POMDP Symposium Home Page
This page contains coordination information on the POMDP symposium.
I'll try to keep it up to date with useful information for
participants and others.
Some important links:
1998 American Association for Artificial Intelligence Fall Symposium
Planning with Partially Observable Markov Decision Processes
Any deployed planning system must be designed to face the fact that
the real world is infused with uncertainty. Partially observable
Markov decision processes (POMDPs) are an elegant way of modeling
uncertainty in sensing and acting, and can be used as the foundation
for robust plan generation and execution. Of course, such
expressiveness and mathematical elegance comes with a price: only the
simplest of planning problems can be easily described and solved
exactly as POMDPs.
However, the last five to ten years have seen advances in our
understanding of POMDP algorithms and heuristics, improvements in the
overall speed of computing, and increases in the demand for computer
support for decision making under uncertainty. Now is the perfect
time to assess the POMDP approach in light of these changes, to figure
out where we stand, and to plot a course for continued research and
development.
The aim of this symposium is to bring together researchers who have
worked in any of several key focus areas in the study of POMDPs:
- mathematical and algorithmic foundations,
- approximations and heuristics,
- factored models,
- reinforcement learning,
- robotic applications,
- other applications,
- extensions and specializations.
Researchers will present their work in short, highly interactive,
presentations, to allow all participants to form their own conception
of the state of the art and the most promising directions for future
research.
Organizing Committee: Michael Littman
(co-chair), Duke University, mlittman@cs.duke.edu; Tony Cassandra
(co-chair), MCC, cassandra@mcc.com; Steve Hanks, U. Washington,
hanks@cs.washington.edu; Leslie Pack Kaelbling, Brown University,
lpk@cs.brown.edu.
Schedule
Friday, October 23
- 9:00 am - 10:30 am Session: Introduction
[30 min] Littman (overview)
[10 min] Introductions
[15 min] Platzman
[15 min] Opening Discussion
- 10:30 am - 11:00 am Break
- 11:00 am - 12:30 pm Session: Mathematical and Algorithmic
Foundations
[20 min] Zhang (overview)
[15 min] Goldsmith, Li, Wells and Sittiner
[15 min] Hansen
[15 min] Madani
[15 min] Zhang and Zhang
- 12:30 pm - 2:00 pm Lunch
- 2:00 pm - 3:30 pm Session: Approximations and Heuristics
[20 min] Hauskrecht (overview)
[15 min] Geffner and Bonet
[15 min] Littman
[15 min] Lusena
[15 min] White
- 3:30 pm - 4:00 pm Break
- 4:00 pm - 5:30 pm Session: Extensions and Specializations
[20 min] Kaelbling (overview, wrap up)
[15 min] Bayer
[15 min] Chawla
[15 min] Engelbrecht
[15 min] Washington
- 6:00 pm - 7:00 pm Reception
Saturday, October 24
- 9:00 am - 10:30 am Session: Reinforcement Learning
[20 min] Singh (overview)
[15 min] Lanzi
[15 min] Singh
[15 min] Pyeatt
[15 min] Discussion and Slack
- 10:30 am - 11:00 am Break
- 11:00 am - 12:30 pm Session: Robotic Applications
[20 min] Mahadevan (overview)
[15 min] Bagnell (et al.?)
[15 min] Roy
[15 min] Touzet
[15 min] Discussion and Slack
- 12:30 pm - 2:00 pm Lunch
- 2:00 pm - 3:30 pm Session: Other Applications
[20 min] Cassandra (overview)
[15 min] Darrell
[15 min] Kemp and Eckerman
[15 min] Mahadevan
[15 min] Hauskrecht
[15 min] Peek
- 3:30 pm - 4:00 pm Break
- 4:00 pm - 5:30 pm Session: Factored Models
[20 min] Slack
[20 min] Boutilier (overview)
[15 min] Boutilier
[15 min] Kaelbling
[15 min] Majercik
- 6:00 pm - 7:30pm Plenary Session
Sunday, October 25
- 9:00 am - 10:30 am Session: Impromptu Presentations
- 10:30 am - 11:00 am Break
- 11:00 am - 12:30 pm Session: Wrap up
- Michael L. Littman, Ph.D.
- Assistant Professor
- Department of Computer Science
- Duke University, Durham, NC 27708-0129
- Office: D209 LSRC
- Phone: 919-660-6537
- Fax: 919-660-6519
- mlittman@cs.duke.edu
-
Last modified: Fri Jul 31 09:18:17 EDT 1998