The invention belongs to the technical field of rail transit vehicle operation control, and particularly relates to a deep reinforcement learning method suitable for automatic parking of a high-speed train, which solves the technical problems in the background technology, and comprises the following steps: constructing a dynamic model and a braking model by analyzing the motion characteristics in the automatic parking braking process of the high-speed train; a multi-input single-output neural network structure combining a long-short-term memory network and a full-connection network is constructed to carry out deep reinforcement learning on the parking braking process of the high-speed train; and applying a deep reinforcement learning method to a simulation environment established by adopting actual line data, and calculating to obtain the parking braking control method of the high-speed train. The method is suitable for the automatic parking control process of the rail transit high-speed train. According to the method, the problem that the parking brake control parameters are difficult to optimize in an existing method is solved, the reinforcement learning environment based on the multi-vehicle multi-line model enables the application scene to be wider, and the purpose that the optimal control parameters are adjusted in real time in the braking process through the deep reinforcement learning method is achieved.
本发明属于轨道交通车辆运行控制技术领域,具体为一种适用于高速列车自动停车的深度强化学习方法,解决了背景技术中的技术问题,其包括通过分析高速列车自动停车制动过程中的运动特点,构建动力学模型和制动模型;构建一种结合长短期记忆网络与全连接网络的多输入单输出神经网络结构对高速列车停车制动过程进行深度强化学习;将深度强化学习方法应用到采用实际线路数据建立的仿真环境,计算得到高速列车停车制动的控制方法。本发明适用于轨道交通高速列车自动停车控制过程。本发明解决了现有方法停车制动控制参数寻优难的问题,基于多车多线路模型的强化学习环境使适用场景更加广泛,实现了深度强化学习方法在制动过程中实时调整最优控制参数。
Deep reinforcement learning method suitable for automatic parking of high-speed train
一种适用于高速列车自动停车的深度强化学习方法
2022-06-14
Patent
Electronic Resource
Chinese
Train accurate parking control method based on deep reinforcement learning
European Patent Office | 2022
|Train temporary parking strategy optimization method based on deep reinforcement learning
European Patent Office | 2023
|Automatic parking method based on deep reinforcement learning
European Patent Office | 2020
|Full-automatic parking lot scheduling method based on deep reinforcement learning
European Patent Office | 2020
|