Off-line improvement of multivariable model predictive control robustness

This paper presents a unified off-line methodology to robustify an initial multivariable model predictive stabilizing controller through the convex optimization of a multivariable Youla parameter. The main innovative purpose is to guarantee robust stability under model unstructured uncertainties and nominal performance specifications for the disturbances rejection. This robustification technique is applied to the control of a stirred tank reactor to reduce the influence of unstructured uncertainties on the system, while respecting time-domain templates for disturbances rejection.