System utility based resource allocation for D2D multicast communication in software-defined cellular networks

Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143

Contents lists available at ScienceDirect

International Journal of Electronics and Communications (AEÜ) journal homepage: www.elsevier.com/locate/aeue

Regular paper

System utility based resource allocation for D2D multicast communication in software-defined cellular networks q Wenrong Gong ⇑, Guomin Li, Baiping Li College of Communication and Information Engineering, Xi’an University of Science and Technology, No. 58, Yanta Road, Xi’an, China

a r t i c l e

i n f o

Article history: Received 21 March 2018 Accepted 24 August 2018

Keywords: D2D communication SDN Social aware Resource allocation Multicast

a b s t r a c t Device-to-device (D2D) multicast communication is a useful way to improve the communication efficiency of local services. This study considers a scenario of D2D multicast communication in software defined cellular network and investigates the frequency resource allocation problem. Firstly, we build the system model and formulate the optimization problem. Secondly, a hierarchical scheme to achieve a suboptimal solution is proposed. To select appropriate user equipments (UEs) as potential D2D transmitters (PDTs), a social aware PDT selection method is proposed. Then, a resource allocation algorithm considering users’ priorities is proposed. Furthermore, to study the resource allocation for general system that UEs without priorities, a non-priority considered allocation algorithm is proposed also. Numerical simulation results show that the proposed schemes are effective in improving the system utility and reducing the resource consuming for D2D communications. Ó 2018 Published by Elsevier GmbH.

1. Introduction With rapid development of smart devices, high data transmission rates applications becoming increasingly prevalent, which is putting new traffic pressure on cellular network. As an essential way of local communication, device-to-device (D2D) communication has become a research hot-spot in recent years [1,2]. Different from the traditional cellular communication, user equipments (UEs) can communicate to each other directly without the relay of base station (BS) via D2D communication, which can reduce the traffic load of the core network. D2D communication was introduced systematically in [3]. Unlike small cell BS, the setup of D2D communication is not costly, and the maintenance of D2D communication is not overly complex. As an underlay to cellular communication, D2D communication may generate harmful interference to cellular user equipments when they share the same frequency resources. Therefore, it is crucial for D2D communication networks to allocate frequency resource properly. To improve performance of D2D communication, various complicated methods have been employed in D2D communication system [4,5], such as network coding [6], graph theory method [7], game theory [8–10], and machine learning [11].

q

Fully documented templates are available in the elsarticle package on CTAN.

⇑ Corresponding author.

E-mail address: [email protected] (W. Gong). https://doi.org/10.1016/j.aeue.2018.08.030 1434-8411/Ó 2018 Published by Elsevier GmbH.

Based on the existing studies about D2D communications, there are two important issues to deal with. Firstly, considering limited memory and energy resources, not all UEs want to be D2D transmitter selflessly. Secondly, with the development of study on D2D communication, resource allocation algorithms have grown in complexity, which brings challenges to network. To deal with the first issues, researchers have begun to study socially aware based D2D communications [12–14]. A socially aware distributed caching strategy based on a learning automation for D2D communications was proposed in [12], which was proved to improve system throughput gain significantly. In [13], Wu etc. proposed a price-based multicast video distribution system and a grid-based clustering method. Furthermore, based on users’ social attributes, Wu’s team proposed a video coding sharing scheme in [14]. Those schemes have stronger applicability than traditional study of D2D communications. Software defined network (SDN) is a promising technology to settle the second problem. As a new paradigm which decouples the control plane and the data plane, SDN enabling network administrators to program the network in a dynamic and flexible manner [15,16]. There are several benefits of SDN-based D2D communication. Firstly, a hybrid D2D communication architecture where a centralized SDN controller can reduce the number of communication links thereby improving energy efficiency. Secondly, since SDN controller has a global view of the network, operators can manage and optimize resource allocation efficiently in response to time-varying network conditions. In addition, SDN

