Adaptive Neural Network Observer Based PID-Backstepping Terminal Sliding Mode Control for Robot Manipulators

Abstract

In this paper, a single weight RBF neural network based state and disturbance observer and the observer based proportional integral differential backstepping terminal sliding mode controller (PID-BTSMC) are proposed for control of robot manipulators subject to system uncertainties, external disturbances and unmeasured states. The single weight RBF neural network is first time used in design of state and disturbance observer to improve the online learning efficiency for practical engineering applications. The observer based backstepping terminal sliding mode controller (BTSMC) is introduced with the merits of high robustness, fast transient response, finite time convergence and globally asymptotic stability. Then a PID-BTSMC is proposed which preserves the merits of both PID and BTSMC. The proposed controller is applied for tracking control for a single link robot system and compared with the related PID, Backstepping and nonsingular fast terminal sliding mode controller. The superior performance of the proposed approach is demonstrated in the comparison results.