A Scalable Method for Multiagent Constraint Optimization

We present in this paper a new complete method for distributed constraint optimization. This is a utility-propagation method, inspired by the sum-product algorithm (Kschischang et al 2001). The original algorithm requires fixed message sizes, linear memory and linear time in the size of the problem. However, it is correct only for tree-shaped constraint networks. In this paper, we show how to extend that algorithm to arbitrary topologies using a pseudotree arrangement of the problem graph. We compare our algorithm with "standard" backtracking algorithms, and present experimental results. For some problem types we report orders of magnitude less messages, and even the ability to deal with arbitrary large problems. Our algorithm is formulated for optimization problems, but can be easily applied to satisfaction problems as well.