Distributed Predictive Control of Stochastic Linear Systems with Chance Constraints

This paper presents a distributed Model Predictive Control algorithm for linear dynamically interconnected subsystems subject to additive, possibly unbounded, stochastic noise. The method is based on the reformulation of the probabilistic chance constraints on the states and the inputs of the subsystems in terms of tighter deterministic ones, on the use of ideas borrowed from the robust tube-based control, and on the inclusion of suitable terminal constraints at the end of the prediction horizon. The convergence and recursive feasibility properties of the proposed algorithm are established and the tuning and implementation issues are discussed.