Agent Cooperation Can Compensate for Agent Ignorance in Constraint Satisfaction

Abstract

A team of constraint agents with diverse viewpoints can find a solution to a constraint satisfaction problem (CSP) when the individual agents have an incomplete view of the problem. In this paper we present a method of solving constraint satisfaction problems (CSPs) using cooperating constraint agents where each agent has a different representation of a particular CSP. Agents assist one another by exchanging information obtained during preprocessing and as a result improve problem solving efficiency. Unlike previous distributed and multiagent CSP techniques this agent-oriented technique provides fault-tolerance and redundancy. The technique is illustrated using cooperating constraint agents solving logic puzzles. Introduction A team of cooperating constraint-based reasoning agents may be able to solve a constraint satisfaction problem (CSP) even when members of the team have underconstrained representations of the problem. Agents with diverse representations can compen...