Observers in nonlinear model-based predictive control

A closed-loop reformulation of the dual-mode paradigm affords significant advantages in model-based predictive control of systems subject to uncertainty and/or disturbances and/or nonlinear dynamics. This paper considers earlier results based on ellipsoidal invariant sets and proposes extensions to the output feedback nonlinear case by defining invariant sets both for the system state and the error dynamics of a nonlinear observer. Particular attention is paid to the class of systems with separable nonlinearities but the results carry over to the general case.