efficient Multicast Transmission Scheme with Patching Stream Exploiting User Behavior in Wireless Networks,” Proc. IEEE GLOBECOM ’12, Anaheim, CA, Dec. 2012, pp. 3537–41. [10] C. Gini, “Memorie di Metodologia Statistica,” Variabilitae Concentrazione, vol. 1, 1912; http://en.wikipedia. org/wiki/Gini-coefficient. [11] Voronoi Cell, http://en.wikipedia.org/wiki/Voronoi_ diagram. [12] 3GPP TR 36.814 V9.0.0, “Evolved Universal Terrestrial Radio Access (E-UTRA); Further Advancements for EUTRA Physical Layer Aspects,” Mar. 2010. [13] EU FP7 EARTH Project, https://www.ict-earth.eu/publications/deliverables/deliverables.html. BIOGRAPHIES X ING Z HANG [M’10] (zhangx@ieee.org) received his Ph.D. degree from Beijing University of Posts and Telecommunications (BUPT), China, in 2007. Since July 2007, he has been with the School of Information and Communications Engineering, BUPT, where he is currently an associate professor. His research interests are mainly wireless communications and networks, green communications and energy-efficient design, cognitive radio and cooperative communications, traffic modeling, and network optimization. He is the author/co-author of two technical books and has published more than 40 papers in top journals and international conferences, and filed 25 patents (11 granted). He has served as a TPC member for a number of major international conferences, including IEEE ICC, IEEE GLOBECOM, and the Wireless Communications and Networking Conference (WCNC). YAN ZHANG [SM’10] (yanzhang@ieee.org) received a Ph.D. degree from Nanyang Technological University, Singapore. He works at Simula Research Laboratory, Norway, and is an adjunct associate professor at the University of Oslo, Norway. He is an Associate Editor or Guest Editor of a number of international journals. He serves as an Organizing Committee Chair for many international conferences. His IEEE Wireless Communications • April 2014 ZHANG1_LAYOUT.qxp_Layout 4/29/14 8:14 PM Page 17 research interests include resource, mobility, spectrum, energy, and data management in wireless communications and networking. RONG YU [S’05, M’08] (yurong@ieee.org) received his Ph.D. from Tsinghua University, China, in 2007. After that, he worked in the School of Electronic and Information Engineering of South China University of Technology (SCUT). In 2010, he joined the Institute of Intelligent Information Processing at Guangdong University of Technology (GDUT), where he is now a full professor. His research interest mainly focuses on wireless communications and mobile computing, including cognitive radio, machine-to-machine communications, wireless sensor networks, home networks, and vehicular networks. He is a co-inventor of 15 patents, and author or co-author of over 70 international journal and conference papers. He is currently serving as the Deputy Secretary General of the Internet of Things (IoT) Industry Alliance, Guangdong, China, and Deputy Head of the IoT Engineering Center, Guangdong, China. He is a member of the Home Networking Standard Committee in China, where he leads the standardization work of three standards. WENBO WANG (wbwang@bupt.edu.cn) received B.S., M.S., and Ph.D. degrees from BUPT in 1986, 1989, and 1992, respectively. He is currently a professor and executive vice dean of the Graduate School, BUPT. Currently, he is the assistant drector of the Key Laboratory of Universal Wireless Communication, Ministry of Education. He has published more than 200 journal and international conference papers, and six books. His current research interests include IEEE Wireless Communications • April 2014 radio transmission technology, wireless network theory, and software radio technology. M O H S E N G U I Z A N I [S’85, M’89, SM’99, F’09] (mguizani@ieee.org) is currently a professor and associate vice president for Graduate Studies at Qatar University, Doha. He was chair of the Computer Science Department at Western Michigan University from 2002 to 2006, and of the Computer Science Department at the University of West Florida from 1999 to 2002. He also served in academic positions at the University of Missouri-Kansas City, University of Colorado-Boulder, Syracuse University, and Kuwait University. He received his B.S. (with distinction) and M.S. degrees in electrical engineering, and M.S. and Ph.D. degrees in computer engineering in 1984, 1986, 1987, and 1990, respectively, from Syracuse University, New York. His research interests include wireless communications and mobile computing, security, cloud computing, and smart grid. He currently serves on the editorial boards of six technical Journals, and is the Founder and Editor-inChief of the Wireless Communications and Mobile Computing” Journal (Wiley) (http://www.interscience.wiley.com/ jpages/1530-8669/). He is the author of eight books and more than 300 publications in refereed journals and conferences. He guest edited a number of special issues in IEEE journals and magazines. He has also served as a member, chair, and general chair of a number of conferences. He served as Chair of the IEEE Communications Society Wireless Technical Committee (WTC) and of the TAOS Technical Committee. He was an IEEE Computer Society Distinguished Lecturer from 2003 to 2005. He is a Senior member of ACM. 17