Relational production systems

Abstract

A relational production system (rps) is a general purpose, formal information processing model developed to support research in artificial intelligence and related areas where a conjunction of predicate calculus literals is a convenient state description language. Rps maintains a strong analogy with type O string grammars. It consists of a “situation”, which is a conjunction of literals, and an unordered set of “relational productions”, analogous to type O string productions. These productions cause the replacement of a subset of literals in the situation by other literals, just as type O string productions cause the replacement of substrings by other substrings. Predictably, this system resembles the more empirical, existing knowledge representation systems, particularly STRIPS, while maintaining a mathematical precision and simplicity which allows proof of useful results. Rps is first defined and exercised on some familiar examples. A relational production composition theorem is then formulated and proved. It is demonstrated that without a sponge-like component in the antecedent of a production, composition of two arbitrary productions is in general impossible.