The invention relates to the field of new energy automobile braking, in particular to an AEB control method and system based on DDPG deep reinforcement learning, and the method comprises the steps: constructing an action-evaluation network based on a strategy network and a value network, selecting an action in a current driving state according to the strategy network, and evaluating the selected action through the value network; constructing a target network based on a target strategy network and a target value network, and randomly obtaining training samples from sample values by using an experience playback mechanism to perform offline training and updating on the strategy network and the target value network; the target network updates the strategy network according to the evaluation value of the state after the action selected by the action-evaluation network is executed in the current driving state and a reward function of the system; inputting a current driving state, and selecting an optimal action by constructing an action-evaluation network based on a strategy network and a value network; by the adoption of the optimized braking strategy, the economic benefits of the new energy automobile can be improved, invalid oil consumption is reduced, and the aging speed of automobile parts is slowed down.

    本发明涉及新能源汽车制动领域,具体涉及一种基于DDPG深度强化学习的AEB控制方法及系统,包括构建基于策略网络和值网络构建动作‑评价网络,根据策略网络选择在当前行驶状态下的动作,值网络对选择的动作进行评价;构建基于目标策略网络和目标值网络的目标网络,使用经验回放机制从样本值中随机获取训练样本对策略网络和值网络进行离线训练和更新;目标网络根据在当前行驶状态下执行动作‑评价网络选择的动作后状态的评价值和系统的奖励函数更新策略网络;将当前行驶状态输入,通过基于策略网络和值网络构建动作‑评价网络选择最优的动作;本发明采用优化后的制动策略能够提高新能源汽车的经济效益,降低无效油耗,减缓车辆零件老化速度。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    AEB control method and system based on DDPG deep reinforcement learning


    Weitere Titelangaben:

    一种基于DDPG深度强化学习的AEB控制方法及系统


    Beteiligte:
    SHI JUNREN (Autor:in) / LI KEXIN (Autor:in) / PARK CHANG HO (Autor:in) / MA YIWEI (Autor:in) / LIU MINGJIE (Autor:in) / CHEN JUNSHENG (Autor:in) / HUANG JIAN (Autor:in) / SU YONGKANG (Autor:in)

    Erscheinungsdatum :

    2022-09-30


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    B60T Bremsanlagen für Fahrzeuge oder Teile davon , VEHICLE BRAKE CONTROL SYSTEMS OR PARTS THEREOF / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    A Control Method for Quadrotor Based on DDPG

    Chen, Shuo / Xiao, Jin / Hu, Cun Yi | TIBKAT | 2022


    A Control Method for Quadrotor Based on DDPG

    Chen, Shuo / Xiao, Jin / Hu, Cun Yi | Springer Verlag | 2021


    Autopilot Strategy Based on Improved DDPG Algorithm

    Zuo, Xiaochao / Li, Xiaoning / Tian, Zhewen | SAE Technical Papers | 2019


    Deep Deterministic Policy Gradient (DDPG) Agent-Based Sliding Mode Control for Quadrotor Attitudes

    Wenjun Hu / Yueneng Yang / Zhiyang Liu | DOAJ | 2024

    Freier Zugriff

    An Intelligent Attitude Control Method for UAV Based on DDPG Algorithm

    Xian, Yuewei / Wang, Peng / Xin, Hongbo et al. | IEEE | 2023