A study on one distributed CSPS solving algorithm based on concurrent search

Distributed constraint satisfaction problem (DisCSP) has being become to a new research branch of artificial intelligence. This paper improves the variable selecting strategy in concurrent search. At the same time, a method for dynamic variable ordering which is fit for distributed multi-agent environment is proposed to solve DisCSP, and value support proportion is defined to calculate variable ordering. Experiments on randomly generated DisCSP demonstrate that concurrent backtrack search with the proposed method for dynamic variable ordering performs more efficiently than random variable ordering based concurrent backtrack search, and communication overhead is decreases as well.