关于车位分配的智能体强化学习算法探讨

INVESTIGATION ON AGENT REINFORCEMENT LEARNING ALGORITHM FOR PARKING ALLOCATION

停车难问题主要体现在停车资源时空分配不合理——现有的停车场管理中缺乏动态协调不同停车需求与停车资源的匹配机制,未能合理利用停车设施的时空特性,导致停车资源的浪费。在此背景下,通过将停车资源分配问题转换为马尔科夫决策过程,提出一种多智能体深度强化学习算法来对停车场内的差异化停车资源进行动态化分配,并通过算例证明其效果。

The difficulty of parking is mainly reflected in the unreasonable space-time distribution of parking resources-the existing parking lot management lacks the matching mechanism of dynamic coordination between different parking needs and parking resources. In this context, by transforming the parking resource allocation problem into a Markov decision-making process, this paper proposes a multi-agent deep reinforcement learning algorithm to dynamically allocate the differentiated parking resources in the parking lot.