Model predictive control

Abstract

The chapter contains the results of the original research dealing with robust and fault-tolerant predictive control schemes. The first part of the chapter is devoted to nonlinear predictive control developed by means of neural networks. Some of the most important issues connected with optimization and stability are investigated in detail. The next part introduces the sensor fault-tolerant control (For this purpose, predictive control is equipped with a fault-diagnosis block.) Binary diagnostic matrix as well as multivalued diagnostic matrix are used in this context. The proposed control strategy is tested using the tank unit example provided. We develop a robust version of predictive control based on a robust model of a plant. We investigate two approaches: uncertainty modelling using model error modelling and statistical uncertainty estimation via statistical analysis. The proposed control schemes are tested on the example of a pneumatic servomechanism.