The invention discloses a multi-vehicle joint longitudinal control method based on an improved MADDPG, and the method comprises the steps: S1, carrying out the modeling of traffic walking and stopping waves through employing simulation software SUMO, carrying out the modeling of each vehicle, and designing a single-lane driving route of the vehicle; s2, using a traffic control interface Traci built in the simulation software SUMO to interact with Python for realizing deep reinforcement learning; s3, calibrating and verifying the simulation model; s4, designing a joint state space, an action space and a joint reward of the multiple agents according to the multi-vehicle longitudinal control task; constructing an SRM-MADDPG algorithm to improve a state communication mode and a reward communication mode among multiple agents; and S5, setting a strategy objective function and a value function. According to the method, the cooperation among multiple agents can be enhanced, the calculation complexity is reduced, the performance of a multi-agent system is improved, the traffic safety is effectively optimized, and the traffic wave walking and stopping phenomena are inhibited.

    本发明公开了一种基于改进MADDPG的多车联合纵向控制方法,包括步骤:S1,采用仿真软件SUMO对交通走停波进行建模,对每辆车进行建模并设计车辆的单车道行进路线;S2,运用仿真软件SUMO内置的交通控制接口Traci与实现深度强化学习的Python进行交互;S3,对仿真模型进行标定与验证;S4,根据多车纵向控制任务,对多智能体的联合状态空间、动作空间以及联合奖励进行设计;构建SRM‑MADDPG算法改进多智能体间的状态通信方式和奖励通信方式;S5,设置策略目标函数和值函数。本发明能增强多智能体之间的合作,降低计算复杂度,提高多智能体系统的性能,有效优化交通安全并抑制交通走停波现象的发生。


    Access

    Download


    Export, share and cite



    Title :

    Multi-vehicle joint longitudinal control method based on improved MADDPG


    Additional title:

    基于改进MADDPG的多车联合纵向控制方法


    Contributors:
    LI ZHIBIN (author) / LI GEYANG (author)

    Publication date :

    2023-10-24


    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 / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Research on autonomous formation of Multi-UAV based on MADDPG algorithm

    Zhang, Yaozhong / Wu, Zhuoran / Ma, Yunhong et al. | IEEE | 2022



    Pulse type track pursuit game method based on PRD-MADDPG algorithm

    ZHAO LIRAN / DANG CHAOHUI / TANG SHENGYONG et al. | European Patent Office | 2022

    Free access