Decentralized Model Predictive Control of Dynamically-Coupled Linear Systems: Tracking under Packet Loss

For large-scale processes whose dynamics can be represented as the interaction of several dynamically-coupled linear subsystems, this paper proposes a decentralized model predictive control (MPC) design approach for set-point tracking under input constraints and possible loss of information packets. Following earlier results in (Alessio and Bemporad, 2007, 2008), the global model of the process is approximated as the decomposition of several (possibly overlapping) smaller models used for local predictions. We present sufficient criteria for asymptotic tracking of output set-points and rejection of constant measured disturbances when the overall process is in closed loop with the set of decentralized MPC controllers, under possible intermittent lack of communication of measurement data between controllers. The effectiveness of the approach is shown in a simulation example on distributed temperature control in the passenger area of a railcar.