Joint Torque Feedback-based Decentralized Neuro-optimal Control of Input-constrained Modular Robot Manipulator System

Abstract

This paper presents a decentralized neuro-optimal control for input-constrained modular robot manipulator (MRM) system based on joint torque feedback (JTF) technique. Utilizing the joint torque sensor measurement, the dynamic model of MRM subsystem is constructed. An improved performance index function is formulated by utilizing the hybrid tracking errors, measured model information, constrained control input torque and uncertainties. On the basis of adaptive dynamic programming (ADP) approach, the corresponding Hamilton-Jacobi-Bellman (HJB) equation is settled via critic neural network (NN) structure, thus, the decentralized optimal control strategy is obtained. The trajectory tracking error of the MRM subsystem is proved to be ultimately uniformly bounded (UUB) under the Lyapunov stability theorem. The effectiveness of the developed control strategy is guaranteed by experimental results.