Input-output linearizing nonlinear model predictive control

Abstract

Continuous-time nonlinear feedback control of multivariable processes with deadtimes is considered. The problem of deadtime compensation for a class of nonlinear multivariable processes is solved. The nonlinear controllers are derived by using two conceptually different approaches, a differential geometric approach and a model predictive approach. The model predictive approach uses shortest prediction horizons which allow one to establish the connections between model predictive control and input-output linearization. These two approaches lead to identical feedback controllers when there are no constraints or when the characteristic (decoupling) matrix of process can be made diagonal by manipulated-input/controlled-output rearrangements. Connections between the derived controllers and the linear internal model control are established. The application and performance of one of the derived nonlinear controllers is illustrated by a reactor example.