Stochastic linear model predictive control using nested decomposition

Abstract

We begin with a traditional model predictive control problem using the l1 norm in the objective function, and then allow the model parameters to be stochastic, with discrete distributions and finite support. We apply the nested decomposition algorithm for multistage stochastic linear programming to the resulting problem. The result is an algorithm for model predictive control that features the realism of model uncertainty, the potential speed of linear objective functions, and can be implemented in parallel.