Approximation de la commande prédictive par un système d’inférence flou.

Abstract

In this thesis, we exploited the universal approximation property of Takagi-Sugeno fuzzy systems to construct a fuzzy approximation of the optimal linear and nonlinear model (MPC) predictive control solution. The nonlinear systems considered are affine in the control law. The fuzzy approximator presents some properties generally not shown by the preceding ones. In particular, it consists of a set of state feedback control laws which are merged to obtain the final nonlinear control law. The constructed control law is very similar to the explicit solution of the linear MPC. The stability is analyzed a posteriori, for the linear and non-linear MPC, on the basis of a new theoretical test developed for fuzzy systems of the Takagi Sugeno type. The examples treated in the simulation have shown the promptness and the good results which prove the effectiveness of the control strategy developed.