Model-based predictive sampled-data control and its robustness

This paper proposes a model-based predictive sampled-data controller with a large fixed sampling rate h. Although the linear-time-invariant (LTI) plant is unknown, a nominal model is available. This nominal model is used to predict and compensate the influence of the large sampling using the measured information from the plant. The controller is designed on the basis of the nominal model. The robustness and performance of this model-based predictive sampled-data controller are explored with respect to the sampling rate h, the mismatches between the nominal model and plant as well as the choice of the feedback gain matrix K. It is interesting to observe that the robustness of the proposed method is not proportional to the sampling rate h, neither a small h nor a large h is robust. Maximum robustness requires a well-chosen finite sampling rate h.