Adaptive MPC of a class of Switched Linear Systems with Unknown System Matrices

This paper presents an adaptive model predictive control for a class of switched linear stochastic systems with unknown system matrices. The proposed adaptive MPC is designed by solving a finite horizon constrained linear-quadratic optimal control problem of on-line estimated models, which are built on a switching recursive weighted least-squares (WLS) algorithm together with a random regularization method. By incorporating an attenuating excitation signal into adaptive MPC, the switched system satisfies the state constraints and control constraints after finite steps.