5G车联网多接入边缘计算卸载策略优化
首发时间:2021-01-29
摘要:随着5G通信技术及车联网的发展,大量车辆联网应用不断出现,车辆设备有限的计算能力已成为降低用户应用体验的一大原因。多接入边缘计算(MEC)的出现能够有效解决该问题,通过将计算任务卸载到位于网络边缘侧的MEC服务器上处理,可以加速数据计算,让用户享有不间断的高质量网络体验。为了有效降低计算时延和设备能耗,本文设计了一个车辆感知多接入边缘计算网络,综合考虑用户任务卸载需求、车辆的移动性、系统计算资源分布以及网络通信带宽等因素,提出了基于深度强化学习的联合计算卸载和任务迁移优化算法(CORA)。实验证明,相比于其他计算卸载方案,该算法在不同的车辆数、任务数据量、MEC服务器数量和计算能力下均具有最低的计算卸载成本,说明CORA算法能够有效提高系统资源利用率,优化MEC计算卸载策略。
关键词: 多接入边缘计算 车联网 5G 计算卸载 深度强化学习
For information in English, please click here
Optimization of Computation Offloading Strategy for Multi-access Edge Computing in 5G Internet of Vehicles
Abstract:With the development of 5G technologies and Internet of Vehicles (IoV), a large number of IoV applications have appeared. The lack of computing capacity of vehicle equipment has become a major limitation for reducing the application experience of users. The emergence of Multi-access Edge Computing (MEC) can effectively solve this problem. By offloading tasks to the MEC server located at the edge of the network, the data computing can be accelerated, allowing users to enjoy uninterrupted high-quality network experience. In order to effectively reduce the latency and energy consumption of tasks, we design a vehicle-aware Multi-access Edge Computing network (VAMECN), and propose a deep reinforcement learning-based joint computation offloading and resourceallocation optimization algorithm (CORA), taking into account the offloading requirements of users, vehicle mobility, computing resource distribution, and communication bandwidth, etc. Experiments have proved that compared with other computation offloading schemes, the CORA algorithm has the lowest offloading cost under different number of vehicles, data size, number of MEC servers, and computing capacity. It indicates that the CORA algorithm can effectively improve system resource utilization and optimize the offloading and resource allocation strategy.
Keywords: multi-access edge computing internet of vehicles 5G computation offloading deep reinforcement learning
引用
No.****
动态公开评议
共计0人参与
勘误表
5G车联网多接入边缘计算卸载策略优化
评论
全部评论0/1000