Contribution à la Commande Adaptative Robuste des Processus Physiologiques par Mode Glissant, Contrôleur Flou et Modèle de Référence

Abstract

The problem of the closed-loop control of physiological systems is investigated in this thesis. Three main controller types are employed: model reference adaptive control, fuzzy logic control and the positive sliding mode control. Our major contribution is the control of these systems; where we have chosen one of the most important physiological systems which is the glucose-insulin system for regulating the glycemia of type 1 diabetes (DT1). Two important models are given and discussed: The first is a long-term model that is derived from clinical data of DT1 subjects and developed in LS2N laboratory in France. The second is the Bergman model. To control these models, the three types of commands are used. First of all, the adaptive reference model control is designed, we created different types of controller by this command: with and without constraints. The first one is based on the standard forms: CGT, and MRAC established on the rules (MIT), which do not constitute a general solution for the control problems of the physiological systems. The second is based on creating constraints to design new saturation methods that have been added to constrain the controller and the output to be non-negative. Secondly, a fuzzy controller is used for the first time for Madeleine’s long-term model for blood glucose regulation. Moreover, this controller is developed for the Bergman model. Finally, a positive guaranteed sliding mode control is designed for the first time for regulating the blood glucose. The last control is chosen for the regulation of blood glucose because of its well-known robustness properties. However, our major contribution is the assurance of a rigorously positive order. This command is designed to be positive throughout an invariant set of the plasma insulinemia subsystem. The three controllers’ types are tested in silico on both DT1 models, and a comparative study between these commands is included in this thesis.