We argue that the development of artificial intelligence techniques is bringing about fundamental change in the way we represent and reason about the physical world. From a control engineering perspective, such methods offer a significant extension of the available method for systems modelling, and hence open up exciting prospects for the diversification of control methods to other application areas, e.g. automated fault diagnosis, simulation and training. However, such diversification brings with it the need to clearly establish the principles, and hence the limitations, behind each technique. Accordingly, we propose a classification of system models in terms of their knowledge classes and characteristics, and relate these to existing approaches to the use of Al methods in control. Such a classification is a necessary precursor to developing a methodological approach to identifying the most appropriate technique for a given generic class of applications.
|Number of pages||11|
|Journal||Computing and Control Engineering|
|Publication status||Published - Jul 1992|