Stochastic Adaptive Control of Multivariable Systems with Dead-Zone Nonlinearities

Abstract

This paper presents a novel scheme for the direct stochastic adaptive control of a class of nonlinear dynamic systems. This class is characterized by a cascade of dead-zone nonlinearities and a linear multivariable system with a general interactor matrix. A piece-wise linear preload vector is introduced to invert the dead zone nonlinearity vector. An optimal adaptive control law is derived using a cost function in which the nonlinear parameter vector of the model is included. A switching gain sequence vector is employed in order to overcome problems of parameter estimation. This scheme is applicable even to systems whose linear parts are open-loop unstable and/or non-minimum phase processes. The algorithm ensures global stability and convergence.