摘要
在能源互联网中引入无人机进行电力线路巡查,并借助移动边缘计算技术实现巡检任务的接入和处理,可降低服务成本,提高工作效率。但是,由于无人机数据传输需求和地理位置的动态变化,易造成边缘服务器负载不均衡,致使巡检业务处理时延和网络能耗较高。为解决以上问题,提出基于深度强化学习的能源互联网智能巡检任务分配机制。首先,综合考虑无人机和边缘节点的运动轨迹、业务差异化的服务需求、边缘节点有限的服务能力等,建立面向时延、能耗等多目标联合优化的双层边缘网络任务卸载模型。进而,基于Lyapunov优化理论和双时间尺度机制,采用近端策略优化的深度强化学习算法,对固定边缘汇聚层和移动边缘接入层边缘节点间的连接关系和卸载策略进行求解。仿真结果表明,所提机制能够在保证系统稳定的情况下降低服务时延和系统能耗。
In order to reduce the cost and improve efficiency of power line inspection,UAV(unmanned aerial vehicle),which use mobile edge computing technology to access and process service data,are used to inspect power lines in the energy internet.However,due to the dynamic changes of UAV data transmission demand and geographical location,the edge server load will be unbalanced,which causes higher service processing delay and network energy consumption.Thus,an intelligent inspection task allocation mechanism for energy internet based on deep reinforcement learning was proposed.First,a two-layer edge network task offloading model was established to archive joint optimization of multi-objectives,such as delay and energy consumption.It was designed by comprehensively considering the route of UAV and edge nodes,different demands of services and limited service capabilities of edge nodes.Furthermore,based on Lyapunov optimization theory and dual-time-scaled mechanism,proximal policy optimization algorithm based deep reinforcement learning was used to solve the connection relationship and offloading strategy of edge servers between fixed edge sink layer and mobile edge access layer.The simulation results show that,the proposed mechanism can reduce the service request delay and system energy consumption while ensuring the stability of system.
作者
徐思雅
邢逸斐
郭少勇
杨超
邱雪松
孟洛明
XU Siya;XING Yifei;GUO Shaoyong;YANG Chao;QIU Xuesong;MENG Luoming(State Key Laboratory of Networking and Switching Technology,Beijing University of Posts and Telecommunications,Beijing 100876,China;Information and Communication Branch,State Grid Liaoning Electric Power Co.,Ltd.,Shenyang 110004,China)
出处
《通信学报》
EI
CSCD
北大核心
2021年第5期191-204,共14页
Journal on Communications
基金
国家重点研发计划基金资助项目(No.2019YFB2102302)
国家自然科学基金资助项目(No.61702048)
工业互联网创新发展工程基金资助项目(基于泛在电力物联网的工业互联网测试床)。
关键词
巡检无人机
任务卸载
近端策略优化
李雅普诺夫优化
人工智能
patrol UAV
task offloading
proximal policy optimization
Lyapunov optimization
artificial intelligence
