On model uncertainty in adaptive controller design

The accurate representation of processes for controller design is, in many situations, a convoluted problem; lack of process information, neglected process dynamics, time varying characteristics and inherent nonlinearity all complicate the modelling process. Two distinct approaches have emerged to deal with the problem of model uncertainty: robust control and adaptive control. By their nature, robust controllers are conservative with corresponding performance degradation. Adaptive controllers, on the other hand, are sensitive during process changes until the new information is reflected in the process model. In this paper, a methodology is proposed to enhance the properties of an adaptive controller using concepts from robust control theory. The use of a robust design procedure coupled to an adaptive process model allows the controller to be tuned for high performance when the model is accurate and to be sufficiently de-tuned to maintain system performance during periods of model inaccuracy. The procedure is considered with reference to an adaptive inferential controller, an algorithm developed for systems with inherent uncertainties.