Arc Consistency for Dynamic CSPs

Abstract

Constraint Satisfaction problems\,(CSPs) are a fundamental concept used in many real world applications such as interpreting a visual image, laying out a silicon chip, frequency assignment, scheduling, planning and molecular biology. A main challenge when designing a CSP-based system is the ability to deal with constraints in a dynamic and evolutive environment. We talk then about on line CSP-based systems capable of reacting, in an efficient way, to any new external information during the constraint resolution process. We propose in this paper a new algorithm capable of dealing with dynamic constraints at the arc consistency level of the resolution process. More precisely, we present a new dynamic arc consistency algorithm that has a better compromise between time and space than those algorithms proposed in the literature, in addition to the simplicity of its implementation. Experimental tests on randomly generated CSPs demonstrate the efficiency of our algorithm to deal with large size problems in a dynamic environment.