毫米波MEC系统资源管理技术

Mobile edge computing (MEC) technique with distributed cloudlike architecture, which allows nearby infrastructure to provide mobile users low delay, low energy consumption and high performance computing services, is a new frontier of mobile communications. In order to overcome the drawback that limited resources are not efficiently managed to guarantee the requirements of communication and computing tasks, this project designs millimeter wave (mmW) communication based MEC technique, namely mmW MEC, and studies exploits the resource management in mmW MEC systems in following three aspects. 1. Delay minimization for dense computing task offloading. In particular, based on leveraging priority weighted marking strategy, the M2O-ME matching algorithm is proposed for offloading decision making, so that the total system delay is minimized. 2. Communication and computing double-motivated resource allocation. Specifically, with the aid of converting the multi-objective problem by using the dimension-reducing and fitting approach, the POA monotonic optimization algorithm is developed to find the optimal solutions to the communication- and computing-oriented low-complexity resource allocation strategies. 3. Joint blocking and interference management in dynamic systems. Aiming to maximize the system reliability, the joint blocking and interference management scheme is designed, in which the small-cell base station cooperation is used for blocking management, while potential game method based spectrum orthogonality algorithm is leveraged to mitigate interference. Aforementioned research issues in this project provide a basis to realize low-delay, low-complexity and high reliable resource management strategies for mmW MEC systems.

移动边缘计算（MEC）技术采用分布式云化架构，在邻近用户的基础设施，为用户提供低时延、低能耗的高性能计算服务，是移动通信的新方向。本项目设计基于毫米波通信的MEC技术，针对有限通信资源难以配置满足计算与通信业务需求问题，从三方面开展毫米波MEC资源管理研究。1. 密集计算业务单驱快速卸载。利用加权优先级考订方法进行计算卸载判决，通过M2O-ME匹配论方法设计卸载传输调度策略，以获得计算时延最小化。2. 通信与计算业务双驱资源分配。基于多维域降维拟合方法转化多目标问题，利用单调优化理论设计POA资源分配算法，实现通信与计算业务并存低复杂度资源优化配置。3. 动态环境遮挡与干扰联合管理。利用微基站协作策略实现遮挡中断链路复建，基于势博弈方法设计频域扰动正交干扰抑制算法，以高概率恢复中断或受损链路，保障通信可靠性。本项目的研究对于实现低时延、低复杂度及高可靠的毫米波MEC资源管理技术有支撑意义。

面向5G/B5G移动网络多元业务爆发式增长需求，移动边缘计算（MEC）技术受到学界和工业界的广泛关注，其采用分布式云化架构，在邻近用户的基础设施，为用户提供低时延、低能耗的高性能计算服务，是移动网络边端服务的重要组成。为了解决频谱资源日趋紧缺问题，本项目将毫米波通信技术引入MEC卸载传输，从如下几方面开展了MEC卸载传输资源管理研究。首先，研究了密集计算业务单驱快速卸载技术。利用加权优先级考订方法进行计算卸载判决，通过匹配论方法设计卸载传输调度策略，以获得计算时延最小化。其次，研究了通信与计算业务双驱资源分配。基于多维域降维拟合方法转化多目标问题，利用单调优化理论设计多维资源分配算法，实现通信与计算业务并存低复杂度资源优化配置。再者，研究了动态环境遮挡与干扰联合管理。利用微基站协作策略及多波束选择实现遮挡中断链路复建，以高概率恢复中断或受损链路，保障通信可靠性。最后，额外研究了计算与缓存业务并存下的无线资源联合管理技术。利用计算卸载与缓存传输的链路对称优势，化解计算与缓存调度矛盾，提出了全新的基于包络更新的多目标深度强化学习算法，给出了系统时延与能耗的最佳权衡。本项目的研究对于实现低时延、低复杂度及高可靠的毫米波MEC资源管理技术有支撑意义。