Using Market Mechanisms to Control Agent Allocation in Global Information Systems

Abstract

Multi-agent based global information systems present one of the broadest and most compelling areas for market-based control. Their inherently distributed and heterogeneous nature combined with their vast size makes traditional centralized control methods impractical. We use the ‘invisible hand’ of markets to guide agent decisions without centralized management. While most market-based studies impose auction mechanisms to drive market convergence, in our system the continual adaptation of the ‘information economy’ is an emergent effect, fuelled by competition between autonomous self-interested agents actinf on locally gathered information. We argue that due to market-based competition an environment tailored to consumer requirements will emerge. We focus on agents competing using mobility to adapt their location in response to consumer demand. In this modern version of the data allocation problem, we demonstrate that even simple decentralized mobility strategies can produce agent allocations comparable to centralized heuristic algorithms. Furthermore, we demonstrate that where agents implement competing strategies, due to market-based competition the agents that best satisfyconsumer requirements will populate the emergent environment.