139

W. Gong et al. / Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143

enables fast control over devices in a vendor independent way by means of standardized interfaces [17]. There are some articles studying the software defined D2D communication [18]. F. R. Yu’s research team has investigated the resource allocation of SDN-based D2D communications [19,20]. Literature [20] illustrated the system model of SDN-based cellular network and proposed a resource allocation algorithm with imperfect CSI. However, those researches are mainly focusing on SDN-based D2D unicast communication. Few research studies SDN-based D2D multicast communication. With the recent popularity of smart terminals, the demand for multimedia and video services is growing rapidly. The adoption of D2D multicast technology can significantly improve the transmission efficiency with the better reuse of the spectrum. Based on the above statements, in this paper, we study about applying D2D multicast communication into software defined cellular networks (SDCNs) and propose a resource allocation scheme. Firstly, we build the system model and formulate the optimization problem. Secondly, a social aware potential D2D transmitter (PDT) selection method is proposed. Then, to obtain the improvement of system utility and guarantee the performance of UEs with high priority, a resource allocation algorithm with priority considered is proposed. Further, to study the resource allocation for general system that UEs without priorities, a non-priority considered resource allocation algorithm is proposed also. Simulation results show that the proposed schemes improve the system utility and reduce resource consumption. The contributions of our work are: (1) Software defined network technology is applied to cellular network to assist with resource allocation of D2D communications. (2) This paper mainly studies D2D multicast communication, which would lead to an improvement of resources’ utilization. (3) A social aware PDT selection method is proposed to select appropriate PDTs. (4) The resource allocation algorithms for UEs with priorities and without priorities are studied respectively and simulation results are performed. The rest of this paper is organized as follows. System model and the optimization problem are introduced in Section 2. In Section 3, we give a detailed introduction to the proposed hierarchical scheme, which includes PDT selection and resource allocation. Numerical simulation results and analysis of results are given in Section 4. Finally, Section 5 concludes the paper and highlights our findings.

2. System model and problem formulation 2.1. System model We consider a SDCN with D2D communication whose system model can be seen in Fig. 1. Components with both control and data functions in traditional network are partitioned. Specifically, Serving gateway (SGW) divided into SGW Control Unit (SGW-C) and SGW Data Unit (SGW-D). Similarly, PDN gateway (PGW) divided into PGW Control Unit (PGW-C) and PGW Data Unit (PGW-D). Mobility management entity (MME), SGW-C, PGW-C are implemented as part of the SDN controller. The control plane and data plane communicate with each other via OpenFlow protocol. The PDT m means that one UE which can be selected to be a D2D transmitter. Let U be the set of PDTs, U ¼ f1; 2; . . . ; m; . . . ; Mg; m 2 U. There is an especial case which with eNB to be transmitter. When m ¼ 0 means that eNB is the transmitter for user. Then we set U0 ¼ U [ f0g. W is the set of receivers, W ¼ f1; 2; . . . ; Ng, we denote receiver n 2 W. Therefore, there are I (I ¼ M þ N) UEs in the system. The independent homogeneous Poisson point process (HPPP) is used for the location of UEs. Receivers’ locations are modeled by a HPPP with density kr . H ¼ f1; . . . ; Kg denotes the set of all sub-channels in the system. We assume that frequency resource allocated to D2D communication is orthogonal to that of cellular communication, and the location information about each UE can be gain from SDN controller. The wireless link between two nodes is subject to independently Rayleigh fading, propagation path loss, and additive white Gaussian noise (AWGN). 2.2. Problem formulation It is assumed that receiver n has a content request C ðnÞ for netðnÞ

ðnÞ

work. Let am be the content distribution indicator, namely, am ¼ 1 indicates that content C ðnÞ is readily stored in the memory of transðnÞ

mitter m, otherwise, am ¼ 0. The size of C ðnÞ denoted by scðnÞ . SimðnÞ

ilarly, the binary variable bm denotes whether or not transmitter ðnÞ

m is allocated to receiver n. In addition, let cm;k be the allocation indicator variable, i.e., if sub-channel k is allocated to the link ðnÞ

between receiver n and transmitter m; cm;k ¼ 1, otherwise, ðnÞ cm;k

¼ 0.

ðnÞ em

¼ 1 indicates that receiver n receives content from

SDN Controller OpenFlow Protocol

UE SGW-D

PGW-D

UE eNB

SGW-D

Fig. 1. System model of software defined cellular network.

