Adaptive learning control of robot manipulators via incremental hybrid neural network

Abstract

A novel hybrid neural network based learning control method is proposed to improve trajectory tracking accuracy for complex robot manipulators in this paper. Firstly, a hybrid neural network is presented to improve the model accuracy and data efficiency, which is integrated by the Differential Newton-Euler Algorithm (DiffNEA) and Radial Basis Function Neural Network (RBFNN). In this hybrid neural network, the DiffNEA takes in charge of modeling the known rigid dynamics, while RBFNN takes in charge of capturing the unmodeled phenomena and external disturbances. Secondly, an incremental design method is proposed to determine optimal structure and parameters of the hybrid neural network. Thirdly, an adaptive learning controller based on the aforementioned hybrid neural network is designed to further reject the effects of unmodeled dynamics and external disturbances on trajectory tacking performance. Finally, experimental results are presented to validate the effectiveness of the proposed method.