Towards the application of multiple model adaptive control schemes

When there is a high degree of uncertainty associated with the modelling of the plant dynamics, the parameters in the adaptation algorithm cannot be initialised “close” to their real values and the identification process can take a considerable time to complete due to the slowness of the algorithm. Such difficulties have led to the emergence of a new general approach to the adaptive control of plants prone to these and related problems. Instead of using a single model to describe the plant dynamics and hence controller design/evaluation, multiple models are used to describe the underlying dynamics. Also a controller is designed for each of the models employed based on the performance specifications. This paper concentrates on the development of a software framework for the evaluation of schemes based on this overall approach prior to trials on the “real” plant. The key part of this approach is the multiple models and controllers and the switching between them where the overall goal is perfect model reference adaptive control of the underlying plant