Implementing an automated reasoning system for multi-agent knowledge and time

Abstract

We present an implementation of an automated reasoning system for multi-agent knowledge and time, which can be used to describe multi-agent environments. Our reasoning procedure is based on a so-called semantic method. That is, suppose that a multi-agent environment is given by a multi-agent knowledge and time modal logic formula set U. To see whether a multi-agent knowledge and time modal logic formula F can be derived from U, all formulas in U and the negation of F are first translated, according to the possible worlds semantics, into a semantically equivalent first-order formula set, from which we then derive a set of first-order clauses. Thus, F logically follows from U iff there is a refutation in the translated clause set. Since we have to use some transitive axioms and deal with inequalities when reasoning about the set of translated first-order clauses, we augment a general purpose first-order theorem proof procedure ME (the Model Elimination) [11] with the capabilities of using transitive axioms and dealing with inequalities. Theory resolution is incorporated into our reasoning procedure for using transitive axioms efficiently. We present our implementation and show some experimental results.

This research is supported in part by the Japanese Ministry of Education and the Artificial Intelligence Research Promotion Foundation.