基于深度增强学习的无人机赋能雾无线电接入网络的能效优化  

Deep reinforcement learning to enhance the energy-efficient performance of UAV-enabled F-RAN

作　　者：梅海波[1] 杨鲲[1] 范新宇[1] MEI Haibo;YANG Kun;FAN Xinyu(University of Electronic Science and Technology of China,Chengdu 610054,China)

机构地区：[1]电子科技大学,四川成都610054

出　　处：《物联网学报》2021年第2期48-59,共12页Chinese Journal on Internet of Things

基　　金：国家自然科学基金资助项目(No.61620106011,No.U1705263,No.61871076)。

摘　　要：雾无线电接入网络适合用于广域范围内的诸如管线管网监测等国家重要行业的物联网应用场景。然而基于地面雾接入节点的网络将受到环境、地形等影响,无法及时有效地提供雾接入服务。利用低空无人机作为雾接入点实现空地的边缘通信和雾计算方面引起了普遍的关注。本文探讨怎样利用深度增强学习来提高无人机雾接入点的能效,延长无人机的任务时间。深度增强学习可以保障无人机雾接入点及时地调整空地通信和计算的配置策略,包括资源优化、动态任务卸载以及缓存,也可以优化无人机在三维空间中的飞行航迹,提高无人机赋能的雾无线电接入网络的总体性能。研究的创新性在于综合论述了深度增强学习用于无人机赋能的雾无线电接入网络要解决的主要优化问题,并且总结了解决相关优化问题的技术细节,最后对深度增强学习应用于无人机赋能的雾无线电接入网络的技术挑战和未来研究方向展开讨论。Fog radio access network(F-RAN)is suitable for Internet of things applications of national important industries,such as pipeline network monitoring in wide area.However,the performance of the F-RAN based on the territorial fog access point will be affected greatly by the complicated territorial environment.This causes F-RAN not able to provide fog access service in a timely and effectively manner.To this problem,the research was proposed to utilize low altitude UAV as the fog access point to realize air ground edge communication and fog computing,which has attracted enormous research interests.How to use deep reinforcement learning(DRL)to improve the energy efficiency of UAV fog access point and extend the mission time of UAV were discussed.Deep reinforcement learning can ensure the UAV fog access point to adjust the configuration strategy timely of air ground communication and computing,including resource optimization,dynamic task offloading and caching.DRL can also optimize the UAV trajectory in 3-D space,and improve the overall performance of UAV enabled fog access network.The innovation of the research lies in the comprehensive discussion of the main optimization problems to be solved in the UAV-enabled F-RAN using DRL.The technical details were also summarized to solve the related optimization problems.Finally,the technical challenges and future research directions of the application of DRL in the UAV-enabled F-RAN were discussed.

关 键 词：无人机 雾无线电接入网络 深度增强学习 航迹规划 网络配置 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]