Computational Markets for Decentralization of Complex Time-Dependent Activities

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

This research project investigates the design and analysis of market-based mechanisms for decentralized scheduling. Market price systems are well known to effectively decentralize decisions in idealized situations, but real scheduling problems present constraints and interdependencies that can defeat simple market schemes. In this project, market failures arising from the inability to couple interdependent resources are addressed through more sophisticated market structures that allow traders to express complex contingencies. The resulting system will be analyzed using game-theoretic and evolutionary techniques. Results from this work will lead to novel practical scheduling mechanisms that can be deployed within e-commerce and other distributed contexts. New design and analysis methodologies will demonstrate how to extend the resulting mechanisms to other classes of decentralized allocation problems.