A stabilizing model-based predictive control algorithm for nonlinear systems (Q5947644)

A receding-horizon predictive control scheme for nonlinear discrete-time systems subject to state and input constraints is considered. By working out suitable bounds, it is shown how to truncate the series which express the infinite-horizon cost (associated with the given auxiliary linear control law) without losing stability. Also, the use of a long prediction horizon in the cost function is made possible by dividing the horizon into two parts. The optimization (over the control horizon) is carried out only with respect to the first few input moves whereas the state is predicted and penalized over a longer horizon in which the remaining input moves are calculated on the basis of the auxiliary linear control law. It is argued that the proposed control scheme allows the stability region to be enlarged without substantially increasing the computational burden. As an illustrative example, the control of a highly nonlinear chemical reactor is discussed.