Efficient Handling of Complex Local Problems in Distributed Constraint Optimization
David A. Burke, Kenneth N. Brown · European Conference on Artificial Intelligence · 2006
Many distributed constraint optimisation algorithms require each agent to have a single variable. For agents with multiple variables, a standard approach is to compile the local problem down to a new variable whose domain is the set of all local solutions. We present two modifications to this method, which (i) reduce problem size by removing interchangeable and dominated local solutions, and (ii) speed up search by identifying values that are interchangeable with respect to specific agents. We show that the modifications give orders of magnitude improvement over the basic compilation.