[Deadline extended to March 5, for an updated list of confirmed invited
speakers see http://www.isr.ist.utl.pt/~mtjspaan/POMDPPractioners/ ]
=============================================================
CALL FOR PAPERS
POMDP Practioners Workshop: solving real-world POMDP problems
=============================================================
Workshop at ICAPS-2010 (co-located with AAMAS-2010)
May 12 or 13, 2010
Toronto, Canada
http://www.isr.ist.utl.pt/~mtjspaan/POMDPPractioners/
Over the past decade, much advancement has been achieved in the field of
decision-theoretic planning under sensing and actuation uncertainty,
i.e., Partially Observable Markov Decision Processes (POMDPs). The
size of domains that POMDP solvers can handle has increased by orders
of magnitude. Solvers developed ten years ago were hardly able to
handle more than 10 states, while modern solvers scale up to models
with millions of states.
The advances in scalable POMDP algorithms have opened the door to many
POMDP applications. Many researchers have successfully applied POMDPs
to their problems, for instance using off-the-shelf software. Many
more acknowledge the applicability of POMDPs to their problems.
This workshop is intended to bring together POMDP practitioners with
researchers developing POMDP algorithms. Thus, we hope to help
practitioners to overcome the difficulties in adopting POMDPs to their
domains, as well as point researchers to the real world challenges.
Some of the questions this workshop will address are:
- Which application domains have successfully adopted POMDPs?
- In which domains POMDPs could be applied in the future?
- What are the crucial components of a problem that make it amenable
(and worthwhile) to a POMDP treatment?
- What off-the-shelf software is currently available?
- How can we tackle modeling questions (e.g., design by hand, learn
from data)?
- How easy is it for non-experts to map a desired goal task into a
reward-cost model representation?
- How can we map planning and scheduling problems to POMDPs?
- What type of structure is required by current solvers to scale up to
real-world problems?
- What types of approximations are acceptable in real-world problems?
Workshop format:
The workshop will consist of a short POMDP tutorial, covering POMDP
basics as well as several key developments in POMDP solvers. Then, we
will present several invited speakers who will discuss case studies of
POMDP applications. The workshop will also contain talks and posters
presenting unsolved or partially solved case studies.
A significant amount of time will be dedicated for informal and
in-depth discussion between POMDP researchers and practitioners during
a poster session.
Call for Contributions:
We invite researchers to submit papers in the standard ICAPS 2010
format (8 page limit). We welcome new and previously published papers
on POMDP applications.
We invite demonstrations of POMDP applications, or applications where
a POMDP would make a valuable addition. For this venue, please submit
a 2-page description of the application and the demonstration.
We also solicit brief (2-page) position statements / extended
abstracts, that discuss the challenges of POMDP applications or the
merits of applying POMDPs to specific applications.
All submissions will either be presented orally and/or as a poster,
depending on scheduling constraints. Submission deadline is March 5,
2010, and should be submitted through the Easy Chair website
(http://www.easychair.org/conferences/?conf=pomdp2010).
Organizing Committee:
Joelle Pineau, McGill University, Canada
Pascal Poupart, University of Waterloo, Canada
Guy Shani, Ben Gurion University of the Negev, Israel
Matthijs Spaan, ISR, Instituto Superior Técnico, Portugal (main
contact)
Jason Williams, AT&T Labs, USA
_______________________________________________
uai mailing list
uai@ENGR.ORST.EDU
https://secure.engr.oregonstate.edu/mailman/listinfo/uai
