Conception des controleurs flous multivariables.

Abstract

This work is a contribution for control of MIMO nonlinear systems, in the first contribution we consider a decoupling matrix based on linguistic reasoning between the laws control defined by the Backstepping SISO controllers. The approach proceeds by two steps, in the first step we built the backstepping SISO controllers by ignoring the interactions. In the second step we construct a fuzzy supervisor under the Takagi-Sugeno reasoning and working in feedforward. It has as inputs the outputs of the SISO controllers and produces compensations signals; the final control vector is the result of addition of the loop- control and compensations signals In the second contribution we consider a decentralization approach in order to minimize the interactions effect coming from the internals states and the parameters variations sensibility. This approach uses the nonlinear damping theory. In order to proceed by optimal design we introduce the multi- objectives genetic algorithms. Three processes are considered to these two approaches: the wastewater station, the two-linked robot and the two – linked pendulum.