基于深度强化学习的V2V通信资源分配

V2V Communication Resource Allocation Based on Deep Reinforcement Learning

在集中式的资源管理结构中,基站作为决策节点收集数据的时延较高,为此提出了基于深度强化学习的联合信道和功率等级分配的分布式自主方案。首先确定车对基础设施链路分配方案,然后根据车与车链路与环境交互情况选择信道和传输功率,计算奖励函数,更新Q网络并选择策略。结果表明,车与车链路满足链路时延约束和容量要求,同时也降低了对车与基础设施链路的干扰,提升了通信性能。

In a centralized resource management structure,the base station can be used as a decision node to collect data but has a high delay.Therefore,a distributed autonomous scheme of joint channel and power level allocation based on deep reinforcement learning was proposed.First,the vehicle-to-infrastructure(V2I)resource allocation scheme was determined,and the channel and transmission power were then selected based on the vehicle-to-vehicle(V2V)link interaction with the environment.The reward function was calculated,and the Q network was updated,with a strategy selected.The experimental results show that the V2V link can satisfy the delay constraint and capacity of the link,and it minimizes interference to the V2I link and improves communication performance.