140

W. Gong et al. / Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143 ðnÞ

transmitter m successfully, and 0 otherwise. Specially, e0 ¼ 1 indicates that receiver n receives content from eNB, and 0 otherwise. The association variable

ðnÞ bm

must be larger than or equal to the ðnÞ

sub-channel allocation variable cm;k as well as the caching variable ðnÞ

em . Therefore, the following conditions should be satisfied. ðnÞ ðnÞ

ðnÞ

bm cm;k P em;k ; 8m 2 U; 8n 2 W; 8k 2 H:

X

ðnÞ

bm 6 1; 8n 2 W:

ð2Þ

m2U

There exists at most N max receivers for each D2D group, which means

X ðnÞ bm 6 Nmax ; 8m 2 U:

ð3Þ

n2W

One receiver could only caches content from at most one transmitter at a time, and one transmitter’s content could be received by at most N max receivers. Therefore, the following conditions should be satisfied,

X

eðnÞ m 6 1; 8n 2 W;

ð4Þ

X eðnÞ m 6 N max ; 8m 2 U:

ð5Þ

m2U

ðnÞ r mk

¼

if

ðnÞ

bm ¼ 1:

ð10Þ

m2U

8 ðnÞ p h > mk > Blog2 ð1 þ min ð mk ÞÞ; > N0 B < 8n2W

X

if

m

ðnÞ

bm > 1;

m2U

ðnÞ > pmk hmk > > : Blog2 ð1 þ N0 B Þ;

if

X

ðnÞ

ð11Þ

bm ¼ 1:

m2U

From the point of view of network, we denote an as the revenue of network from receiver n per unit of received data rate, bðnÞ m as the price for per unit of consumed radio bandwidth when receiver n associates with transmitter m, cm as the price of per unit of conðnÞ

sumed backhaul bandwidth for receiver n on transmitter m; /ðnÞ as the revenue of network per unit of estimated backhaul bandwidth reduction by means of caching content on receiver n, and wðnÞ as the price of per unit of space in the memory of receiver n, respectively. The utility for the link between user n to transmitter m can be calculated by ðnÞ um ¼

X ðnÞ ðnÞ ðnÞ ðnÞ ðnÞ cm;k ðan r mk  bðnÞ m B  cm ð1  am Þr mk Þ k2H ðnÞ

ðnÞ

ðnÞ ðnÞ ðnÞ ðnÞ ðnÞ ðnÞ þ eðnÞ m ð/ Be  w scðnÞ Þ þ e0 ð1  a0 Þð/ Be  w scðnÞ Þ;

n2W

ð12Þ

We assume that one subchannel can be allocated to, at most, N max potential links if and only if there has a D2D multicast group which has N max receivers.

XX n2W m2U

where BðnÞ denotes estimated backhaul bandwidth. When m ¼ 0, e ðnÞ

(i.e. the transmitter is eNB), e0 ¼ 1, the utility for the link between ðnÞ

ðnÞ

cm;k 6 Nmax ; 8k 2 H:

ð6Þ

To satisfy the caching limitations on the memory of D2D transmitters and the eNB,

X

pmk hmk Þ; N0 B

According to (9) and (10), the achievable rate by receiver n on sub-channel k, when associating with transmitter m, can be calculated as

ð1Þ

One receiver could only be connected to at most one D2D transmitter, and the constraint is denoted by

X

ðnÞ

ðnÞ

rmk ¼ Blog2 ð1 þ

eðnÞ m scðnÞ 6 Sm ;

ð7Þ

n2Wm

ðnÞ

receiver n to eNB adds the term e0 ð1  a0 Þð/ðnÞ BeðnÞ  wðnÞ scðnÞ Þ compared with that of the case m – 0. The optimization problem is modeled as maximizing the sum utility of all links between receivers to transmitters, which is given by

max

XX

uðnÞ m ;

n2W m2U0

ð13Þ

s:t:ð1Þ—ð8Þ:

X ðnÞ e0 scðnÞ 6 S0 ;

ð8Þ

n

where Sm and S0 denote the storage capacity of transmitter m and eNB, respectively. Wm denotes the set of receivers served by transmitter m. To transmitter m, if it has been connected by more than one P ðnÞ D2D receivers, namely m2U bm > 1, and the content requests of these D2D receivers are the same, then one D2D multicast communication group is established. Since the data rate of a multicast group is limited with the minimum rate of all the links in the group, to D2D multicast group with transmitter m, the achievable rate by receiver n on sub-channel k can be calculated as ðnÞ

ðnÞ

r mk ¼ Blog2 ð1 þ min ð 8n2Wm

pmk hmk ÞÞ; N0 B

if

X

The optimal solution to (13) is infeasible to achieve because discrete and continuous variables are included in the optimization problem. Furthermore, it would be impractical to gain the channel conditions of all links, which would result in a huge overhead to the central controller. Therefore, we propose a hierarchical scheme to achieve a suboptimal solution. 3. Hierarchical resource allocation scheme The sub-optimization problem is divided into two subproblems in this paper: PDTs selection and resource allocation. 3.1. Social aware PDT selection method

