On Problems of Knowledge in Fuzzy Control

The field of Fuzzy Control have enjoyed tremendous success in the last decade, with both theoretical and industrial developments being introduced at an increasing rateI. However, fuzzy control is just one application of Soft Computing methods in general, and Fuzzy Sets theory in particular. This principled approach to approximate reasoning is not limited only to control problems, but is useful also in closely related fields, such as Artificial Intelligence (AI) in its various forms and guises, Decision Sciences, Quantitative and Qualitative Decision Theory, theoretical studies of Uncertainty, Information and Knowledge, etc. As we seek to expand the success of fuzzy control to these fields, it is very tempting to attempt use of known methods in a familiar way to solve problems in these new domains. However, a principled analysis of the goals and assumptions underlying different fields may reveal important differences from the field of control which may make current methods insufficient. We take the position that as we look at real-world decision problems and domains, the familiar techniques of fuzzy control will be insufficient to provide adequate solutions, because these techniques are designed with the assumptions of the field of control in mind --assumptions which do not hold in such domains. The techniques are not incorrect, but simply insufficient, and can be augmented to provide the necessary theoretical and practical infrastructure for the new domains. This short abstract will attempt to point out our initial approach to such necessary augmentations of fuzzy control techniques. We will motivate the discussion with a very small scale decision problem, which despite its size, captures some of the underlying problems with current techniques and points the way at the necessary directions for development that will bridge the gap between fuzzy control techniques and the required technology.