IJCAI 2005 Workshop on
Planning and Learning in A Priori Unknown or Dynamic Domains

Many testbeds in Artificial Intelligence are a priori known and static. However, as Artificial Intelligence systems are applied to more realistic problems, they need to be able to cope with a priori unknown and dynamic domains. The state of an environment, for instance, can change over time due to the presence of other (cooperative or competitive) agents or exogenous events. In such domains, an agent cannot predict the effects of its actions with certainty and thus needs to be able to plan under uncertainty with its current domain model as well as learn improved domain models, often in real time. Example domains include real-time games, mobile robotics, space applications, supply-chain management, and decision support for crisis management.

We solicit papers on all aspects of planning and learning in a priori unknown and dynamic domains, including but not limited to:

• finding robust plans and enhancing the robustness of existing plans,
• time-dependent planning,
• trading off planning time and plan quality,
• interleaving planning and plan execution,
• fast replanning and plan adaptation,
• planning when and what to sense,
• multi-agent planning and learning,
• life-long learning of domain models,
• acting robustly in the presence of unknown but important state features or discovering them
• reinforcement learning, and
• learning with very few training data.

The one-day workshop will be held on August 1, 2005 as part of the IJCAI 2005 workshop series.

If you are interested in submitting a paper or participating in the workshop, please send an email to ijcai05-pludd-request@cs.ualberta.ca with the line:

subscribe

in the body of the message. This will add you to the mailing list for the workshop. The mailing list is moderated against spam abuse.

Manuela Veloso, Carnegie Mellon University
Michael Van Lent, University of Southern California

April 4, 2005: paper submissions
May 2, 2005: accept/reject decisions
May 23, 2005: submission of camera-ready papers

Paper submissions must be send in pdf format to bulitko@ualberta.ca with the subject line "IJCAI Workshop Submission". Note that we cannot accept paper submissions in formats other than pdf. Paper submissions and final papers must not exceed 6 pages. The format is NOT double-blind, but otherwise has to follow the IJCAI 2005 formatting guidelines.

When additional information becomes available, it will be posted on this website.

The most important part of the schedule is the poster session, where everyone (including the plenary speakers) will present their posters. There are two parts to the poster session so that every author does not need to be at their poster all the time and gets the chance to walk around and look at the other posters. Everybody needs to bring a poster (likely of A1 size). Authors whose name is listed in the poster advertisement session, also need to bring 3 slides or so for their 3 minute poster ad. We encourage them to bring overhead slides to avoid the delays of switching laptops every three minutes.