ðnÞ

bm > 1;

ð9Þ

m2U

where B is the bandwidth of each sub-channel. pmk is the transmit ðnÞ

power of D2D transmitter m on sub-channel k. hmk denotes the channel condition between transmitter m and receiver n on subchannel k. N 0 is the noise power per sub-channel. To transmitter m, if it has been connected by only one D2D P ðnÞ receivers, namely, m2U bm ¼ 1, then one D2D unicast communication pair is established. The achievable rate by receiver n on sub-channel k, when associating with transmitter m, can be calculated as

In order to facilitate the analysis system utility, we assume that PDTs have been selected in Section 2. However, not all UEs want to be D2D transmitter selflessly because of limited memory and energy resources. Therefore, PDTs should be selected from D2D UEs according to their characteristics. To satisfy the data rate requirement and ensure good performance of D2D communication, the energy consumption, storage space and the mobility of UEs need to be considered when selecting potential D2D transmitters. The mobility of UEs will influence the data rate performance of D2D communication. The data delivery rate of UEs can be improved when it have high stay probability. According to [13],

W. Gong et al. / Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143

stay probability of UE is related with time users remain in each cluster. The stay probability of UE i is given by:

Pi ¼

ð14Þ

where T s;i means the total history stay time in the area of a D2D cluster, and T c;i denotes the history stay time in the current cluster. T c;i and T s;i can be obtained from SDN controller. In this paper, the area of a D2D cluster is the area of D2D transmission distance radius from a centre of D2D transmitter. Fitness F can be used to measure whether one UE is an appropriate PDT. For UE i, the fitness is defined as:



lDi þ mSi þ jEi þ ePi ;

if

0;

Ei < fEi max ;

if

Ei P fEi max ;

ð15Þ

where l; m; j and e are weighting parameters satisfying l þ m þ j þ e ¼ 1 . Di denotes the social importance of UE i. Si is the available storage space of UE i. Ei denotes the available battery power of UE m. Ei max denotes the full battery power of UE i. f is a tunable parameter which value is between 0 and 1. For example, if one D2D transmission will consume one-quarter of the full battery power of a transmitter UE, f can be set to 0.3 or other appropriate value in the range between 0.25 and 1. Only when the current battery power of UE i is higher than fEi max , UE i could be chosen as a PDT. Otherwise, UE i cannot be chosen as a PDT. In this paper, we assume that single hop clustering structure is adopted in D2D communication. Therefore, no UE needs to relay information form one UE to another. The social importance of UE i can be calculated as

PI Di ¼

Table 1 The priority-considered resource allocation (PCRA) algorithm. Algorithm 1: PCRA

T c;i T s;i

Fi ¼

141

j¼1;j–i Psðdi;j Þ

I1

;

ð16Þ

where di;j is the distance between user i and user j. According to [12], Psðdi;j Þ denotes a social relationship with respect to the physical distance di;j as

(

Psðdi;j Þ ¼

1; 2

A2 =di;j ;

if

0 < di;j 6 A;

if

di;j > A:

ð17Þ

where A is a predefined distance. After calculating each user equipment’s fitness, SDN controller can get a list of the fitness of UEs in the descending order which is denoted by F ¼ fF 1 ; F 2 ;    ; F i ;   g. The M UEs corresponding to the first M elements in F are picked to be PDTs and be stored in set U, then U ¼ fUEF 1 ; UEF 2 ;    ; UEF M g. In the following, we rewrite UEF m as m for notational convenience. Then, we get U ¼ f1; 2; . . . ; Mg. These selections are carried out by SDN controller. 3.2. Resource Allocation Algorithms After finishing the processes of PDTs selection, the procedure of resource allocation can be performed. 3.2.1. The priority-considered resource allocation algorithm In this section, we propose a priority-considered resource allocation (PCRA) algorithm for D2D communications in SDCN. Based on the priority of D2D receivers, we formulate the PCRA. The PCRA is summarized in Table 1. Receivers are listed in W from highest to lowest priority. Receiver n is the highest priority UE in the set of receivers that have not been allocated resource (line 2). To receiver n, find the potential D2D transmitter which can achieve the maximum utility and mark it as mn , calculated the data rate rmn n of link between receiver n and transmitter mn . Then determine if r mn n is greater than the threshold data rate (line 4–5). If rmn n is greater

