The invention provides an urban traffic scene-oriented intelligent automobile reinforcement learning cooperative driving method, which comprises the following steps that: an intelligent agent interacts with a traffic environment to obtain observation information, action information and adjacent vehicle information; the observation information, the action information and the adjacent vehicle information are stored in an experience pool; sampling the data in the experience pool to obtain sample data; inputting the sample data into a value network and a strategy network, adding feature splicing in the value network, and adding an attention mechanism, a coding mechanism and feature splicing in the strategy network; and updating parameters of the strategy network based on the value function and the action value, and generating an automatic driving strategy. According to the method, features of different scales are spliced and introduced into the strategy network and the value network, feature reuse is promoted, the network representation capability is improved, an attention mechanism is used in the strategy network, effective extraction of features of a complex scene is enhanced, and the applicability of different scenes and the stability and accuracy of decision making are improved.

    本发明提供了一种面向城市交通场景的智能汽车强化学习协作驾驶方法,包括:智能体与交通环境进行交互,得到观测信息、动作信息以及相邻车辆信息;将观测信息、动作信息以及相邻车辆信息储存至经验池;对经验池中的数据进行采样,得到样本数据;将样本数据输入价值网络和策略网络,在价值网络中增加特征拼接,在策略网络中增加注意力机制、编码机制以及特征拼接;基于价值函数及动作值对策略网络的参数进行更新,生成自动驾驶策略。本发明的方法将不同尺度的特征拼接引入策略网络和价值网络,促进特征重用及提升网络表征能力,并在策略网络中使用注意力机制,增强对复杂场景的特征的有效提取,提高了不同场景的适用性,以及决策的稳定性和准确性。


    Access

    Download


    Export, share and cite



    Title :

    Intelligent automobile reinforcement learning cooperative driving method for urban traffic scene


    Additional title:

    面向城市交通场景的智能汽车强化学习协作驾驶方法


    Contributors:
    LIU JIA (author) / LIANG QINGYI (author) / YIN JIANWEN (author) / JIANG ZHENGMIN (author) / XU KUN (author) / LI HUIYUN (author)

    Publication date :

    2024-01-09


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G06Q Datenverarbeitungssysteme oder -verfahren, besonders angepasst an verwaltungstechnische, geschäftliche, finanzielle oder betriebswirtschaftliche Zwecke, sowie an geschäftsbezogene Überwachungs- oder Voraussagezwecke , DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES



    Decision planning method for automatic driving vehicle in urban traffic scene based on reinforcement learning

    CHEN XIYUAN / LIU WEIYAN / FANG LIN et al. | European Patent Office | 2024

    Free access

    Intelligent automobile traffic scene mapping method considering dynamic target

    DONG HONGHUI / LIU TONGTONG / ZHOU YUE | European Patent Office | 2024

    Free access

    Intelligent automobile traffic scene-oriented objective marginal quantitative evaluation method

    ZHU BING / LI WENXU / ZHAO JIAN et al. | European Patent Office | 2024

    Free access


    Scene extraction method based on intelligent automobile traffic situation risk cognition

    ZHU BING / XUE JINGWEI / ZHAO JIAN et al. | European Patent Office | 2024

    Free access