能量收集蜂窝网络中面向移动边缘计算的资源分配理论与方法

能量收集(EH)与移动边缘计算(MEC)结合能为计算密集型和时延敏感型应用提供有力的技术支撑，构建绿色节能的自治移动通信系统。传统的EH与MEC技术无法适应EH与MEC结合时多维资源和信道状态的随机变化，也不能解决多维资源及业务QoS约束下的联合资源分配问题。为此，本项目对EH蜂窝网络中面向MEC的资源分配理论和方法进行研究，包括：充分考虑任务到达、能量收集和无线信道状态的随机变化，研究EH蜂窝网络中面向MEC的系统模型；充分考虑能量、无线带宽、计算资源和业务QoS的多维约束，研究时延、能耗、任务处理速率等多目标优化模型；针对系统状态确知、随机但统计分布已知（或可预测）和随机且统计特性不可预测三种情况，利用凸优化、随机规划、马尔科夫决策过程和强化学习等方法，研究优化多目标的多维资源联合分配策略。上述研究有望在基础理论和关键算法上取得创新研究成果，为新一代移动通信的发展提供理论参考。

Integration of Energy-Harvesting (EH) and Mobile Edge Computing (MEC) can effectively support computation-intensive and delay-sensitive mobile applications (such as augmented reality, virtual reality and smart factory). In the meanwhile, it is a promising technique for constructing an energy-saving and green autonomous mobile communications system. Traditional EH and MEC technologies cannot adapt to the random changes of task arrival, harvested energy and wireless channel, and do not solve the resources allocation for optimization problem under multi-dimension constraints in EH and MEC mobile cellular networks. To address the above mentioned problems, this project will conduct research on the resources allocation method for EH and MEC mobile cellular networks, which include: 1) Building the system models that take the random features of task arrival, harvested energy and wireless channel into consideration; 2) Under multiple constraints of energy, computation and wireless bandwidth resources and applications’ quality of service, formulating the optimization problem for optimizing the delay, energy consumption, and rate of tasks processing; 3) Taking deterministic and random system states into consideration, the effective resource allocation methods will be proposed based on convex optimization, multi-stage stochastic programming, Markov decision processing and reinforcement learning. The research of this project is expected to produce new theoretic and technical results, which provides theoretic foundation for the evolution of future mobile communication system.