An object-oriented And/Or graph inference engine

Most AI problems can be specified by first order logic. Such a specification can be mapped onto the nodes of an And/Or graph. Traditional And/Or graph inference engines either in depth-first or breadth-first fashion, take a procedural view on And/Or graphs. For a given problem, they either decompose it into a set of procedures and embed goals as consecutive procedure calls, or translate it to a passive data structure and carry out a search procedure by using various stacks. In this paper, we propose an object-oriented inference engine for And/Or graph-based reasoning. This approach takes advantage of object-oriented programming techniques, that is, every component of an And/Or graph is treated as an object. As objects are self-maintained entities cooperating through messages, our approach offers great potential for exploring different interesting techniques, such as parameter hashing, intelligent backtracking and concurrency. The proposed inference engine has been used in the implementation of LU-Prolog