【摘要】 由于受到非完整约束不可积性的影响，机器人移动轨迹跟踪相对复杂，控制困难，为了降低轨迹跟踪控制误差，提出基于LSTM的非完整约束机器人轨迹跟踪控制方法。通过数学模型表达非完整约束形式，基于拉格朗日法构建非完整约束下移动机器人的动力学模型，通过上述模型描述非完整约束移动机器人的系统动能。基于LSTM神经网络建立轨迹跟踪误差模型，分析机器人在移动过程中是否符合期望轨迹。根据轨迹跟踪的动态误差，在定向循环辅助下匹配适宜的期望控制参数，利用李雅普诺夫稳定函数设计轨迹跟踪控制律，对控制律函数进行微分处理，逐渐实现误差收敛，完成轨迹跟踪控制。经实验证明所提方法的运动轨迹跟踪控制误差较小，精准度高，可适用性强，控制效果好。更多还原

【Abstract】 Due to the influence of the non-integrability of nonholonomic constraints, tracking the robot trajectory is relatively complex. In order to reduce the trajectory tracking control error, this paper presented a method of controlling and tracking robot trajectory under nonholonomic constraints based on LSTM. Firstly, the form of nonholonomic constraint was expressed through a mathematical model, and then the Lagrangian method was adopted to construct a dynamic model of mobile robot under nonholonomic constraints. On this basis, the kinetic energy of mobile robot under nonholonomic constraints was described by this model. After that, the trajectory tracking error model was built based on LSTM neural network to analyze whether the robot was in accord with the expected trajectory in the process of moving. According to the dynamic error of trajectory tracking, the appropriate expected parameters were matched with the aid of directional circulation. Moreover, Lyapunov stability functions were used to design a trajectory-tracking control law. In the meanwhile, the functions of control law were differentiated to gradually realize error convergence. Finally, the trajectory tracking control was achieved. Experiment results show that the proposed method has smaller control error, higher accuracy, stronger applicability as well as better control effect.更多还原