面向超密移动边缘计算的协作式任务卸载研究

The deployment of ultra-dense heterogeneous base stations provides mobile edge computing (MEC) with more flexible access service, while posing challenges on the energy consumption of the mobile devices (MDs), the computation tasks’ processing delay, network resource allocation, quality-of-experience (QoE), etc. In order to integrate the superiorities of both technologies, new computation offloading strategies are required. Note that available computation offloading research only focused on single-tier base station or simple computation task scenarios. Little works can be found on computation offloading for mobile-edge computing in ultra-dense networks (UDN), especially the works on the description and modeling of the important factors such as the stochastic generation of multiple computation tasks, dynamic changing of the available resources at the edge servers, channel interference, etc. Toward this end, this project is intended to describe and model the above factors based on the characteristics of MEC in UDN, where he mathematical tools like the queuing theory, stochastic geometry, etc., may be adopted. The formal definition for the energy consumption optimization problem of multi-user multi-task collaborative computation offloading will be presented, and collaborative task offloading schemes with given task processing delay will be proposed as our solution. After that, considering the influence of dynamic CPU frequency and transmit power scaling on the MDs’ energy consumption, optimal schemes with minimum energy consumption will be further presented by jointly adjusting the MDs’ CPU frequency and transmit power dynamically and conducting computation offloading. The successful implementation of this project, can not only be able to establish the fundamental basis for the future fusion of UDN and MEC, but also provide both theoretical and technical support for the application of UDN and MEC in mobile Internet and Internet-of-Things.

超密异构基站的部署为移动边缘计算提供更灵活终端接入的同时，也在终端能耗、计算任务执行时延、网络资源分配、用户QoE保障等方面提出了挑战。如何将两者优势充分结合，需要设计全新的计算任务卸载策略。现有研究多考虑单层基站或简单计算任务场景，缺乏对多计算任务随机产生、边缘服务器可用资源动态变化、无线信道干扰等重要因素的刻画；超密网中协作式计算任务卸载研究亦鲜有涉及。鉴于此，本项目拟针对超密移动边缘计算的特性，结合排队论、随机几何等理论工具建模刻画上述要素；形式化定义超密网下多用户多计算任务卸载能耗优化问题，并设计满足计算任务执行时延约束的最优协作式卸载方案；在此基础上，进一步考虑终端CPU频率、发射功率等因素的影响，设计终端能耗最优的频率、功率调节及计算任务卸载方案。本项目的成功实施，可为未来超密网和移动边缘计算技术融合打下必要的理论基础，并为其在移动互联网、物联网等的广泛应用提供理论和技术支持。

超密异构基站的部署为移动边缘计算提供更灵活终端接入的同时，也在终端能耗、计算任务执行时延、网络资源分配、用户QoE保障等方面提出了挑战。如何将两者优势充分结合，需要设计全新的计算任务卸载策略。现有研究多考虑单层基站或简单计算任务场景，缺乏对多计算任务随机产生、边缘服务器可用资源动态变化、无线信道干扰等重要因素的刻画；超密网中协作式计算任务卸载研究亦鲜有涉及。本项目针对超密异构蜂窝网络、车联网和无人机辅助通信等场景，建立超密移动边缘计算任务卸载模型，形式化定义多用户多计算任务卸载能耗优化问题，结合博弈论、组合优化、深度强化学习等方法求解，提出了相应的任务卸载及资源优化分配方案，解决了超密边缘计算、车联边缘计算、无人机边缘计算中的智能任务卸载、资源优化分配、负载均衡等问题，突破了超密物联网络环境带能量感知的计算任务卸载及传输功率优化、车联边缘计算网络计算任务卸载、无人机辅助边缘计算任务卸载及资源优化分配、超密边缘计算网络环境资源优化配置及分配、边缘计算网络中电动汽车智能弹性充电策略、车联边缘计算网络任务卸载负载均衡、边缘智能任务卸载等关键技术。本项目的实施，为超密网和移动边缘计算技术融合打下必要的理论基础，并为其在移动互联网、物联网等的广泛应用提供理论和技术支持。