1 Input: N; M; Q ¼ ð0ÞNðMþ1Þ 2 for n ¼ 1 : N Un ¼ U, 3 ðnÞ

4 mn ¼ argm2Un maxðum Þ. 5 if rmn n P r 0 &&sumðQ ð:; mn ÞÞ < N max 6 qnmn ¼ 1, goto 2 7 else 8 Un ¼ Un n fmn g 9 if Un – £ 10 goto 4 11 else 12 qn;Mþ1 ¼ 1, goto 2. 13 end if 14 end if 15 end for 16 Output: Q NðMþ1Þ 17 SDN controller calculates the minimum number of resource blocks required for each D2D communication group and allocates corresponding number of PRBs to them.

than the threshold data rate and the number of receivers link with the current mn is smaller than the threshold number N max , choose transmitter mn as the transmitter of receiver n. Otherwise, exclude transmitter mn in Un and find the next potential D2D transmitter as new mn . Turn to execute line 4 repeatedly, until find the transmitter of receiver n. If there is no suitable PDT for receiver n, choose eNB as its transmitter. This procedure will repeat until all receivers have selected their transmitter. In order to meet the need of data rate requirement of each receiver, determine the minimum number of resource blocks required for each D2D communication group and allocate corresponding number of PRBs to them. 3.2.2. The non-priority considered resource allocation algorithm To study the resource allocation for general system that receivers without priorities, a non-priority considered resource allocation algorithm (NPCRA) is proposed. The NPCRA is summarized in Table 2. Potential D2D transmitters are listed randomly in U. Table 2 The non-priority considered resource allocation (NPCRA) algorithm. Algorithm 2: NPCRA 1 Input: N; M; Q ¼ ð0ÞNðMþ1Þ 2 for m ¼ 1 : M 3 Wm ¼ W, 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

ðnÞ

nm ¼ argn2Wm maxðum Þ if r mnm P r 0 if sumðQ ð:; mÞÞ < N max qnm m ¼ 1; W ¼ W n fnm g , Wm ¼ Wm n fnm g, if Wm – £ goto 4 end if else goto 2 end if else if Wm ¼ Wm n fnm g&&Wm – £, goto 4 end if end for for n ¼ 1 : N if sumðQ ðn; :ÞÞ ¼ 0 qn;Mþ1 ¼ 1 end if end for Output: Q NðMþ1Þ SDN controller calculates the minimum number of resource blocks required for each D2D communication group and allocates corresponding number of PRBs to them.

142

W. Gong et al. / Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143

To PDT m, find the receiver which can achieve the maximum utility and mark it as nm , calculated the data rate rmnm of link between receiver nm and transmitter m. Then determine if rmnm is greater than the threshold data rate. If r mnm is greater than the threshold data rate, mark that transmitter m as the transmitter of receiver nm , and exclude nm in W. Otherwise, exclude transmitter nm in Wm and find the next receiver, turn to execute line 4 repeatedly, until find N max receivers or all of receivers have been determined. This procedure will repeat until all potential D2D transmitters have selected their receivers. For each receiver, determine weather it have found its transmitter. If have not, choose eNB as its transmitter. Then, allocate PRBs to each receiver. 4. Simulation results In this section, we evaluate the performance of the proposed algorithms of D2D communication in SDCN. The simulation results are compared with the scheme without D2D communication and the distance based resource allocation (DBRA) scheme in [4]. In system without D2D communication, all user equipments obtain information by cellular communication. The basic parameters in simulation are shown in Table 3. Fig. 2 shows the resource consumption per receiver versus the ratio of receivers and PDTs. The resource consumption per receivers is denoted by the ratio of the sum resource blocks consumed by receivers to the number of receivers N. From this figure, we have the following observations. Firstly, the system with NPCRA algorithm consumes the least PRBs, followed by the system with PCRA algorithm. The reason for the lower consumption of resource in

Table 3 Basic simulation parameters. Parameter

Value

Maximum D2D pair distance Available bandwidth eNB Tx power D2D Tx power Noise spectral density The number of PDTs kr r0 Path-loss exponent

