An efficient method for asynchronous distributed constraint optimization using upper and lower bound

Distributed constraint optimization problem (DCOP) is an important area of study of multi-agents. Recently, a distributed search algorithm for DCOP is proposed. The algorithm enables asynchronous search based on blanch-and-bound and depth first tree. However, the algorithm has overhead of backtracking and re-exploration. In this paper, some effective methods which reduce overhead of former algorithm are proposed. To reduce overhead of backtracking, partial solution which is related with lower/upper bound is considered. And the information of lower bound is sent to upper level nodes, by a shortcut. To reduce re-exploration, leaning of solution is introduced.