Abstract: To improve the working performance of robotic trajectory tracking controller,the robotic trajectory tracking control method based on reinforcement learning was proposed.The basic principle of reinforcement learning was introduced,and then the robot trajectory tracking control strategy based on SARSA was proposed.By using the reinforcement learning,the unknown disturbance factors were compensated and the adaptability to the unknown was improved after the PD control method was applied.The experimental results verified the feasibility and effectiveness of the reinforcement learning method in the trajectory tracking problem of robot arms,and the learning speed of the controller was enhanced.

Key words: robot, reinforcement learning, trajectory tracking, PD controller, feedforward neural network

摘要： 为了提高机器臂轨迹跟踪控制器的工作性能,提出基于增强学习的机械臂轨迹跟踪控制方法。介绍了增强学习的基本原理,提出基于SARSA算法的增强学习补偿控制策略。利用比例—微分(PD)控制器完成了基本的稳定任务后,再利用增强学习算法实现了对未知干扰因素的补偿,提升了对不同未知情况的适应能力。实验结果验证了自适应离散化增强学习方法在机械臂轨迹跟踪问题中的可行性和有效性,明显提高了控制器的学习速度。

关键词: 机器人, 增强学习, 轨迹跟踪, 比例&mdash, 微分控制器, 前馈神经网络