25 m 10 MHz 46 dBm 24 dBm 174 dBm/Hz 20 100=250p(Except for Figs. 2 and 3) 1 Mbps(Except for Figs. 4 and 5) 4 10, 10, 1, 10, 1.5

ðnÞ ðnÞ ðnÞ an ; bðnÞ m ; cm ; / ; w

N max

4

Fig. 2. Resource consumption vs. the ratio of receiver UEs and PDTs.

systems with our proposed schemes is that the adoption of D2D multicast technology has better utilization of frequency resources. Secondly, as the ratio of receivers and PDTs increases, resource consumption with NPCRA increases slowly when the ratio of receivers and PDTs is bigger than 4. It is duo to that when the ratio of receiver UEs and PDTs is smaller than the threshold number of receivers in one D2D multicast group N max , almost all receiver UEs can find a D2D transmitter. Therefore, resource consumption can be held steady in system with NPCRA. While when Rto is larger than N max , some receivers need to obtain signals via eNB, which leads more resource consumption. Thirdly, the number of PRBs consumed in the system without D2D communication is independent of the ratio of receivers and PDTs. That is because that each cellular user is allocated one PRB in cellular communication according to the data rate required. In Fig. 3, we compare the system utility under different ratio of receivers and PDTs. The simulation results show that system utilities with all the four schemes increase as the ratio of receivers and PDTs increases. In system without D2D communications, when there is a higher proportion of receivers, the sum data rate increases and then the system utility increases. In addition, the utility of system with NPCRA algorithm performs best, followed by the system with PCRA algorithm. It demonstrates that the proposed schemes are effective in improving the system utility. Fig. 4 shows the average number of PRBs costing per receiver versus the request data rate of receivers. In Fig. 4, we can obtain some conclusions. Firstly, the numbers of PRBs costing per receiver rise as the request data rate of UE increases except for the system without D2D communication. The main reason is that more resources are required to meet the increasing data rate demands of users in system with NPCRA algorithm, PCRA algorithm and DBRA algorithm. While in the system without D2D communication, one PRB per receiver is enough to satisfy the data rate request of receivers by cellular communication. Secondly, when r0 6 3 Mbps, system with NPCRA algorithm costs the lowest number of PRBs compared with systems with PCRA algorithm and DBRA algorithm. When r 0 > 3 Mbps, system with PCRA algorithm costs the lowest number of PRBs compared with NPCRA algorithm and DBRA algorithm. The cause of this phenomenon is that when the data rate request is relatively low, the system with NPCRA can complete the data transmission with fewer amounts of PRBs. To guarantee services to receivers with high priorities, the system with PCRA needs more PRBs. When the data rate request is relatively higher, the system with NPCRA needs more PRBs to ensure the data rate performance of all receivers.

Fig. 3. System utility vs. the ratio of receiver UEs and PDTs.

W. Gong et al. / Int. J. Electron. Commun. (AEÜ) 96 (2018) 138–143

143

allocation for general system that UEs without priorities, a nonpriority considered resource allocation algorithm is proposed. Simulation results show that both proposed algorithms could improve the system utility performance and reduce frequency resource consumption greatly. In our future work, social aware power allocation for D2D multicast communication in SDCN would be considered. Acknowledgement This work is supported by the China NSFC Project (Grant No. 61701393/No. 61701392), the Science Research Project of Education Department of Shaanxi Province (17JK0514), the Outstanding Youth Science Foundation of Xi’an University Of Science And Technology (2018YQ3-07), the Doctoral Scientific Research Foundation of Xi’an University Of Science And Technology (2015QDJ054), and the Talent Training Funds of Xi’an University of Science And Technology (201664). Fig. 4. Resource consumption vs. the request data rate for UEs(Mbps).

Fig. 5. System utility vs. the request data rate for UEs(Mbps).

Fig. 5 describes the system utilities under different request data rates of receivers. It can be observed that the NPCRA algorithm achieves the highest system utility, which is followed by PCRA algorithm. The system utility achieved by PCRA algorithm is lower than NPCRA algorithm since more resources are needed to serve receivers with high priority in PCRA algorithm and thus sacrificing the performance of receiver with lower priorities. Moreover, system utilities of schemes increase as the request data rate of receivers increases. It is because that the higher request of data rate makes the higher data rate obtained by receivers. Higher data rate of receivers leads to a higher system utility. 5. Conclusion Resource allocation problem for D2D multicast communication in software defined cellular network system is studied in this paper. A social aware method is proposed to select appropriate PDTs from UEs. To make better use of the available frequency resource and guarantee the performance of UEs with high priorities, a priority considered resource allocation algorithm for D2D multicast communication is proposed. To study the resource

