Constraint Relaxation Techniques to Aid the Reuse of Knowledge Bases and Problem Solvers

Tomas Eric Nordlander, Kenneth N. Brown, Derek H. Sleeman

Abstract: Effective re-use of knowledge bases requires the identification of plausible combinations of both problem solvers and knowledge bases, which can be an expensive task. Can we identify impossible combinations quickly? The capabilities of combinations can be represented using constraints, and we propose using constraint relaxation to help eliminate impossible combinations. If a relaxed constraint representation of a combination is inconsistentthen we know that the original combination is inconsistent as well. We examine different relaxation strategies based on constraint graph properties, and we show that removing constraintsof low tightness is an efficientstrategywhich is also simple to implement