Learning from adaptive neural control of manipulator using an ISS-modular approach

Abstract

This paper studies learning from adaptive neural control of manipulator using an input-to-state stability (ISS) modular approach. In order to remove the limitation of the derivative of the affine term in robot manipulator, an ISS modular method is introduced into this paper. Meanwhile, the proposed adaptive neural control law, based on the ISS modular method and backstepping technology, guarantees the boundedness of all signals in closed-loop system and exponential convergence of tracking errors. Moreover, by using a state transformation strategy and coupling with the established partial persistent excitation (PE) condition, neural network(NN) weight values converge exponentially to their optimal values. The convergent weights which are stored as the constant experienced knowledge make the learning control for a similar task successful. Simulation results are shown to demonstrate the effectiveness of the proposed control method.