Detail of publication

Abstract

This paper deals with performance and robustness of simple model based predictive controller with manipulated value constraints. To simplify the open-loop optimization procedure, the manipulated value is restricted to be constant over and beyond the control horizon. Even such simple algorithm can result in very good closed loop performance but it is necessary to use a prediction horizon significantly shorter than the settling time of the controlled system. This modification injects derivative character into the resulting controller. The overall performance of closed loop with respect to prediction horizon length is evaluated using ITAE criterion. It is shown that this "ultra-simple" predictive control approach gives unique results for stable, linear and t-invariant FIR systems with monotonous step response, moreover it is quite robust to model uncertainties: first order plus dead-time or second order plus dead-time approximation of the controlled system can be used for prediction.