A-system: Problem solving through abduction

Abstract

This paper presents a new system, called the A-System, performing abductive reasoning within the framework of Abductive Logic Programming. It is based on a hybrid computational model that implements the abductive search in terms of two tightly coupled processes: a reduction process of the high-level logical representation to a lower-level constraint store and a lower-level constraint solving process. A set of initial "proof of principle" experiments demonstrate the versatility of the approach stemming from its declarative representation of problems and the good underlying computational behaviour of the system. The approach offers a general methodology of declarative problem solving in AI where an incremental and modular refinement of the high-level representation with extra domain knowledge can improve and scale the computational performance of the framework.