Enhancing the robustness of GPC via a simple choice of the Youla parameter (Q1130309)

A two-step procedure for Generalized Predictive Control (GPC) design is presented. In the first step, a nominal controller is designed in order to minimize the GPC tracking performance index. In the second step the nominal controller is modified to take account of perturbations and unmodelled dynamics. The modification is based on the suboptimal (but simple and computationally fast) choice of a Youla parameter. The overall controller is shown to yield a better compromize between closed-loop performance and robust stability than is obtained with prevailing GPC methods.