Vibration suppression of flexible multi-arm space manipulator system based on neural network
- Paper number
IAC-20,D1,1,5,x58389
- Author
Prof. Shuang Li, China, Nanjing University of Aeronautics and Astronautics
- Coauthor
Dr. Yinkang Li, China, Nanjing University of Aeronautics and Astronautics
- Year
2020
- Abstract
As one of the important actuators in the future space operation tasks, multi-arm space manipulator system has very important research significance. Due to the flexibility of the space flexible manipulator system, residual vibration will inevitably occur during the execution of the task. Such vibration will reduce the positioning accuracy of the system and adversely affect the stability and reliability of the control system. The dynamic model of the flexible multi-arm space manipulator system is more complicated, which not only generates more disturbances and errors in control system, but also brings greater challenges to the vibration suppression method. The purpose of this research is to provide a vibration suppression method based on neural network for flexible multi-arm space manipulator system. In view of the above difficulties, corresponding methods are proposed . Considering the complicated dynamic model of the flexible multi-arm space manipulator system, various disturbances and errors will inevitably occur in the control system. As for this, a neural network fitting algorithm is adopted to realize the observation and feedforward of control system errors, the dynamic disturbances are evaluated online by deep learning algorithm, and the weights are learned online by using the basis function. Through the above operations, the current system errors can be accurately estimated and feed-forwarded into the control algorithm, thereby improving the accuracy and reliability of the control algorithm of the flexible multi-arm space manipulator system. In order to suppress the residual vibration of the flexible multi-arm space manipulator system, the method of combining input shaping control with RBF neural network is adopted in this paper. The traditional input shaping control method needs to design the system's input shaper through the relevant parameters of the system (such as natural frequency, damping ratio, etc.). However, the use of traditional modal function theory to calculate the flexible vibration displacement of manipulator has disadvantages such as low accuracy and heavy calculation. Therefore, a deep learning algorithm is adopted in this paper to fit the vibration displacement of the manipulator endpoint by the speed, angular velocity of the joints and current control input, and then to identify the system's natural frequency, damping ratio and other relevant parameters. After that the input shaper can be designed by the obtained parameters, so as to suppress the residual vibration of the flexible manipulator endpoint and improve the control accuracy.
- Abstract document
- Manuscript document
(absent)