Radial-basis-functions neural network sliding mode control for underactuated mechanical systems

Abstract

This paper presents an indirect adaptive neural network sliding mode Control (IANSMC) technique and a neural network sliding mode control (NNSMC) for underactuated robot manipulators. The adaptive neural network (NN) based on radial basis functions (RBF) is used to estimate the equivalent control and to compensate model uncertainties. In IANSMC, the adaptive learning algorithms are derived using Lyapunov stability analysis. Sliding mode control and indirect adaptive technique are combined to deal with modeling parameter uncertainties and bounded disturbances. The stability of the mixed controller is then proved. NN parameters are tuned on-line, without an off-line learning phase. For the NNSMC, the NN control is used to learn the equivalent control due to the unknown nonlinear system dynamics and the robust sliding mode control (SMC) is designed for a trajectory tracking control. Simulation results show that the NNSMC and IANSMC are better than the classical SMC to control underactuated manipulators. Although the proposed controllers can eliminate the chattering phenomena and estimate matching uncertainties. The IANSMC can also reject mismatched perturbations. Discussions and comparisons between proposed controllers are presented.