For more information, contact Steve Chien.

Description

Some of the issues we will discuss include:

For more suggestions on unifying frameworks and problem encodings, see the home pages for the IJCAI-97 Challenges in Bridging Plan Synthesis Paradigms ( http://rakaposhi.eas.asu.edu/challenge.html) and Propositional Reasoning and Search ( http://www.research.att.com/~kautz/challenge) respectively.

Workshop Format

Attendance and Submission Requirements

Please send plain ASCII text only.

If you have recently submitted a relevant research paper to AIPS or to another AI conference, please include a copy of the abstract. At your option, you may also include an extended abstract (6 page maximum, in text or postscript format) of new work you may wish to discuss. Participants will be invited to submit a final version of their research statement and other supplementary material to be printed in the working notes.

The workshop fee will be minimal for participants attending the main conference, the Fourth International Conference on Artificial Intelligence Planning Systems (AIPS '98). See http://www.cs.cmu.edu/~aips98 for details.

Deadlines

• Submit statement of interest by March 3, 1998.
• Accepted participants will be notified no later than April 20, 1998. If you need earlier notification in order to plan your travel to AIPS, please indicate so on your statement.
• Submit materials for working notes by May 15, 1998.

Organizing Committee

Motivation:

Planning/scheduling systems and agents must often function in highly dynamic and uncertain environments in which objectives, demands, and resources change rapidly due to the influences of the environment, and the actions of friendly, neutral, and hostile agents. Such situations pose a number of challenges to planning/execution systems. They must interleave planning, scheduling, and execution, and must do so in a way that takes account of the passing time. They must form plans and schedules that are robust with respect to changes that can be anticipated as likely to occur. And they must be able to modify previously formed plans and schedules, in compliance with the most recent available information, attempting minimum disruption of earlier plans and still aiming for the most effective possible use of resources and achievement of goals.

Workshops at previous conferences have demonstrated that there is already considerable work in this area. Now, the time is ripe to explore this area in more depth by bringing together different researchers.

Workshop Topics:

• Rigorous formulations of the planning/scheduling/execution problems in dynamic and uncertain environments.
• Ontologies and representations: How useful are the ontologies and representations (of plans, actions, and planning problems) that have been developed for static or certain domains?
• Models, algorithms, and architectures for planning/scheduling in dynamic and uncertain environments:
  • interleaving planning, scheduling, and execution;
  • planning and execution with incomplete and uncertain information;
  • managing limited computational resources;
  • forming plans that are robust with respect to changes;
  • using cost models in the context of planning and execution.
• Learning issues:
  • How can one exploit information from execution failures/successes,
  • How to reuse planning and execution experience, e.g., case-based reasoning?
• Real-world applications and implemented systems:
  • descriptions of real-world applications and lessons learned;
  • real-world requirements for the integration of planning and execution;
  • new promising application domains;
  • evaluation criteria.

Paper Submissions

Number of Participants:

Co-ordinators/ Organizers:

Co-ordinators:

International Organizing Committee:

Time Table

Further Information:

Introduction

• Analysis of knowledge used in planning (domain knowledge, problem-solving knowledge).
• Ontologies and schemas of plans and planning processes, and their use in knowledge acquisition and engineering.
• Reusable and modular planning components.
• Case studies reporting on the knowledge engineering and acquisition process of constructing planners for concrete applications.
• Relations between characteristics of domains (domain features) and planning techniques or planning problem-solving methods (also called assumptions of problem-solving methods).
• Why are some representation techniques (e.g. HTN, ADL-operators, fluents) more natural for particular domains than others?
• Knowledge acquisition tools for planning (including software demos).

Workshop Format

Submission of Papers

Workshop Co-Chairs

• Leliane Nunes de Barros, Lab of Integrated System, University of Sao Paulo, Brazil, leliane@lsi.usp.br
• Richard Benjamins, Artificial Intelligence Research Institute, CSIC, Spain and SWI, University of Amsterdam, The Netherlands, richard@swi.psy.uva.nl
• Yuval Shahar Section on Medical Informatics, Stanford University, USA, shahar@smi.stanford.edu
• Austin Tate, Artificial Intelligence Applications Institute, University of Edinburgh, Scotland, UK, a.tate@ed.ac.uk
• Andre Valente, Information Science Institute, University of Southern California, USA, valente@isi.edu

Contact Person

Program Committee

Important Dates

For more information on the workshop, send an e-mail to leliane@lsi.usp.br

Introduction

Planning has long been a key problem for AI research. Recent technological developments are making it possible to build planning systems in realistic domains that can solve hard problems quickly. Nonetheless, there seems to be a clear need to integrate these automated systems with the abilities of human planners and decision-makers. There will always be problems that are too hard for the automated planners to solve, or for which it is too difficult for the human users to formulate their requirements precisely enough for the automated systems to use effectively. Even in planning systems with more autonomy, issues of tasking, authority and reporting require that these systems interact with humans at some point. Thus having the human ``in the loop'' is both a practical necessity and an intellectual opportunity. The ultimate goal of interactive and collaborative planning systems is for humans and computers working together to be able to solve harder problems than either could solve alone.

Workshop Topics:

• Mixed-initiative planning
• Incremental planning
• Plan visualization
• Plan explanation and exploration
• Architectures for interactive planning systems
• Collaborative environments and mechanisms
• Use of existing planning technology in interactive and collaborative systems
• Advanced I/O modalities for planning (speech, gestures, maps, GUIs, etc.)
• Interleaved planning and execution (online planning)
• Automated support for groups of human decision makers
• Use of planning and plan recognition in collaborative systems
• Integration of planning and information gathering (situation assessment)

Submissions

Submissions should consist of three (3) copies of full-length or position papers and must be received at the address below no later than Tuesday, March 3, 1998:

Dr. George Ferguson
Dept. of Computer Science
734 Computer Studies Bldg.
University of Rochester
Rochester, NY, 14627-0226

Organizers

• Karen Myers myers@ai.sri.com
• Stephen Smith sfs@cs.cmu.edu
• George Ferguson ferguson@cs.rochester.edu