摘要
对于带有柔性原件的空间机器人,其精确的动力学模型往往难以建立。柔性原件所引入的非线性因素也同时增加了机器人控制系统的复杂度。为了提高柔性关节机器人的轨迹跟踪性能,提出了无模型的柔性关节智能控制方法,建立了新的协同学习框架。并将此框架用于的智能控制器的训练。最终在轨迹跟踪实验中,和传统PID控制器、非协同学习的RBF神经网络PID控制器相比较,智能控制器的跟踪误差减少了39.24%和32.70%。这表明了新的协同学习框架和智能控制器的有效性和可行性。
For space robots with flexible components,its precise dynamic model is often difficult to establish.The non-linear factors introduced by the flexible components also increase the complexity of the robot control system.In order to improve the trajectory tracking performance of flexible joint robots,this paper proposes a model-free intelligent control method for flexible joints and establishes a new collaborative learning framework,and uses this framework for the training of the intelligent controller.Finally,in the trajectory tracking experiment,compared with the traditional PID controller and the non-cooperative learning RBF neural network PID controller,the tracking error of the intelligent controller is reduced by 39.24%and 32.70%.This shows the effectiveness and feasibility of the new collaborative learning framework and intelligent controller.
出处
《工业控制计算机》
2021年第5期97-99,101,共4页
Industrial Control Computer
关键词
柔性关节
协同学习
智能控制器
神经网络
flexible joints
collaborative learning
intelligent controller
neural network
