Workshop on Multi-Agent Sequential Decision Making in Uncertain Domains(MSDM) 2006

Held in conjunction with the Fifth International Joint Conference on Autonomous Agents and Multi-Agent Systems (AAMAS), Future University-Hakodate, 9th May, 2006

Important Dates

Overview

Sequential decision making under uncertainty is the problem an agent faces when it tries to maximize its performance through interacting with its environment (and possibly other agents) based upon its observations of the world. Single-agent decision-theoretic approaches to this problem have centered around two primary models, the Markov Decision Problem (MDP) and the Partially Observable Markov Decision Problem (POMDP), depending on whether the agent's knowledge about the world is complete or partial.

These mathematically rigorous models have been used very successfully in single-agent systems so it is only natural to apply them to systems with many agents. Just as in single-agent decision-theoretic work, the decision-theoretic multi-agent community has focused on two kinds of models: i) where each agent has complete knowledge about the state of the world, and ii) where each agent has partial (and potentially different) knowledge about the state of the world.

The high computational complexity of finding optimal solutions in these multi-agent models has been a significant barrier. Much of the work in this area relates to addressing this complexity through exploiting problem structure like locality of interaction, decomposition of reward and independence between the agents, and through approximate algorithms that converge to a local optimum instead of a global optimum.

The purpose of this workshop is to bring together researchers in the field of sequential decision-making in stochastic multi-agent systems to present and discuss promising new work, to discuss the relationships between the various models in use, and to establish important directions and goals for further research and collaboration. This workshop will strive to develop consensus within the community on benchmarks and evaluation methodology in order to contrast the alternative approaches and models, and to study the tradeoffs associated with the use of each. Furthermore, we will discuss the creation of online problem sets for testing the various algorithms to facilitate comparison.

Topics

The workshop will address a range of topics relating to new and existing models of multi-agent systems (i.e. MMDP, Dec-MDP, Dec-POMDP, Dec-MDP_Com, MTDP, COM-MTDP, R-MTDP, E-MTDP, EMT, I-POMDP, POSG, POIPSG, ND-POMDP, TI-Dec_MDP) including:

• Relationships between the models and their assumptions
• Algorithms for policy generation and coordination
• Comparisons of algorithms
• Distributed vs. centralized planning
• Online vs. offline planning
• Communication during policy generation
• Communication decisions during execution
• Techniques for scaling problems
• Identifying subclasses of problems and their complexity
• Cooperative and competitive agent systems
• Theoretical and empirical results
• Benchmarks and evaluation methodologies for comparing different approaches

Submission Procedure

Authors are encouraged to submit papers up to 15 pages in length in the standard LaTeX Article format (12 pt font). Submissions should be sent to ranjit.nair@honeywell.com or raphen@cs.umass.edu in PostScript or PDF form. Each submission will be reviewed by at least two Program Committee members.

Organizing Committee

Ranjit Nair
Knowledge Services Group, Honeywell Laboratories
3660 Technology Dr, #2703, Minneapolis MN 55418, USA
Phone: +1(612) 951-7198
Fax: +1(612) 951 7438
ranjit.nair@honeywell.com
http://teamcore.usc.edu/nair

Raphen Becker
Department of Computer Science, University of Massachusetts Amherst
140 Governor's Drive, Amherst, MA 01003, USA
Phone: +1(413)545-1985
raphen@cs.umass.edu
http://www.cs.umass.edu/~raphen

Program Committee

Schedule

Each paper has 25 minutes for both presentation and discussion. We recommend saving the last five minutes of the slot for questions and discussion (20 min presentation, 5 min questions).