AIPS-00 Workshop on Decision-Theoretic Planning
April 14, 2000
Breckenridge, Colorado
http://www.cc.gatech.edu/fac/Sven.Koenig/aips-dtp.html
part of the
Fifth International Conference on
Artificial Intelligence Planning and Scheduling Systems
April 14-17, 2000
Decision-theoretic planning combines classical and non-classical
planning techniques from artificial intelligence with decision theory
(including probability theory and utility theory) to provide a more
expressive planning model. Research on decision-theoretic planning has
made significant progress over the past few years. The challenges now
are to understand the advantages and disadvantages of the different
approaches, to study how they can be extended and combined (both with
each other and with classical planning techniques), and to develop
even better approaches.
We use a broad definition of decision-theoretic planning that includes
planning techniques that deal with all types of uncertainty and plan
evaluation. In this workshop, researchers from the areas of "planning"
and "knowledge representation and reasoning" will exchange ideas about
techniques for representing uncertainty, plan generation, plan
evaluation, plan improvement, and acquisition of the necessary
knowledge. Issues include
- - Modeling -- Which representations to use for modeling uncertainty,
including effector and sensor uncertainty, incomplete knowledge of the
current state, exogenous events, other agents, and how the world
operates? How does the nature of the planning problem affect the
choice of representation? Which representations are best suited for
the acquisition of planning knowledge?
- - Objectives -- Which planning objectives to adopt? How to explain
why one plan is better than another or how it could fail? Which
representations to use for planning objectives, and how do these
representations facilitate or hinder plan generation and improvement?
- - Planning -- Which planning paradigms to use in conjunction with the
different representations of uncertainty? Which planning objectives to
adopt? How to plan with specifications of planning problems that are
incomplete due to lack of time or information? When to use qualitative
and when to use quantitative techniques?
- - Plan Execution -- How to combine planning and plan execution, and
how much to plan in advance? How to reason about when and what to
sense? How to acquire missing domain knowledge by either learning or
eliciting models from humans? How to gather and refine domain
knowledge incrementally? When to re-plan? How to adapt the current
plan to changes during plan execution?
- - Scaling -- How to best combine the advantages of different
representations and planning techniques? How to extend individual
techniques to larger domains and additional applications?
The workshop is intended for researchers from both the "planning"
community and the "knowledge representation and reasoning"
community. We are interested in having a good mix of theoretical and
applied researchers, with the common thread being an interest in
decision-theoretic planning. We welcome both researchers who develop
individual techniques for decision-theoretic planning and knowledge
representation and researchers who build complete systems.
Submission Guidelines
- --------------------
Researchers wishing to participate in the workshop may submit either
full-length papers (of at most 8 pages) or 1-2 page position papers in
AAAI conference format. Work being submitted to the AIPS conference
can also be submitted to the workshop. Electronic submissions are
highly preferred, either in postscript or PDF format. Please forward
submissions to [EMAIL PROTECTED] Hard copy submissions my be sent
to:
Sven Koenig
College of Computing
Georgia Institute of Technology
Atlanta, Georgia 30332-0280
USA
Important Dates:
- ----------------
Submission Deadline: December 15, 1999
Acceptance Notifications: January 30, 2000
Deadline for Final Paper: February 28, 2000
Organizing Committee:
- ---------------------
Jim Blythe, ISI and University of Southern California
([EMAIL PROTECTED])
Craig Boutilier, University of Toronto
([EMAIL PROTECTED])
Richard Goodwin (Co-Chair), IBM T.J. Watson Research Center
([EMAIL PROTECTED])
Lloyd Greenwald, Drexel University
([EMAIL PROTECTED])
Peter Haddawy, University of Wisconsin-Milwaukee
([EMAIL PROTECTED])
Sven Koenig (Co-Chair), Georgia Institute of Technology
([EMAIL PROTECTED])
Alexander Kott, Carnegie Group
([EMAIL PROTECTED])
Michael Littman, Duke University
([EMAIL PROTECTED])
Martha Pollack, University of Pittburgh
([EMAIL PROTECTED])
Reid Simmons, Carnegie Mellon University
([EMAIL PROTECTED])
