Connecting AFB_BJ+ with soft arc consistency

Abstract

The Distributed Constraint Optimization Problem (DCOP) is a major and powerful paradigm for modeling and solving problems in multiagent coordination. AFB BJ + is one of the most excellent algorithms for solving DCOPs. Recently, researchers have shown that including soft arc consistency (AC * ) in DCOP algorithms causes significant improvements in their performance. In this paper, we introduce AFB BJ + -AC * algorithm which connects AFB BJ + with soft arc consistency (AC * ). It relies on pruning non-optimal values of an agent domain, using AC * , and propagating them, without adding other types of messages, for generating other deletions that will likewise be propagated. Our experimental analysis on several benchmarks shows that thanks to AC * , AFB BJ + -AC * improves the basic AFB BJ + .