The multi variable multi constrained distributed constraint optimization framework

Service coordination in domains involving temporal constraints and duration uncertainty has previously been solved with a greedy algorithm that attempts to satisfy service requests one by one rather than reasoning over their combinations to deal with intractable problems. We propose preprocessing using a model that abstracts away duration uncertainty, so that the simplified problem can be solved faster and suggest which requests should be addressed first based on reward and satisfiability. We define a new generalization of MC-DCOP, called MV-MC-DCOP, and use this as our model. We present a Mixed Integer Linear Program, a Partially Centralized MV-MC-DCOP, and a fully decentralized MVMC-DCOP approach to this preprocessing, and analyze the strengths and weaknesses of each approach.