• 9:00 Opening remarks [Koenig and Bulitko]
• 9:15 Invited Talk
  • State Estimation, Action Selection, and Learning in Multi-Robot Adversarial Domains [Veloso]
• 9:45 Plenary Session 1 (15 minutes talk + 5 minutes discussions each)
  • An Algorithm better than AO*? [Bonet and Geffner]
  • Improving Convergence of LRTA*(k) [Hernandez and Meseguer]
  • Modeling and Planning in Large State and Action Spaces [Kochenderfer and Hayes}
• 10:45 Coffee Break
• 11:00 Plenary Session 2 (15 minutes talk + 5 minutes discussions each)
  • Learning Planning Rules in Noisy Stochastic Worlds [Zettlemoyer, Pasula and Kaelbling]
  • Simulation Methods for Uncertain Decision-Theoretic Planning [Aberdeen and Buffet]
  • Hedged Learning: Regret-Minimization with Learning Experts [Chang and Kaelbling]
• 12:00 Lunch Break
• 1:30 Invited Talk (25 minutes talk + 5 minutes discussion)
  • AI and Learning in Games and Simulators [van Lent]
• 2:00 Poster Advertisements (3 minutes each)
  • Speeding up Learning in Real-Time Search via Automatic State Abstraction [Bulitko, Sturtevant and Kazakevich]
  • Best-First Utility-Guided Search [Ruml, Crawford and Do]
  • Hypothetical Planning [Babaian]
  • Adaptive Partitioning of State Spaces using Decision Graphs for Real-Time Modeling and Planning [Kocherderfer and Hayes]
  • Generating Temporally Contingent Plans [Foss and Onder]
  • Deduction and Exploratory Assessment of Partial Plans [Malec and Nowaczyk]
  • Heuristic Speed-Ups for Learning in Complex Stochastic Environments [Darken]
  • Learning Task Allocation via Multi-Level Policy Gradient Algorithm with Dynamic Learning Rate [Abdallah and Lesser]
  • Fast Reachability Analysis for Uncertain SSPs [Buffet]
  • Risk-Directed Exploration in Reinforcement Learning [Law, Coggan, Precup and Ratitch]
  • Apprenticeship Learning for Initial Value Functions in Reinforcement Learning [Maire and Bulitko]
  • Bayesian Models of Nonstationary Markov decision processes [Jong and Stone]
  • Supervised Learning of Options: A Pilot Study [Paduraru and Bulitko]
  • Putting Olfaction into Action: Anchoring Symbols to Sensor Data using Olfaction and Planning [Loutfi, Coradeschi, Kaelsson and Broxvall]
  • Path Planning for Unmanned Underwater Vehicles [Petres and Patron]
  • Using LTL Assumptions to Generate Safe Plans for Partially Known Domains [Albore and Bertoli]
  • Dynamic Domains in Data Production Planning [Golden and Pang]
  • Robust and Opportunistic Planning [Gaines, Estlin, Chouinard, Fisher, Castano, Anderson and Judd]
• 3:15 Poster Session - Part 1 (all posters, including posters for the plenary talks)
• 4:15 Plenary Session 3 (15 minutes talk + 5 minutes discussions each)
  • Using LTL Assumptions to Generate Safe Plans for Partially Known Domains [Albore and Bertoli]
  • Dynamic Domains in Data Production Planning [Golden and Pang]
  • Robust and Opportunistic Planning [Gaines, Estlin, Chouinard, Fisher, Castano, Anderson and Judd]
• 5:15 Poster Session - Part 2 (all posters, including posters for the plenary talks)
• 6:15 IJCAI Reception

Vadim Bulitko, University of Alberta (bulitko@ualberta.ca)
Sven Koenig, University of Southern California

Douglas Aberdeen, The Australian National University (Australia)
Blai Bonet, Universidad Simon Bolivar (Venezuela)
Adi Botea, University of Alberta (Canada)
James Bruce, Carnegie Mellon University (USC)
Greg Calbert, DSTO (Australia)
Alessandro Cimatti, IRST (Italy)
Alan Fern, Oregon State University (USA)
Natalia Gardiol, MIT (USA)
Russ Gayle, University of North Carolina (USA)
Enrico Giunchiglia, Universita di Genova (Italy)
Russ Greiner, University of Alberta (Canada)
Charles Gretton, National ICT Australia (Australia)
Richard Korf, University of California at Los Angeles (USA)
Lihong Li, University of Alberta (Canada)
Michael Littman, Rutgers University (USA)
Yaxin Liu, University of Texas (USA)
Frederic Maire, Queensland University of Technology (Australia)
Mausam, University of Washington (USA)
Martin Mueller, University of Alberta (Canada)
Doina Precup, McGill University (Canada)
Ioannis Refanidis, University of Macedonia (Greece)
Matt Rudary, University of Michigan (USA)
Jonathan Schaeffer, University of Alberta (Canada)
Masashi Shimbo, Nara Institute of Science and Technology (Japan)
Bill Smart, Washinton University in St. Louis (USA)
Trey Smith, Carnegie Mellon University (USA)
Finnegan Southey, University of Alberta (Canada)
Peter Stone, University of Texas at Austin (USA)
Nathan Sturtevant, University of Alberta (Canada)
Rich Sutton, University of Alberta (Canada)
Sylvie Thiebaux, The Australian National University (Australia)
Vincent Vidal, CRIL (France)
Eric Wiewiora, University of California in San Diego (USA)
Martin Zinkevich, Brown University (USA)