基于多智能体协作学习的车联网联合通信计算资源分配

Multi-agent Cooperative Learning Based Joint Communication and Computing Resource Allocation in Internet of Vehicles

边缘计算赋能下车联网固有的高移动性和计算资源实时变化性对通信与计算资源分配提出了全新要求。考虑以最小化计算迁移时延与车辆用户传输能耗之间长期权衡为优化目标,利用协作多智能体强化学习方法设计联合通信与计算资源分配方案,使得各车辆用户分别作为智能体通过共享训练经验以完成中心化训练,进而可仅依据本地观测信息即可协作完成分布式决策。经数值仿真验证,所提方案相较于基准方案能更好地适应部分观测车联网环境,且可以更有效地同时降低计算迁移时延和传输能耗。

The inherent high mobility and real-time changes of computing resources in the edge computing enabled vehicular networks have put forward new requirements for communication and computing resource allocation.Considering the long-term trade-off between the minimum computing migration delay and the transmission energy consumption of vehicular users as the optimization objective,the collaborative multi-agent reinforcement learning method is used to design a joint communication and computing resource allocation scheme.All vehicular users act as agents to complete centralized training by sharing training experience,and then the distributed decisions can be cooperatively completed based on the local observation information.Numerical simulation results have shown that compared to the benchmarks,the proposed scheme can better adapt to the partially observable vehicular communication environment and can effectively reduce both the computation migration delay and transmission energy consumption.