Feedback model predictive control by randomized algorithms

In this paper we present a further development of an algorithm for stochastic disturbance rejection in model predictive control with input constraints based on randomized algorithms. The algorithm presented in [1] can solve the problem of stochastic disturbance rejection approximately but with high accuracy at the expense of a large computational effort. The algorithm described here uses a predefined controller structure in the optimization and it is significantly less computationally demanding but with a price of some performance loss. Via an example it is shown that the algorithm gives considerable reduction in the computational time and that performance loss is rather small compared to the algorithm in [1].