A Microeconomic Approach to Intelligent Resource Sharing in Multiagent Systems
Abstract
We have analyzed characteristics of sharable resources and developed techniques for intelligently sharing resources specifically, communication channels among agents in multiagent systems. Our techniques allow agents to nearly optimize their communication behavior in a self-organizing and distributed fashion, involving the use of a microeconomic pricing system based on economic laws of supply and demand and trading among agents in real-time. Our analyses are based on three measures of performance: fairness of resource allocation, waiting time for resources, and utilization of resources. Our initial analysis indicates that fairness and utilization are conflicting, in that the best utilization with a fair allocation is equivalent to the worst utilization with an unfair resource allocation, assuming the allocation policy is statically defined. To strike a balance in performance, we have developed mechanisms that establish an artificial economy, where agents can dynamically reallocate goods (resource access) using a competitive market pricing mechanism. However, unlike more common market-oriented methods, our approach does not demand convergence to equilibrium, but permits more rapid, heuristic trading, leading to near optimal performance where both buyers and sellers of resources can benefit. Our studies show that agents employing our mechanisms can dramatically improve utilization while still providing fair access to the resources.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1995
- Accession Number
- ADA462235
Entities
People
- Edmund H. Durfee
- Jaeho Lee
Organizations
- University of Michigan