An Automated Reasoning System for Multi-Agent Environments

概要

論文の詳細を見る
This paper introduces a formal approach and a practical system for studying reasoning problems in multi-agent environments. In our system, multi-agent knowledge and time, a kind of multi modal logic, is used to describe multi-agent reasoning problems. Thanslating method is applied for implementation. We first define the syntax of our multi-agent knowledge and time and introduce its possible worlds semantics. Then, according to the possible worlds semantics, we give the translation procedure, which can translate an arbitrary multi-agent knowledge and time formula into its corresponding first-order formula. In this way, instead of reasoning about the given multi-agent knowledge and time formula set, we only need reason about the translated first-order formula set. As there exist transitive axioms and inequalities in the translated first-order formula set, a general purpose first-order theorem proffer, ME(Model Elimination), is extended to be able to deal with the transitive axioms and inequalities properly. Theory resolution is incorporated into our system for using transitive axioms efficiently. Experimental results, including the well-known three wise men problem, show that our system appears to be of reasonable efficiency for solving multi-agent reasoning problems.
社団法人人工知能学会の論文
1999-05-01