列车自动驾驶迭代学习控制研究

Study on Iterative Learning Control in Automatic Train Operation

地铁列车精确停车研究对装有屏蔽门的车站乘客换乘有着现实意义。根据城轨列车在停车阶段的重复特性,通过迭代学习调节列车初始状态来克服停车阶段的重复不确定性。在求解列车制动模型微分动态的基础上,获得系统梯度,进而求取满足收敛条件的学习参数。针对列车停车涉及多目标的特点,定义多目标优化函数,将该方法推广到多变量、多目标调节的情况,可以同时满足城轨列车停车精度和乘坐舒适性的要求。仿真结果验证了该方法的有效性。最后,提出基于迭代学习方法研究列车停车问题的未来研究思路。

Study on accurate automatic train stopping holds great practical significance to passenger tansfor at subway stations installed with platform screen doors. In view of the repetitive characteristics of subway trains in stopping the iterative learning control scheme was implemented to adjust the initial state of the trains to eliminate repetitive uncertainty of stop. By solving the differential dynamic train braking model, the system gradient information was obtained and the iterative learning parameters satisfying convergence conditions were found. The multi-object optimization function was defined in consideration of the multiple objectives involved in train stopping. The method was extensively applied in adjustment of multiple various and multiple objectives, which fulfilled the requirements of stopping accuracy, riding comfort and energy saving. The simulation results verified the effectiveness of the method. Finally, future research direction of ATO train stopping was pointed out.