ADAPTIVE NEURAL DYNAMIC SURFACE CONTROL WITH TRUNCATED ADAPTATION FOR UNCERTAIN SATURATED NONLINEAR SYSTEMS

摘要

This paper considers adaptive neural control for a class of uncertain saturated nonlinear systems. To overcome the problem of "explosion of complexity" that exists in the traditional backstepping design, dynamic surface technique is utilized to avoid the repeated differentiations of virtual controllers. Novel truncated adaptation technique is proposed to attenuate the effect cause by input saturation. Radial basis function (RBF) neural networks (NNs) are used to online approximate uncertain system dynamics. Auxiliary signals generated by properly designed auxiliary system are used to truncate the training signal of RBF NNs when input saturation happens. The stability of closed-loop system is guaranteed and proved using Lyapunov stability theorem. The tracking error can be made to arbitrarily small by tuning design parameter in an explicit way even with input saturation in effect, and the compact sets that steady tracking error and transient tracking error are confined in are given, which are characterized by design parameters. Simulation results are presented to verify the effectiveness of the proposed control scheme.