A repetitive learning method based on sliding mode for robot control

Abstract

In order to make a robot precisely track desired periodic trajectories, this work proposes a sliding mode based repetitive learning control method, which incorporates characteristics of sliding mode control into repetitive learning control. The learning algorithm not only utilizes shape functions to approximate influence functions in integral transforms, but also estimates inverse dynamics functions based on integral transforms. It learns at each sampling instant the desired input joint torques without prior knowledge of the robot dynamics. To carry out sliding mode control, a reaching law method is employed, which is robust against model uncertainties and external disturbances. Experiments are performed to validate the proposed method.