A framework for managing uncertainty in embedded expert systems

Abstract

The uncertainty management system (UMS) shell and its management of uncertainty are described. UMS is an expert system building tool dedicated to real-time embedded applications. This system allows the user to specify knowledge using a set of linguistic approximations, without being conscious of data-processing internally performed. Supported by a development of the axioms of probability, an interval-based framework for managing uncertainty is presented. As generally admitted, an interval number is used to capture features of fuzziness and incompleteness. However, a new parameter, called the degree of dependency, is shown to be naturally induced by M.J. Frank's (1979) family of triangular norms.<>