Composite NNs learning full-state tracking control for robotic manipulator with joints flexibility

Abstract

In this paper, the tracking control problem of flexible-joint manipulator (FJM) system subjected to system uncertainties and time-varying external disturbances is addressed. Firstly, a novel composite neural networks learning (CNNL) control framework that blends the advantage of neural networks (NNs) and terminal sliding mode disturbance observer (TSMDOB) is proposed. Furthermore, based on the backstepping theory, a full-state CNNL tracking control scheme is developed for the FJM system. Then the link-side performance is enhanced without using high-order derivatives of the link states. NNs are employed to approximate the system uncertainties of FJM and TSMDOB is mainly used to deal with the time-varying external disturbances. Significantly, the system convergence speed and accuracy are improved with added prediction errors under a composite framework. Finally, to validate the effectiveness of the proposed method, numerous simulation results on 2-link flexible-joint robotic manipulator are provided. In comparison to the other state-of-the-art approaches, the proposed control strategy possesses several advantages.