基于内容感知的D2D异构网络绿色资源分配与缓存协同技术研究

This proposal focuses on the data-sharing scenario of D2D users that are densely deployed and have similar service requests. We provide a more individualized, fine-grained optimization design for green resource allocation and cooperative caching by exploring a dimension that has been often overlooked so far, i.e., users' context. Firstly, the system model of D2D heterogeneous networks is based on the Poisson point process (PPP) stochastic geometry theory. We investigate how key system parameters such as energy efficiency are affected by interference distribution models. Then, we adopt Bayesian network and collaborative filtering to mine users' individual characteristics and design a prediction model for user’s context information. Finally, under practical constraints of battery capacity and caching capability, we have realized dynamic matching between resource allocation with interference level, channel condition, and service type by developing context-aware energy-efficient resource allocation and cooperative caching design. The achievements of this proposal have broken through the limitations of conventional models such as single type of service, scenario, and user. By taken into consideration the application scenario, service type, and user behavior that have caused the D2D communication demands, we not only develop a holistic framework for context-aware resource allocation design, but also provide a novel way for solving the high energy consumption problem faced by smart phones.

本课题针对密集部署的、具有相似业务请求的D2D用户数据共享场景，从内容感知的角度，提出了更具个性化、精细化的绿色资源分配与缓存协同优化方案。首先，基于PPP随机几何理论建立D2D异构网络系统模型，分析能效等关键参数随干扰模型变化的缩放规律；其次，利用贝叶斯网络和协同过滤算法，挖掘用户个性特征，建立用户内容请求预测模型；最后，在移动终端电池容量和缓存容量受限的条件下，通过设计基于内容感知的绿色资源分配与缓存协同算法，实现资源使用与干扰环境、信道状态、业务内容之间的动态适配，从而大幅度提升能效性能与用户体验。本课题的研究，突破了传统模型单一业务、单一场景、单一用户类型的局限性，通过考虑产生D2D通信需求的应用场景、业务类型和用户行为，不但形成了一套从内容感知角度优化资源分配的整体性框架，而且为解决智能移动终端的高能耗问题提供了一种新的研究思路和方法。

本课题针对密集部署的、具有相似业务请求的D2D用户数据共享场景，从内容感知的角度，提出了更具个性化、精细化的绿色资源分配与缓存协同优化方案。首先，建立le1单跳和多跳D2D异构网络系统模型，分析能效等关键参数随干扰模型变化的缩放规律；其次，利用机器学习和大数据分析，挖掘用户个性特征，建立用户内容请求预测模型；最后，在移动终端电池容量和缓存容量受限的条件下，通过设计基于内容感知的绿色资源分配与缓存协同算法，实现资源使用与干扰环境、信道状态、业务内容之间的动态适配，从而大幅度提升能效性能与用户体验。本课题的研究，突破了传统模型单一业务、单一场景、单一用户类型的局限性，通过考虑产生D2D通信需求的应用场景、业务类型和用户行为，不但形成了一套从内容感知角度优化资源分配的整体性框架，而且为解决智能移动终端的高能耗问题提供了一种新的研究思路和方法。本项目总体进展顺利，在项目支持下，发表论文30篇，包括21篇IEEE系列的本领域国际权威SCI期刊论文，7篇ESI前1%高被引论文，获得4项最佳论文奖。申请发明专利23项；培养博士生1名（其中在读1名），硕士生14名（其中在读10名，毕业4名）；获得教育部自然科学二等奖1项。