Self-Organized Graph-Based Resource Allocation

Abstract

In many technical systems, such as smart grids, the central issue is to enable multiple devices to solve a resource allocation problem. Because centralized solutions usually struggle with an increasing number of agents, regio-central or completely decentralized mechanisms, which solve the problem in a cooperative manner, are of utmost interest in large-scale systems. In this paper, we present a coalition-based algorithm that allows a multi-agent system to cooperatively solve a single-resource allocation problem. Our approach uses self-organization to dynamically identify groups of agents whose decisions have to be coordinated, while optimistically refraining from coordinating the actions between these coalitions. The basic idea of our algorithm is inspired by the way pressure compensates in gas or fluid pipeline systems. It therefore operates on the basis of an overlay network, i.e., a graph, that defines a topology of possible resource flows as well as resistances in the form of costs of transferring a resource from one agent to another. Throughout this paper, the problem of compensating for imbalances between energy production and consumption in autonomous power management systems serves to illustrate our algorithm and results.