具有URLLC任务卸载的无人机辅助移动边缘计算系统的计算时延最小化研究  

Computing Delay Minimization for UAV-enabled Mobile Edge Computing Systems with URLLC-based Offloading

作　　者：易雅倩 吴庆捷 崔苗[1] 张广驰[1] Yi Ya-qian;Wu Qing-jie;Cui Miao;Zhang Guang-chi(School of Information Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学信息工程学院,广东广州510006

出　　处：《广东工业大学学报》2024年第4期70-79,共10页Journal of Guangdong University of Technology

基　　金：广东省基础与应用基础研究基金资助项目(2023A1515011980);广东省科技计划项目(2022A0505050023,2022A0505020008,2023A0505050127);广东省海洋经济发展项目(粤自然资合[2023]24号);广东省特支计划项目(2019TQ05X409);江西省军民融合北斗通航重点实验室开放基金资助项目(2022JXRH0004)。

摘　　要：无人机辅助移动边缘计算(Mobile Edge Computing,MEC)系统利用无人机部署灵活和能与地面设备建立起高质量通信链路的优势,可以提高边缘计算的性能。已有的无人机辅助MEC系统研究通常假设计算任务卸载通信的数据包较长,其结果无法直接用于对计算时延具有严格要求的MEC应用场景。本文考虑了一个具有超可靠低时延通信(Ultra-Reliable and Low-Latency Communication,URLLC)任务卸载的无人机辅助MEC系统,其中一架配备计算服务器的无人机为多个地面设备提供MEC服务,地面设备将部分计算任务通过URLLC的方式向无人机卸载。通过联合优化无人机的部署位置、卸载通信的带宽和无人机与地面设备的中央处理器(Central Processing Unit,CPU)计算频率来最小化系统的计算时延。针对形成的非凸优化问题,采用块坐标下降方法将该问题分解为无人机部署位置优化子问题和任务卸载带宽与CPU频率优化子问题,并交替求解这两个子问题直到收敛。在求解两个子问题的过程中,使用对数函数非线性近似方法对URLLC卸载通信速率表达式进行化简,并通过连续凸逼近方法将每个子问题转换成易于求解的凸优化问题。仿真结果表明,与其他基准算法相比,所提算法能够有效平衡系统的通信能力和计算能力,从而降低系统时延。Unmanned aerial vehicle(UAV)-enabled mobile edge computing(MEC)systems can improve their edge computing performance by taking the advantage of flexible deployment of UAVs and the ability to establish high-quality communication links between UAVs and ground users.The existing works on UAV-enabled MEC systems usually assume that the blocklength of the offloading transmission is long,and their results cannot be directly applied to MEC scenarios with strict requirements on computing delay.A UAV-enabled MEC system with ultra-reliable and low-latency communication(URLLC)-based task offloading is considered,in which a UAV carrying a computing server provides MEC service for multiple ground users and the users offload parts of their computing tasks to the UAV through URLLC.The UAV's deployment location,the offloading bandwidths of the users,and the computing central processing unit(CPU)frequencies of the UAV and users are jointly optimized to minimize the computation delay of the system.To solve the resulting non-convex optimization problem,the block coordinate descent method is used to decompose the problem into two subproblems that optimize the UAV's deployment location,and the offloading bandwidths and CPU frequencies,respectively,and the two subproblems are solved alternately.In solving the two subproblems,logarithmic functions are used to make nonlinear approximation to the expression of URLLC offloading rate to simplify the subproblems,and the two subproblems are both transformed into convex optimization problems by applying the successive convex approximation method.Simulation results show that the proposed algorithm can effectively balance the communication capability and computing capability of the system and reduce the system’s computing delay compared to other benchmark schemes.

关 键 词：无人机 移动边缘计算 超可靠低时延通信 计算时延 

分 类 号：TN929.5[电子电信—通信与信息系统]