RBFNN based adaptive control of uncertain robot manipulators in discrete time

Abstract

The trajectory tracking control problem for a class of n-degree-of-freedom (n-DOF) rigid robot manipulators is studied in this paper. A novel adaptive radial basis function neural network (RBFNN) control is proposed in discrete time for multiple-input multiple-output (MIMO) robot manipulators with nonlinearity and time-varying uncertainty. The high order discrete-time robot model is transformed to facilitate digital implementation of controller, and the output-feedback form is derived to avoid potential noncausal problem in discrete time. Furthermore, the desired controller based on RBFNN is designed to compensate for effect of uncertainties, and the RBFNN is trained using tracking error, such that the stability of closed-loop robot system has been well guaranteed, the high-quality control performance has been well satisfied. The RBFNN weight adaptive law is designed and the semi-global uniformly ultimate boundedness (SGUUB) is achieved by Lyapunov based on control synthesis. Comparative simulation studies show the proposed control scheme results in supreme performance than conventional control methods.