雾计算网络的最优能效理论与技术研究

A large amount of emerging intelligent applications are springing up with the development of fifth generation wireless communications (5G) and internet of things (IoT), it raises more comprehensive requirements for the performance parameters of communication, computation, storage, and energy efficiency, which can not be fully guaranteed by the traditional cellular networks. To address this problem, a fog network is defined as a group of nodes with sharable computing and storage resources, as well as spare spectrum for node-to-node/device-to-device (N2N/D2D) and task scheduling. It promotes more intelligent applications and services in different IoT scenarios, thanks to effective collaborations among neighboring fog nodes via spectrum sharing techniques. At the theoretical layer, the project establishes a comprehensive analytical model that considers spectrum sharing, circuit, computation and offloading energy consumptions for accurately evaluating the overall energy efficiency in heterogeneous/homogeneous fog networks. With this model, the tradeoff relationship between performance gains and energy costs in collaborative task offloading/content caching is investigated, thus enabling us to formulate the energy efficiency optimization problem for resource allocations with practical constraints in available computing and storage resources at helper nodes and unused spectrum in neighboring environments. Based on rigorous mathematical analysis, the maximal intelligent cognition, energy efficient task scheduling and content caching algorithms for fog networks are proposed to derive the optimal scheduling and caching decisions for one/many task nodes and multiple neighboring helper nodes under feasible modulation schemes and time allocations, which achieve the tradeoff relationship between energy efficiency and spectrum efficiency for large-scale resource-sharing fog networks. Finally, at the experimental verification layer, this project builds a simulation platform to verify the theoretical research results of the energy efficient resource sharing strategies and task scheduling algorithms based on the fog networks with 30 nodes.

第五代移动通信与物联网的迅猛发展将催生海量的智能应用和服务，对通信、计算、存储、能耗等系统性能指标提出了更高的综合性要求，难以在传统蜂窝通信架构中得到充分保障。本项目针对这一难题，采用雾计算网络共享频谱、计算和存储等资源的研究思路，重点提升频谱效率和能量效率。在理论层面，研究基于无线通信技术的雾计算网络架构，以及相邻节点共享计算和存储资源的机制，建立高能效的资源共享、任务分配和文件缓存的模型，提出高能效无线资源分配算法、跨节点任务调度算法、以及流行文件的共享缓存与交互机制。基于雾节点协同计算和缓存等技术，创新提出包含计算功率、发射功率和电路功率等核心参数的系统能耗模型，面向雾计算网络来优化智能认知、任务卸载、计算协同和文件缓存等策略和算法，提升整个系统的频谱效率和能量效率。在实验验证层面，通过30个节点的雾计算网络仿真来验证理论研究成果和关键技术，特别是能效优先的资源共享机制和任务调度算法

海量的智能应用和服务对通信、计算、存储、能耗等系统性能指标提出了更高的综合性要求，针对这一难题，本项目采用雾计算网络共享频谱、计算和存储资源的研究思路，建立高能效的内容缓存、资源共享和任务分配的模型，提出高能效无线资源分配算法、内容缓存算法和跨节点任务调度算法。研究了基于非正交多址接入、动态无线网络、机间通信(device to device, D2D)和大规模多入多出(Massive multiple input multiple output, Massive MIMO)的雾/边缘计算框架，其中多个任务节点任务卸载到多个附近的辅助节点执行。提出了联合任务调度和通信资源分配问题，以最小化总时延和能量消耗为目标，同时考虑实际通信和计算约束。其中资源调度包括任务、计算资源和功率、通信带宽分配。. 本项目研究了基于非正交多址接入(Non-orthogonal multiple access, NOMA)和Massive MIMO的雾计算网络系统，当多个任务节点通过NOMA或者Massive MIMO基站对一个帮助节点进行任务调度时，需要具有丰富计算资源的帮助节点从任务节点计算计算任务。我们提出了一个联合任务调度、计算资源分配和功率分配的优化问题，目标是使总成本(即总的能量消耗以及时间开销)最小化。实现了能量与延迟的权衡。对于这样一个组合优化问题，如何获得最优策略是一个挑战。为此，我们提出了一个基于在线学习的优化框架来解决这个问题。仿真结果表明，与传统的算法相比，该方案大大降低了总成本消耗。. 研究了基于无线通信技术的雾计算网络架构，以及相邻节点共享计算和存储资源的机制，建立高能效的存储以及计算资源共享的模型，提出高能效无线资源分配算法、跨节点任务调度算法、以及流行文件的共享缓存与交互机制。基于雾节点协同计算和缓存等技术，创新提出包含计算功率、发射功率和电路功率等核心参数的系统能耗模型，面向雾计算网络来优化任务卸载和文件缓存等策略和算法，提升整个系统的频谱效率和能量效率。