Multiagent SAT (MASSAT): Autonomous Pattern Search in Constrained Domains

In this paper, we present an autonomous pattern search approachto solving Satisfiability Problems (SATs). Our approach is essentially a multiagent system. To solve a SAT problem, we first divide variables into groups, and represent each variable group with an agent. Then, we randomly place each agent onto a position in the correspoding local space which is composed of the domains of the variables that are represented by this agent. Thereafter, all agents will autonomously make searchdecisions guided by some reactive rules in their local spaces until a special pattern (i.e., solution) is found or a time step threshold is reached. Experimental results on some benchmark SAT test-sets have shown that by employing the MASSAT approach, we can obtain performances comparable to those of other popular algorithms.