References [1] Doppler K, Rinne M, Wijting C, Ribeiro C, Hugl K. Device-to-device communication as an underlay to lte-advanced networks. IEEE Commun Mag 2009;47(12):42–9. [2] Lei L, Zhong Z, Lin C, Shen X. Operator controlled device-to-device communications in lte-advanced networks. Wireless Commun IEEE 2012;19 (3):96–104. [3] Gandotra P, Jha RK, Jain S. A survey on device-to-device (D2D) communication: architecture and security issues. J Network Comput Appl 2017;78(C):9–29. [4] Duong Q, Shin Y, Shin OS. Distance-based resource allocation scheme for device-to-device communications underlaying cellular networks. AEUE - Int J Electron Commun 2015;69(10):1437–44. [5] Zhou Z, Ota K, Dong M, Xu C. Energy-efficient matching for resource allocation in D2D enabled cellular networks. IEEE Trans Veh Technol 2017;66 (6):5256–66. [6] Gao C, Li Y, Zhao Y, Chen S. A two-level game theory approach for joint relay selection and resource allocation in network coding assisted D2D communications. IEEE Trans Mob Comput 2017;16(10):2697–711. [7] Yang T, Zhang R, Cheng X, Yang L. Graph coloring based resource sharing (GCRS) scheme for D2D communications underlaying full-duplex cellular networks. IEEE Trans Veh Technol 2017;66(8):7506–17. [8] Zhang Y, Xu Y, Wu Q, Liu X, Yao K, Anpalagan A. A game-theoretic approach for optimal distributed cooperative hybrid caching in D2D networks. IEEE Wireless Commun Lett 2018;7(3):324–7. [9] Xu S, Xia C, Kwak KS. Overlapping coalition formation games based interference coordination for D2D underlaying lte-a networks. AEUE - Int J Electron Commun 2016;70(2):204–9. [10] Wang R, Cheng D, Zhang G, Lu Y, Yang J, Zhao L, et al. Joint relay selection and resource allocation in cooperative device-to-device communications. AEU - Int J Electron Commun 2017;73:50–8. [11] Davide Del Testa MD, Zorzi M. A machine learning-based ETA estimator for Wi-Fi transmissions. IEEE Trans Wireless Commun 2017;16(11):7011–24. [12] Ma C, Ding M, Chen H, Lin Z, Mao G, Liang YC, et al. Socially aware caching strategy in device to device communication networks. IEEE Trans Veh Technol 2018;67(5):4615–28. [13] Wu D, Yan J, Wang H, Wu D, Wang R. Social attribute aware incentive mechanism for device-to-device video distribution. IEEE Trans Multimedia 2017;19(8):1908–20. [14] Wu D, Liu Q, Wang H, Wu D, Wang R. Socially aware energy efficient mobile edge collaboration for video distribution. IEEE Trans Multimedia 2017;19 (10):2197–209. [15] Jain A, Sadagopan NS, Lohani SK, Vutukuru M. A comparison of sdn and nfv for re-designing the lte packet core. In: 2016 IEEE Conference on network function virtualization and software defined networks (NFV-SDN); 2016, p. 74–80. [16] Zhu Y, Yan F, Zhang Y, Zhang R, Shen L. SDN-based anchor scheduling scheme for localization in heterogeneous WSNs. IEEE Commun Lett 2017;21 (5):1127–30. [17] Jagadeesan NA, Krishnamachari B. Software-defined networking paradigms in wireless networks: a survey. ACM Computing Surveys 2014;47(2):1–11. [18] Usman M, Gebremariam AA, Raza U, Granelli F. A software-defined device-todevice communication architecture for public safety applications in 5G networks. IEEE Access 2015;3:1649–54. [19] Cai Y, Yu FR, Liang C, Sun B, Yan Q. Software-defined device-to-device (D2D) communications in virtual wireless networks with imperfect network state information (NSI). IEEE Trans Veh Technol 2016;65(9):7349–60. [20] Wang K, Li H, Yu FR, Wei W. Virtual resource allocation in software-defined information-centric cellular networks with device-to-device communications and imperfect CSI. IEEE Trans Veh Technol 2016;65(12):10011–21.