Abstract: In view of the operating characteristics of a multiple joint robot driven by ultrasonic motors, a new speed-position double closed-loop PID control mode is put forward, which has effectively curbed the dramatic change of speed caused by quick response of the ultrasonic motor, so as to enhance the smooth movement of the robot. To further improve the performance of the robot and to real-time adjust the PID control parameters according to the changes of characteristics of ultrasonic motors, a PID-like wavelet neural networks (WNN) is presented. The speed-position double closed-loop PID control, practical experience, the current trajectory error and its change are all taken into account in the frame of the proposed PID-like wavelet neural networks. A back propagation algorithm based on gradient descent is developed to train the PID-like WNN on line by using δ adaptation law. Experiment results prove that the PID-like WNN controller is superior to PID controller and PID neural networks controller, and has a good anti-interference capability.

Key words: Joint robot, PID, Ultrasonic motor, Wavelet neural network