Layering-ADOPT: ADOPT with Layering Boundary

Abstract

ADOPT and BnB-ADOPT are the two important search-based complete algorithms to solve Distributed Constraint Optimization Problem (DCOP). However, a large number of solution reconstructions appear in ADOPT and sub-optimal branch can't be promptly pruned in BnB-ADOPT. A layering DCOP algorithm to combine ADOPT and BnB-ADOPT is proposed to remedy their weakness, named Layering-ADOPT. In Layering-ADOPT, a layering boundary is introduced to divide all agents into two classes, one performing ADOPT and another for BnB-ADOPT. This paper presents a rule to get a layering boundary and a new strategy to realize the joint between ADOPT and BnB-ADOPT. In the experiment, Layering-ADOPT and ADOPT as well as BnB-ADOPT are compared on random DCOPs. The results show that Layering-ADOPT is superior to ADOPT and BnB-ADOPT on the test problems.