The paper is logically divided into three parts. In the first part the following question is considered. How a standard Model Predictive Control algorithm would perform in a case of model-process discrepancy? It turns out that this discrepancy could lead to control performance deterioration and even to instability of a predictive control system. In the sequel Uncertainty is described as model uncertainty, and then is introduced mathematical description of uncertainty as an analytical uncertainty model. The analytical uncertainty model can be directly used during predictive control algorithm design. In the third part some new ideas and solutions are proposed. First of all, to efficiently cope with negative effects of model-process mismatch the model uncertainty should be taken into account already during predictive control algorithm design. It can be done by the use of the analytical uncertainty model. According to the above solutions like: a fast model selection, dynamical switching sector, and a dual-phase approach to a predictive control algorithm design are presented.
- systems of steering
- arrangement of control