面向停车合作基于深度强化学习的车辆任务卸载  

Vehicle Task Offloading Based on Deep Reinforcement Learning for Parked Vehicle Cooperation

作　　者：王振川 花季伟[1] 朱金奇[1] 孙麒惠 郑敏 李云龙 WANG Zhen-chuan;HUA Ji-wei;ZHU Jin-qi;SUN Qi-hui;ZHENG Min;LI Yun-long(School of Computer and Information Engineering,Tianjin Normal University,Tianjin 300387,China;School of Intelligent Manufacturing,Tianjin Sino-German University of Applied Sciences,Tianjin 300350,China)

机构地区：[1]天津师范大学计算机与信息工程学院,天津300387 [2]天津中德应用技术大学智能制造学院,天津300350

出　　处：《小型微型计算机系统》2023年第3期658-664,共7页Journal of Chinese Computer Systems

基　　金：天津市自然科学基金项目(18JCYBJC85900,18JCQNJC70200)资助;国家自然科学基金项目(61902282,62002263)资助;天津市企业科技特派员项目(20YDTPJC0062)资助。

摘　　要：针对路边基础设施受损或失效情况下卸载任务无法被执行的状况,提出令拥有丰富计算资源的路边停放车辆彼此合作,执行车联网中移动车辆产生的计算密集型任务.在把一条道路的路边停放车辆组织成停车簇后,首先分析各卸载任务所需最佳资源量,接着提出基于深度强化学习(Deep Reinforcement Learning, DRL)的计算任务分块卸载算法,将任务划分为多个子任务后由多个停放车辆并行执行,以最小化任务执行延迟和执行任务能耗开销构成的总成本.大量仿真结果表明,本文所提算法的任务执行完成率大大高于其他对比算法,且具有最低的任务执行成本开销.For the situation that the offloaded task cannot be performed when the roadside infrastructure is damaged or fails, roadside parked vehicles with rich computational resources cooperate with each other to perform the compute-incentive tasks generated by mobile vehicles in Internet ofVehicles(IoVs) is proposed. After organizing the roadside parked vehicles on a road into a parking cluster, firstly, the optimal amount of resources required for each unloading task is analyzed, and then a task uploading algorithm for computing tasks based on deep reinforcement learning(DRL) is proposed. Eachtasksis divided into multiple sub-tasks and executed by multiple parked vehicles in parallel, to minimize the total cost that is composed of the task execution delay and the energy consumption overhead forexecuting the task. A large number of simulation results show that the task execution successful rate of the proposed algorithm is not only much higher than other comparison algorithms, but also has the lowest task execution cost.

关 键 词：边缘计算 停放车辆 任务卸载 深度强化学习 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]