基于负载均衡与上下文感知的D2D网络接入策略研究

The massive and super-dense hierarchal cellular networks will be utilized in the coming mobile communication systems. D2D (Device-to-Device) technology, as one of the key technologies for setting up a bridge over the next generation mobile communication system and the Internet of Things, turns out to be a leading technique, which enables devices to build up massive local connectivity directly with each other. Our team proposed an access strategy for D2D networks based on load balancing and context aware, and proposed a methodology for access mode selection method based on joint decision, in regard of the diversity of transmission and access in D2D networks as well as terminal variability. Our team also created the D2D access strategy based on Software Defined Network technique and D2D route plotting strategy based on the convex optimization theory, explored the solution for managing and controlling network load in terms of overall perspective and the best way for designing routes as well as setting bandwidth in the view of the optimal network performance. Considering that the characteristic information from large scale of D2D network is not mined and utilized well, our team will create a context data source analysis framework for communication system with Deep Learning, in order to guarantee the high transmission and computation efficiency of the D2D context information. The contribution of the work would provide a feasible solution for the establishment and access selection of D2D systems, the load balancing and route planning for D2D networks, the analysis and utilization of context information in communication systems and so forth.

未来移动通信系统将采用大规模和超密集的网络部署。作为下一代移动通信系统引入物联网的关键技术，D2D连接必将成为邻近区域内建立大量本地互联的主流技术。项目组针对D2D无线接入的网络多样性和终端可变性等特点，提出基于负载均衡与上下文感知的D2D网络接入策略及基于联合决策的接入模式选择方法，创建了基于软件定义网络（SDN）的接入策略和基于凸优化理论的D2D路径规划策略，探索从全局角度管理和调度网络负载的解决方案，从网络整体性能最优的角度设计路径及带宽配置方法；针对大规模D2D终端接入的特征信息挖掘利用不充分问题，将构建通信系统的上下文数据分析架构，对数据来源进行深度学习分类和统计数据建模，解决通信领域上下文信息挖掘问题。本项目的研究成果将为D2D通信系统的建立与接入选择、D2D网络负载均衡与路径规划、通信网络上下文信息分析与利用等关键科学问题提供可行方案。

为了实现大规模和超密集的网络部署，D2D连接必将成为下一代移动通信系统引入物联网的关键技术。项目组针对D2D无线接入网络多样性和终端可变性等特点，从D2D通信的无线接入与模式选择、D2D通信负载均衡与路径规划以及基于上下文感知的网络数据分析三方面，进行了D2D网络接入策略研究。在现有的D2D通信网络架构下，项目组提出了基于干扰图的干扰控制、基于优先级、差异化服务质量和Stackelberg博弈的资源分配算法、联合功率分配和中继选择的能效优化方案；为在中继辅助的D2D通信中规划路径、均衡负载，项目组提出了基于社交感知的中继选择、基于贪婪思想的D2D模式选择与信道分配的机制、基于启发式的D2D信道选择算法，以及利用Dinkelbach、拉格朗日乘子、Q学习、匈牙利方法的功率控制、中继选择和信道分配的能效优化与资源分配算法；通过分析获取的网络数据信息，项目组提出了基于速率需求的D2D下行链路资源分配算法、基于进化策略和深度Q网络的垂直切换决策算法和基于社交关系的D2D多播通信资源分配算法。项目成果在满足个性化服务质量需求并实现高效决策，降低接入失败率，提高系统的吞吐量、能效、无线资源利用率和收敛性等方面具有突出表现，为构建未来移动通信系统提供理论依据和方案。