面向蜂窝网业务卸载的D2D通信关键技术研究

The spurting growth of mobile data traffic makes the spectrum resource be in shortage and the core network of operators overload, which becomes a bottleneck restricting the development of mobile communications, thereby, the effective and easily implemented offloading technology is urgent to be investigated. This project is to study on the key technologies of D2D communication to realize the traffic offloading of cellular network and improve the offloading and resource efficiency. The detailed research contents and objects are as follows. The delivering user selection algorithm is studied based on the combination of social and communication layers by using social relationship and prediction model to improve the valid coverage of the delivering users over content requests, as a result, the selection redundancy of delivering users is reduced and the serving efficiency for content requests is improved. The mode selection with high accuracy and low complexity is tried to be realized based on the prediction algorithm of machine learning by using SVM model to model the mode selection problem of traffic offloading. The intelligent caching scheme is studied by using decision model of content caching, which can significantly reduce the control overhead while improve caching efficiency compared to other peer-to-peer techniques. In addition, the studied caching scheme can increase the matching degree between cached contents and requested contents from neighbor users. Through theoretical analysis, simulations and some necessary experiments, the proposed algorithms will be applied to realistic cellular systems in the future.

移动数据业务井喷式增长使得频谱资源短缺、运营商核心网不堪重负，成为制约移动通信发展的瓶颈，研究有效且易于实施的卸载技术迫在眉睫。本项目拟对D2D通信关键技术进行深入研究，实现D2D技术对蜂窝业务的卸载，提高卸载效率和频谱利用率。具体研究内容及目标是：研究基于结合社交层与通信层的分发用户选择算法，利用社交关系和预测模型提高分发用户对内容需求用户覆盖的有效性，解决分发用户选择冗余度高、服务效率低的问题；将业务卸载时的传输模式选择问题建模为SVM模型，尝试利用机器学习的预测算法实现准确性高、复杂度低的模式选择；研究智能的内容缓存算法，建立内容缓存判决模型，在相比与其它点对点技术可以更加有效缓存的同时，又能大幅度减少控制开销，提高缓存内容与邻居用户需求内容的匹配度。通过深入的理论分析、计算机仿真并辅之必要的实验，使得提出的算法能应用到未来的蜂窝网中。

移动数据业务井喷式增长使得频谱资源短缺、运营商核心网不堪重负，成为制约移动通信发展的瓶颈，研究有效且易于实施的卸载技术迫在眉睫。本项目对卸载蜂窝网业务时的device-to-device (D2D)通信关键技术进行深入研究，实现D2D技术对蜂窝业务的卸载，提高卸载效率和资源利用率。本项目重点研究了基于社交网络的分发用户选择算法、分发模式选择算法以及基于智能内容缓存算法。本项目依次完成了上述内容的算法研究，提出了可应用于移动蜂窝的D2D缓存及分发技术算法。算法可以实现智能的缓存以及高效的分发，并能根据网络参数合理的选择分发模式。仿真证明本研究中理论分析的正确性，且性能较已有算法得到大幅度提高，例如，提出的分发算法可降低传输的中断概率达92%。通过本项目对分发用户的选择、分发模式选择以及内容的智能缓存的研究，为移动蜂窝网络提供了有效的边缘缓存策略，可有效的卸载网络业务，也为扩大网络容量，提高网络服务性能提供了可行方案。