Computational offloading and resource allocation algorithm based on deep reinforcement learning in Internet of Vehicles
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
维普
万方数据
针对车辆边缘计算(Vehicular Edge Computing,VEC)卸载与资源分配过程中由于边缘服务器资源受限导致的时延增大的问题,提出一种基于深度强化学习的计算卸载与资源分配(Compute Offload and Resource Allocation Based on Deep Q-Netwrk,CORADQN)算法.构建VEC网络架构,通过拆分计算密集型车载任务及利用空闲服务车辆的计算资源,将计算任务分别卸载至边缘服务器、空闲服务车辆和本地车辆进行处理,以降低VEC网络系统的总时延.将计算卸载与资源分配转化为多约束优化问题,并将平均奖励作为样本的优先级进行采样,从而提高样本的利用率,加快算法收敛速度.仿真结果表明,相较于完全本地(ALL-Local)算法、完全边缘(ALL-Edge)算法、联邦卸载(Federated Offloading Scheme,FOS)算法及深度Q学习(Deep Q-Network,DQN)算法,所提算法能够最小化VEC网络的系统时延.
In order to solve the problems of limited resources and high latency of edge servers in the process of computation offloading and resource allocation of vehicular edge computing(VEC),a computation offloading and resource allocation algorithm based on deep reinforcement learning is proposed.The VEC network architecture is constructed to reduce the total latency of the VEC net-work system by splitting the computation-intensive on-board tasks and using the computing re-sources of the idle service vehicles,the computing tasks are offloaded to the edge servers,the idle service vehicles,and the local vehicles separately for processing.The computational offloading and resource allocation are transformed into multi-constraint optimization problems,and the average re-ward is sampled as the priority of the samples,so as to improve the utilization rate of the samples and accelerate the convergence speed of the algorithm.Simulation results show that compared with the ALL-Local algorithm,the ALL-Edge algorithm,the FOS algorithm,and the DQN algorithm,the proposed algorithm can minimize the system delay of the VEC network.
Internet of Vehiclesmobile edge computingdeep Q-learningcomputation offloadingnetwork latency
维普
万方数据
