Reducing the Number of Samples in Distributed Cooperative Solution Method for Resource Supply Networks

Distributed Constraint Optimization Problems (DCOPs) are applied to resource allocation problems in resource supply networks. In previous studies, distributed cooperative solution methods based on feeder trees have been utilized. However, in most cases with resource supply networks, the size of variable´s domains in the problems is very large, since the variables originally take continuous values. This is critical even if the networks are trees because it increases the number of combinations. Therefore, sampling of solutions is necessary to restrict the size of the problems. In this study, we propose methods to reduce the number of samples for resource allocation problems of resource supply networks. To maintain the feasibility with the samples, boundaries for the resource amount and cost values were introduced. With the proposed methods, the size of problems is reduced while the methods keep relatively better feasibility and quality of the solutions.