A methodology for multiple rule system integration and resolutions within a singular knowledge base

• Authors: Frank N. Kautzmann, III
• Venue: CSC '87 Proceedings of the 15th annual conference on Computer Science
• Year: 1987

Expert Systems which support knowledge representation by qualitative modelling techniques experience problems, when called upon to support integrated views embodying description and explanation, especially when other factors such as multiple causality, competing rule model resolution, and multiple uses of knowledge representation are included. MITRE Corporation, in conjunction with several directorates of NASA at Johnson Space Center, is currently developing a series of prototypes to demonstrate the feasibility of automating the process of systems engineering, design and configuration, and diagnosis and fault management. It is important to study these processes as they pertain to the analysis of design tradeoff issues, knowledge capture and Space Station planning and assessment. The MITRE effort will involve not only a generic knowledge representation; it must also support multiple views at varying levels of description and interaction between physical elements, systems, and subsystems. Moreover, it will involve models of description and explanation for each level. This multiple-model feature requires the development of control methods between rule systems and heuristics on a meta-level for each expert system involved in an integrated and larger class of expert system. This paper describes the broadest possible category of interacting expert systems and proposes a general methodology for the knowledge representation and control of mutually exclusive rule systems within the context of a single knowledge representation scheme.