The invention discloses a multi-agent Actor-Critic traffic signal cooperative control method based on centralized training and decentralized execution, and relates to the field of artificial intelligence. In an urban traffic network, a distributed Actor network and a centralized Critic network are distributed to each intersection agent. In the centralized training stage, information sharing is achieved by sending respective local traffic states, local rewards and other information between the intelligent agents, and perception of the global traffic state is formed. The centralized Critic network evaluates the control strategy effect of the distributed Actor network according to the aggregated global traffic state and reward information, and guides the Actor network to learn how to maximize the overall performance of the traffic network through shared information and collaborative decision in a dynamic time-varying traffic environment; in the decentralized execution stage, the distributed Actor network selects a control action according to local observation of the distributed Actor network and a trained strategy. According to the invention, the intelligent cooperation level and the adaptive decision-making capability of multi-intersection traffic signal control are effectively improved.

    一种基于集中训练、分散执行的多智能体Actor‑Critic交通信号协同控制方法,涉及人工智能领域。在城市交通网络中,每个交叉口智能体分配一个分布式Actor网络和一个集中式Critic网络。在集中训练阶段,智能体之间通过发送各自的局部交通状态、局部奖赏等信息,实现信息共享,形成对全局交通状态的感知。集中式Critic网络根据聚合的全局交通状态和奖赏信息,评估分布式Actor网络的控制策略效果,并指导Actor网络学习如何在动态时变的交通环境中通过共享信息和协同决策最大化交通网络的整体性能;在分散执行阶段,分布式Actor网络根据自己的局部观察和训练好的策略选择控制动作。本发明有效提升了多交叉口交通信号控制的智能合作水平和自适应决策能力。


    Access

    Download


    Export, share and cite



    Title :

    Multi-agent Actor-Critic traffic signal cooperative control method based on centralized training and decentralized execution


    Additional title:

    一种基于集中训练、分散执行的多智能体Actor-Critic交通信号协同控制方法


    Contributors:
    ZHANG YONGNAN (author) / PENG SIRUI (author) / LI YUNXUAN (author) / GAN SHAOJUN (author) / CHEN YANYAN (author)

    Publication date :

    2024-10-11


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Agent-Agnostic Centralized Training for Decentralized Multi-Agent Cooperative Driving

    Yan, Shengchao / König, Lukas / Burgard, Wolfram | ArXiv | 2024

    Free access


    Swarm Cooperative Navigation Using Centralized Training and Decentralized Execution

    Rana Azzam / Igor Boiko / Yahya Zweiri | DOAJ | 2023

    Free access


    Multi-agent deep reinforcement learning with actor-attention-critic for traffic light control

    Wang, Bin / He, ZhengKun / Sheng, JinFang et al. | SAGE Publications | 2024