Non-linear model based predictive control

Linearization, interpolation and end-point constraints are combined to derive a receding horizon predictive control algorithm for continuous-time non-linear systems which are subject to input constraints. The algorithm makes extensive use of the 'tail' of input/output state trajectories, namely the extension to current time of trajectories computed at the previous time instant. Through this device, the algorithm has guaranteed feasibility and stability and has some desirable attributes with respect to dynamic performance. The results of the paper are illustrated by a model of a coupled tank system.