Coordinated problem solving through resource sharing in a distributed environment.

Abstract

An important feature in a distributed problem solving system is that the resources of different nodes can be shared through cooperation. In this paper, the generalized partial global planning (GPGP) approach used for multiagent systems is extended by providing a coordination mechanism for resource sharing across nodes. In our framework, multiple conflicting criteria (or objectives) like quality, cost, and duration may be associated with an input task. Preference ratings expressed subjectively may be assigned to each of the criteria. Task assignment in this system, which is a multiobjective decision making problem, is important for the satisfaction of the criteria. It has to be done with imprecise information since the system is dynamic and preference ratings are specified subjectively. A technique for task assignment using the fuzzy set approach is also presented in this paper. Simulation studies for the coordination mechanism and the task assignment have been performed to demonstrate their effectiveness.