蜂窝网络中面向D2D多播的一种混合分簇策略  

A hybrid clustering strategy for D2D multicast transmission in cellular communication networks

作　　者：江帆[1] 张兰 宫家志 JIANG Fan;ZHANG Lan;GONG Jiazhi(Shaanxi Key Laboratory of Telecommunications and Information Networks and Security,Xi'an University of Posts andTelecommunications,Xi'an 710121,China;Shaanxi Qmedia Technology Co,Ltd Xi'an 710121,China)

机构地区：[1]西安邮电大学,陕西省信息通信网络及安全重点实验室,陕西西安710121 [2]西安企鹅智媒科技有限公司,陕西西安710100

出　　处：《西安邮电大学学报》2020年第4期29-36,共8页Journal of Xi’an University of Posts and Telecommunications

基　　金：国家自然科学基金项目(62071377,61801382,61874087);陕西省自然科学重点基金项目(2019JZ-06);西安邮电大学大学生科技创新活动项目(19-B-289)。

摘　　要：面向蜂窝网络中基于设备对设备(device-to-device,D2D)多播的内容分发场景,为了有效减少分发能耗及延长网络生命周期,提出一种蜂窝网络中基于D2D多播的混合智能分簇簇略(hybrid intelligent clustering strategy,HICS)。该方案根据D2D用户对不同文件类型的偏好差异,使用自组织特征神经网络(self-organizing feature map,SOM)算法对用户进行聚类;再利用凝聚层次聚类算法(agglomerative hierarchical clustering algorithm,AHCA)对每个聚类内用户实现进一步的分簇;以最小化系统总能耗为目标,根据簇内用户的相对簇中心度、相对剩余能量、社交相似度以及距离因子等参数给出最佳的簇首选择算法。仿真结果表明,与已有方案相比,所提策略能够将具有相同文件偏好的用户更加均匀地分布在每个簇中,从而有效减少分发能耗,提高用户体验和系统性能。To effectively reduces the energy consumption and extends the network life cycle,a hybrid intelligent clustering strategy(HICS)for D2D multicast transmission is proposed in the cellular network under content distribution scenario.Firstly,according to D2D users' preference to different file types,self-organizing feature map(SOM)algorithm is adopted to form different clusters.Secondly,in order to ensure the reliable communication between each user among D2D multicast group,the agglomerative hierarchical clustering algorithm(AHCA)is applied to further adjust users within each cluster.Finally,based on parameters such as relative center degree of the users in each cluster,the relative residual energy,the social relationship similarity and distance factor,an optimal cluster head selection algorithm is proposed to minimize the total energy consumption of the system.Simulation results demonstrate that compared with the existing schemes,the proposed scheme can distribute users more evenly into each cluster according to users' preference,which can effectively reduce energy consumption and therefore improve user experience and system performance.

关 键 词：设备到设备通信 分簇 自组织特征映射神经网络 凝聚层次聚类 系统能耗 

分 类 号：TN929.52[电子电信—通信与信息系统]