Stability and optimality of distributed, linear model predictive control∗ Part II: Output Feedback

The distributed model predictive control (MPC) framework described in [38] is extended to handle output feedback. A distributed estimator design strategy is proposed. Each estimator is stable and uses only local measurements to estimate subsystem states. The feasible cooperation-based MPC (FC-MPC) algorithm presented in [38] is used for distributed regulation. Feasibility and closed-loop stability for all (FC-MPC algorithm) iteration numbers are established for the combined distributed estimator-distributed regulator in the case of decaying estimate error. A subsystem-based disturbance modeling framework to eliminate steady-state offset due to modeling errors and unmeasured disturbances is next presented. Conditions to verify suitability of chosen local disturbance models are provided. A distributed target calculation algorithm to compute steady-state targets locally is presented. All iterates generated by the distributed target calculation algorithm are feasible steady states. Conditions under which the FC-MPC framework, with distributed estimation, distributed target calculation and distributed regulation, achieves offsetfree control at steady state are described. The effectiveness of the proposed framework is illustrated through two examples a chemical plant and an irrigation canal network.