Constraint-Based Protocols for Distributed Problem Solving (extended abstract)

Distributed Problem Solving} (DPS) decomposes problems into subproblems to be solved by interacting, cooperative software agents. Thus, DPS is suitable for modeling the solving of problems characterized by many interdependencies among subproblems, both in the context of parallel and distributed architectures. Concurrent Constraint Programming (CCP) is a powerful framework to support DPS, since it provides a direct implementation of the notion of shared, reusable, incrementally refined information, where constraints can declaratively implement both local problem solving as well as exchange of information, and hence DPS, among agents. To optimally solve DPS problems with CCP we need however capabilities for tuning the protocol for constraint communication at the execution level, according to such factors as the specific instance of the DPS problem and the characteristics of the underlying system architecture. In our research we introduce two general protocols for DPS, the Request-Subrequest Protocol and the Local Caching Protocol.