Controle preditivo robusto baseado em modelo aplicado a sistemas lineares com saltos Markovianos

Abstract

In this thesis, robust model predictive control techniques applied to discrete-time Markov jump linear systems are introduced. Two control scenarios are addressed. In the first scenario, it is developed a control solution that minimizes the expected value of an infinite-horizon quadratic cost. As a byproduct, mean square stability is obtained under two cases: i) without constraints, and ii) with constraints on control input and state. The second control scenario considers the minimization of the expected value of quadratic finite-horizon cost. This scenario considers not only stochastic additive noise, but also constraints that are imposed on the second moment of both state and control. Numerical experiments illustrate the results for both scenarios.