Technischer Bericht. In: Proceedings of the 2014 Joint Workshop of Fraunhofer IOSB and Institute for Anthropomatics, Vision and Fusion Laboratory, KIT Scientific Publishing, Karlsruhe, 2015.
The Max-Sum algorithm, an instance of the Generalized Distributive Law family, is known to solve Distributed Constraint Optimization Problems (DCOP) where the summed utility functions of interacting agents are maximized. However, Max-Sum relies on available communication channels between all agents that partake in a utility function. We present a generalization of Max-Sum that solves DCOP exactly in situations where the communication network layout does not match the agents’ utility inter-dependencies.