Abstract

This paper addresses the problem of controlling a class of unknown nonlinear systems with unstable inverse, by using a generalized minimum variance approach. We highlight the limitations of the classic cost function used in this context and we propose a new one based on the Generalized Feedback Linearization concept. The adaptive control algorithm uses a nonparametric model, which provides an estimate of the mean and variance of the system output. Simulation examples illustrate the main characteristics and limitation of the proposed approach. Copyright © 2005 IFAC.