Resource Allocation for UAV-aided Communication in High-Speed Railway Systems: A Multi-Agent Reinforcement Learning Approach
首发时间:2021-03-05
Abstract:The past decades have witnessed the rapid developments of high-speed railways (HSRs) communications. To provide seamless communication services between high-speed trains, both mobile edge computing (MEC) servers and UAVs are integrated into HSRs to provide on-demand resource access. However, the sensitive delay requirements of high-speed services pose significant challenges to the resource allocation in HSRs. This paper will formulate the UAV-aided resource allocation in high-speed railways (HSRs) as a distributed optimization problem to optimize the resource utilization while minimizing the path blocking probability. To address this problem, a multi-agent deep deterministic policy gradient (multi-agent DDPG) approach is proposed. The MEC servers are taken as the agents to make resource allocation decisions in the training phase. The simulation shows demonstrate that multi-agent DDPG outperforms the traditional single-agent method. The proposed multi-agent DDPG-based resource allocation algorithm can achieve satisfactory performance.?????
keywords: High-speed railway deep reinforcement learning unmanned aerial vehicle resource allocation multi-agent DDPG.
点击查看论文中文信息
高速铁路系统中无人机辅助通信的资源分配:一种基于多智能体强化学习方法
摘要:在过去的几十年间,高速铁路(High-Speed Railway, HSR)系统得到了飞速发展。为了给高铁列车提供无缝的通信服务,移动边缘计算(Mobile Edge Computing, MEC)和无人机(Unmanned Aerial Vehicle, UAV)被集成到高铁系统中为用户提供按需的资源接入服务。然而,高铁通信对时延的超高需求给高铁系统的资源分配带来了重大挑战。本文将HSR系统中的无人机辅助资源分配问题制定为分布式优化问题,在优化资源利用的同时最小化路径阻塞概率。针对这一问题,本文提出了一种基于多智能体深度确定性策略梯度(Multi-Agent Deep Deterministic Policy Gradient, Multi-Agent DDPG)的资源分配方法。在训练阶段,将MEC服务器和UAVs作为代理进行资源分配决策。仿真结果表明,多智能体DDPG的学习性能优于传统的单智能体方法,并且资源分配的仿真实验中取得了优越的性能。
基金:
引用
No.****
同行评议
共计0人参与
勘误表
高速铁路系统中无人机辅助通信的资源分配:一种基于多智能体强化学习方法
评论
全部评论0/1000