Neural Network Tracking Controls of SCARA Manipulator System
Yongfeng Lv, Yingbo Huang, Xiaolong Wu, Long Jian · 2021
In this paper, the neural network optimal tracking controller of SCARA manipulator system is designed. First of all, the SCARA manipulator system is modeled, and the reference motion trajectory of the joint is given, and the steady-state control is designed to ensure that the manipulator can keep up with the reference trajectory, but it can not guarantee the various performance in operation. Given the performance index of position and speed tracking error, the neural network approximate feedback control is obtained by learning the optimal performance index function with three-layer neural network based on reinforcement learning (RL). According to the steady-state control and approximate feedback control, the neural network optimization controller of the manipulator is designed to achieve the tracking effect of the position speed of the manipulator with the minimum overshoot and chattering, and the lowest energy consumption.