Decentralized Markov Decision Processes with Event-Driven Interactions

Decentralized MDPs provide a powerful formal framework for planning in multi-agent systems, but the complexity of the model limits its usefulness. We study in this paper a class of DEC-MDPs that restricts the interactions between the agents to a structured, event-driven dependency. These dependencie...

Full description

Saved in:
Bibliographic Details
Published in:Autonomous Agents and Multiagent Systems: Proceedings, 3rd International Joint Conference, New York City, New York, 2004. pp. 302 - 309
Main Authors: Becker, Raphen, Zilberstein, Shlomo, Lesser, Victor
Format: Conference Proceeding
Language:English
Published: Washington, DC, USA IEEE Computer Society 19.07.2004
IEEE
Series:ACM Conferences
Subjects:
Communication system control
Computer science
Computing methodologies > Artificial intelligence > Distributed artificial intelligence > Multi-agent systems
Computing methodologies > Artificial intelligence > Planning and scheduling
Distributed control
Mathematics of computing > Probability and statistics > Probabilistic representations > Markov networks
Mathematics of computing > Probability and statistics > Stochastic processes > Markov processes
Multiagent systems
Permission
Power system modeling
Power system planning
Process planning
State-space methods
Stochastic processes
Theory of computation > Theory and algorithms for application domains > Machine learning theory > Markov decision processes
ISBN:9781581138641, 1581138644
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Decentralized MDPs provide a powerful formal framework for planning in multi-agent systems, but the complexity of the model limits its usefulness. We study in this paper a class of DEC-MDPs that restricts the interactions between the agents to a structured, event-driven dependency. These dependencies can model locking a shared resource or temporal enabling constraints, both of which arise frequently in practice. The complexity of this class of problems is shown to be no harder than exponential in the number of states and doubly exponential in the number of dependencies. Since the number of dependencies is much smaller than the number of states for many problems, this is significantly better than the doubly exponential (in the state space) complexity of DEC-MDPs. We also demonstrate how an algorithm we previously developed can be used to solve problems in this class both optimally and approximately. Experimental work indicates that this solution technique is significantly faster than a naive policy search approach.
Bibliography:SourceType-Conference Papers & Proceedings-1
ObjectType-Conference Paper-1
content type line 25
ISBN:9781581138641
1581138644
DOI:10.5555/1018409.1018761