Report on distributed model predictive control for cyber-physical systems and efficient computation

This deliverable will document the results achieved in tasks 2.2 and 2.3. These tasks concern Model Predictive Control (MPC) of a cyber-physical system composed of interacting, possibly hybrid, subsystems that may operate in an uncertain environment. The goal is to derive distributed control schemes where each subsystem repetitively solves online a local control problem, while accounting for the other subsystems through time-varying and uncertain constraints. Stability and performance results will be investigated for the subsystems and the overall networked system. When the adopted description of the uncertainty is probabilistic, then, the control problem to be solved online naturally reduces to a constrained stochastic optimization problem, which will be addressed using jointly randomized and robust optimization techniques. Computational aspects related to the online implementation of the proposed distributed MPC solutions will be addressed by combining approximation methods ( e.g., model abstraction, constraint convexification, multiple time scale resolution) in an efficient scheme preserving stability while meeting real-time constraints posed by the online implementation.