基于多媒体业务感知的D2D无线缓存网络关键技术研究

As an efficient way to solve high data traffic demand in 5G network, D2D technology has widely concerned by a great many researchers in recent years. D2D wireless caching network, which combines caching architecture and D2D technology, is a new direction of future wireless communication. It is facing new challenges on multimedia collaboration in D2D wireless caching network as the characteristics of dynamic node movement and heterogeneity and diversity of multimedia services in D2D wireless caching network. This project studies the new methods of multimedia collaboration services and caching in D2D wireless caching network from the perspective of multimedia communication and network dynamic environment, and uses multi-objective optimization and cognitive theory as research tools. The main contents are as follows. Firstly, a multi-dimensional evolution time variant graph model of D2D wireless caching network is established. Secondly, cooperative caching mechanism based on multimedia service perception is designed in the study of this project. Thirdly, cooperative incentive mechanism under the restriction of selfishness of nodes is studied. Finally, Quality of Experience is evaluated and forcasted with using machine learning theory. Through the study of this project, a frame of theoretical system will be designed for multimedia collaboration services and caching in D2D wireless caching network. Moreover, the core mechanism of cooperative caching and motivating schemes for multimedia sevices will be defined. These results will provide theoretical guidance for the application of future multimedia sevice in D2D wireless caching network.

D2D通信作为解决5G网络中高数据流量需求的有效途径，近年来一直受到国内外学者的广泛关注。D2D无线缓存网络融合了D2D通信技术与缓存架构，成为未来无线通信的一个新发展方向。由于D2D无线缓存网络环境具有节点动态移动和多媒体业务异质性、多样性等特点，使D2D无线缓存网络中多媒体协作面临新的挑战。本项目从多媒体业务通信、网络动态化环境出发，结合多目标优化、认知理论等方法，研究D2D无线缓存网络中多媒体资源的协作服务和缓存的新方法。主要内容包括：1）建立多维度D2D无线缓存网络动态演化图模型；2）设计基于多媒体业务感知的协作缓存机制；3）研究节点自私性限制下的协作激励机制；4）开发可靠的QoE评估和预测体系。通过项目研究，将面向D2D无线缓存网络，提出多媒体协作服务和缓存的理论体系框架，明确多媒体协作缓存和激励的核心机制，为未来多媒体业务在D2D无线缓存网络中的应用提供理论指导。

D2D通信作为第五代移动通信系统的关键技术之一，与传统的蜂窝网络相比具有更高的信道增益、更高的传输速率、更低的通信能耗、更高的频谱利用率，其研发与应用成为未来无线通信发展的一个新方向。本项目的研究目标是，从多媒体通信、网络动态化环境多视角出发，通过多目标优化的方法，结合凸优化、博弈论、图论、深度学习和认知理论等方法，在分析D2D网络用户行为特征和多媒体服务特点的基础上，建立D2D无线缓存网络模型，完成对D2D无线缓存网络基于业务感知的协作缓存机制和激励机制，并依据用户体验质量QoE以衡量和调整网络性能，重点包括：（1）以节点为研究对象建立了一个D2D终端运动和传输模型，同时考虑文件冗余和中断中缓存的空闲丢失了的影响，设计了自适应缓存策略LAAC；（2）提出了一种基于深度学习和博弈论的，支持D2D的蜂窝网络的社交感知缓存方案（GDC），并使用鹰鸽博弈模型来计算可以使整个网络达到纳什平衡的用户比例；（3）基于辅助设备的可移动性，我们考虑了使用D2D技术卸载辅助设备上的数据问题，设计了利用辅助设备进行数据卸载的传输网络，并以系统成本为指标，对D2D缓存系统成本效用函数进行建模分析；（4）借鉴了贪心算法，以D2D通信距离最小化为贪心策略，基本达到在不影响系统吞吐量的前提下，通过加入移动预测模型来减少信息交互次数、采用动态发射功率来减小实时发射功率以降低系统能耗的目标。. 综上所述，本项目在D2D无线缓存网络中用户移动性预测、社交网络的D2D缓存机制研究、辅助节点在D2D无线缓存网络中的应用与性能优化、D2D用户配对等方面取得了丰富的成果，这些研究成果对D2D通信在实践中的推广具有重要价值。