Robust distributed model predictive control with cooperation

This paper extends a form of distributed model predictive control (MPC) for subsystems with decoupled dynamics and with coupled constraints. By use of an objective function that includes knowledge of the coupling between subsystems, cooperative behaviour is promoted with respect to the global objective. The method is robust to persistent disturbances, by the use of the Tube MPC concept, in which an optimization designs a sequence of invariant state sets for the system to follow, rather than a trajectory. Robust stability is guaranteed for any choice of update sequence and objective function. It is demonstrated that, through cooperation, global performance can improve on that obtained by a simple distributed implementation.