Enhancing AFB_BJ+-AC* algorithm

Abstract

Several problems in multi-agent coordination can be modeled using a Distributed Constraint Optimization Problem (DCOP) paradigm, which can be solved using one of the state-of-the-art algorithms that solve DCOPs. The AFB_BJ+-AC* algorithm is one of the newest of these algorithms. It uses soft arc consistency techniques (AC*) to rapidly reach the optimal solution of a DCOP by deleting non-optimal values from each agent domain. This paper enhances the AFB_BJ+-AC* algorithm to surpass its inefficiency in some problems, especially those where the number of deletions is too limited or remains zero, by increasing the ability of soft arc consistency techniques (AC*) to generate deletions. The idea is centered on increasing the value of the zero-arity constraint, used as a bound to delete values, by redistributing the constraint costs between an agent and its neighbors via extension operations, then via the execution of AC*. Our experiments on different benchmarks show that the new improvements make AFB_BJ+-AC* better in terms of communication load and computation effort.