An Efficient Constraint Optimization Method for Large Multiagent Systems

Constraint satisfaction/optimization is a powerful paradigm for solving numerous tasks in distributed AI, like planning, scheduling, resource allocation. A number of different algorithms (essentially all being backtracking-based) have been developed for this framework. However, up to now, distributed algorithms for constraint reasoning (especially optimization) have not been applied to largescale systems due to their prohibitive complexity in terms of number of messages being exchanged. We present in this paper a new, complete method for distributed constraint optimization, based on dynamic programming. Our approach works on arbitrary topologies using a pseudotree arrangement of the problem graph. It requires a linear number of messages, whose maximal size depends on the induced width along the particular pseudotree chosen. Thus, our method is likely to work very well on large but loose problems. We compare our algorithm with existing approaches and present experimental results in two application domains: distributed meeting scheduling and sensor networks. For some problem types we report orders of magnitude less messages, and even the ability to deal with arbitrarily large problems. Our algorithm is formulated for optimization problems, but can be easily applied to satisfaction problems as well.