The invention provides a vehicle model-free stability control method based on deep reinforcement learning, and the method employs an independently driven electric vehicle as a research object, employs a convolutional neural network to carry out the processing of a pavement image collected by a camera, recognizes the pavement state, and serves as the basis of vehicle stability control. The method comprises the following steps: establishing a vehicle ideal stability state reference model, enabling stability control of a vehicle to be equivalent to a Markov decision process, constructing a reinforcement learning environment, including designing a state space, an action space and a reward function, designing an Actor-Critic network structure, updating network parameters by using a depth deterministic strategy gradient and a value loss function, and performing offline learning. And finally, on-line adjustment is carried out on the vehicle attitude based on the stability control Agent. According to the method, the road surface image information is collected and processed in real time, the road surface type can be judged without establishing a state equation and an observation equation, the influence of process noise and observation noise on an estimation result is avoided, the road identification accuracy is improved, a stability control strategy can be dynamically adjusted according to road conditions, and the road identification accuracy is improved. And the working condition adaptability of the control strategy is improved.

    本发明提供了一种基于深度强化学习的车辆无模型稳定性控制方法,该方法以独立驱动电动汽车为研究对象,利用卷积神经网络对摄像头采集到的路面图像进行处理,辨识路面状态,作为车辆稳定性控制的依据。建立车辆理想稳定性状态参考模型,将车辆的稳定性控制等效为马尔科夫决策过程,构建强化学习的环境,包括设计状态空间、动作空间与奖励函数,设计Actor‑Critic网络结构,并利用深度确定性策略梯度和价值损失函数更新网络参数,进行离线学习,最后基于稳定性控制Agent对车辆姿态进行在线调节。本发明通过对路面图像信息进行实时采集和处理,无需建立状态方程和观测方程即可判断路面类型,避免了过程噪声和观测噪声对估计结果的影响,提高了道路辨识的精确度,并可根据道路条件动态调整稳定性控制策略,提高控制策略的工况适应性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Electric vehicle model-free stability control method based on deep reinforcement learning


    Weitere Titelangaben:

    基于深度强化学习的电动汽车无模型稳定性控制方法


    Beteiligte:
    HE YILIN (Autor:in) / ZHAO XUAN (Autor:in) / MA JIAN (Autor:in) / CHEN YISONG (Autor:in) / ZHENG WEI (Autor:in) / ZHANG JIWEI (Autor:in) / DU KECHUAN (Autor:in) / WANG DONGYANG (Autor:in)

    Erscheinungsdatum :

    29.10.2024


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    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



    Battery Electric Vehicle Control Strategy for String Stability Based on Deep Reinforcement Learning in V2V Driving

    Borneo, Angelo / Miretti, Federico / Acquarone, Matteo et al. | British Library Conference Proceedings | 2023


    Electric vehicle longitudinal control method based on deep reinforcement learning and sliding mode control

    WANG ZHENWEI / DING JINQUAN / HOU JUNJIAN et al. | Europäisches Patentamt | 2025

    Freier Zugriff


    Vehicle following speed control method based on deep reinforcement learning

    FEI RONG / YANG LU / QIU YUAN et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Cluster electric vehicle charging behavior optimization method based on deep reinforcement learning

    HU JUNJIE / ZHAO XINGYU | Europäisches Patentamt | 2020

    Freier Zugriff