AAAI 2006 Spring Symposium on Distributed Plan and Schedule Management

 

Program (as of March 23)

 

Monday, March 27, 9:00-10:30

 

Session:  Plan and Schedule Flexibility

How can distributed plans/schedules be configured to avoid, correct for, exploit, and/or tolerate asynchronously arising local (environment or objective) deviations, and how can patterns of local deviations be diagnosed and responded to in a distributed manner?

 

An Examination of Criticality-Sensitive Approaches to Coordination

Pedro Szekely, Rajiv T. Maheswaran, Robert Neches, Craig M. Rogers and Romeo Sanchez

 

Cooperative Team Plan: Planning, Execution, and Replanning

Olivier Bonnet-Torres and Catherine Tessier

 

Discussant(s): Blazej Bulka and/or Marie desJardins

Position Paper: Integrating Top-Down Planning with a Bottom-Up Approach that Learns to Locally Combine Actions

 

Monday, March 27, 11:00-12:30

 

Session:  Distributed Constraint Reasoning

What principled techniques from distributed constraint reasoning are suited to defining and solving the distributed plan/schedule management problem? 

 

Multiply-Constrained DCOP for Distributed Planning and Scheduling

Emma Bowring, Milind Tambe, and Makoto Yokoo

 

An Any-space Algorithm for Distributed Constraint Optimization

Anton Chechetka and Katia Sycara

 

Discussant(s): Carla P. Gomes, Willem-Jan van Hoeve, and/or Bart Selman

Position Paper: Constraint Programming for Distributed Planning and Scheduling

 

Monday, March 27, 2:00-3:30

 

Session:  Distributed Planning

How can efficient techniques for robust conditional/probabilistic planning, for plan monitoring and repair, for continual plan refinement and replanning, etc., be gainfully employed in distributed plan/schedule management? 

 

GPGP - A Domain Independent Implementation

John Phelps

 

Lateral and Hierarchical Partial Centralization for Distributed Coordination and Scheduling of Complex Hierarchical Task Networks

Mark Sims, Hala Mostafa, Bryan Horling, Haizheng Zhang, Victor Lesser, Daniel Corkill, and John Phelps

 

Discussant(s): Mathijs de Weerdt and/or Cees Witteveen

Position Paper: Multiagent Planning: Problem Properties that Matter

 

Monday, March 27, 4:00-5:30

 

Session:  Open and Heterogeneous Systems

What are the implications of bringing together distributed heterogeneous agents, whose planning and scheduling concerns are at very different temporal grain sizes or planning horizons, whose knowledge and beliefs could be incomplete or inconsistent with each other, or whose preferences and objectives might not be entirely aligned? How must distributed planning and scheduling algorithms change to adapt to open world situations in which the timing of events or activities are externally imposed and must be worked around? 

 

Planning to Explore: Using a Coordinated Multisource Infrastructure to Overcome Present and Future Space Flight Planning Challenges

Edward Balaban, Michael Orosz, Tatiana Kichkaylo, Andre Goforth, Adam Sweet, and Robert Neches

 

Multi-Agent Management of Joint Schedules

Stephen F. Smith, Anthony Gallagher, Terry Zimmerman, Laura Barbulescu, and Zack Rubinstein

 

Discussant(s): Nicoleta Neagu, Klaus Dorer, and/or Monique Calisti

Position Paper: Solving Distributed Delivery Problems with Agent-Based Technologies and Constraint Satisfaction Techniques

 

Tuesday, March 28, 9:00-10:30

 

Session:  Application and Evaluation Issues

What application experiences exist in designing, developing, and deploying distributed plan/schedule management systems, and what can we learn from them? What metrics and testbeds can be used to evaluate plan/schedule management systems?

 

Invited Talk: Human Activity Coordination and the COORDINATORs Program

Tom Wagner

 

Exploring Coordination Properties within Populations of Distributed Agents

Elizabeth Sklar, Martijn Schut, Konrad Diwold, and Simon Parsons

 

Discussant(s): Pauline Berry, Melinda Gervasio, Bart Peintner, Tomás Uribe, and/or Neil Yorke-Smith

Position Paper: Multi-Criteria Evaluation in User-Centric Distributed Scheduling Agents

 

Tuesday, March 28, 11:00-12:30

 

Session:  User-Centered Systems

How do the needs/expectations of human users and the participation of users in the plan/schedule management process affect the design and behavior of the overall system? How is dealing with possible instability in users’ preferences and objectives similar to, and different from, dealing with environmental unpredictability during plan/schedule execution?

 

Challenges in Coordinating Remote Sensing Systems

Robert Morris, Jennifer Dungan, and John Bresina

 

Distributed Interactive Narrative Planning System

Joe Winegarden and R. Michael Young

 

Discussant(s): William Staderman

Position Paper: HCI Issues in Planning and Scheduling Systems

 

Tuesday, March 28, 2:00-3:30

 

Session:  Meta-Level Reasoning

How should agents make meta-level decisions about how to allocate computational (and other) resources to plan/schedule management functions, including how frequently and thoroughly management within, and across, their plans/schedules should be carried out?

 

Other Agents’ Actions as Asynchronous Events

Kurt D. Krebsbach

 

Leveraging Problem Classification in Online Meta-Cognition

Anita Raja, George Alexander, and Verghese Mappillai

 

Discussant(s): David Sarne and/or Barbara Grosz

Position Paper: Timing Interruptions for Better Human-Computer Coordinated Planning

 

Tuesday, March 28, 4:00-5:30

 

Session:  Communication Issues

To what extent can and should communication be effectively utilized in distributed plan/schedule management to formulate, track, and dynamically repair distributed plans and schedules?

 

A Constraint Optimization Framework for Fractured Robot Teams

Mary Koes, Katia Sycara, and Illah Nourbakhsh

 

Adaptive Robotic Communication using Coordination Costs for Improved Trajectory Planning

Avi Rosenfeld, Gal A. Kaminka, and Sarit Kraus

 

Discussant(s): Cees Witteveen and/or Mathijs de Weerdt

Position Paper: Multi-agent Planning for Non-Cooperative Agents

 

Wednesday, March 29, 9:00-10:30

 

Session:  Distributed Policy Formulation

What principled techniques from distributed policy formulation are suited to defining and solving the distributed plan/schedule management problem? 

 

Exploiting Locality of Interaction in Networked Distributed POMDPs

Yoonheui Kim, Ranjit Nair, Pradeep Varakantham, Milind Tambe, and Makoto Yokoo

 

Coordinated Plan Management Using Multiagent MDPs

David Musliner, Edmund Durfee, Robert Goldman, Mark Boddy, Jianhui Wu, and Dmitri Dolgov

 

Discussant(s): Yuqing Tang and/or Simon Parsons

Position Paper: Using Argumentation-Based Dialogues for Distributed Plan Management

 

Wednesday, March 29, 11:00-12:30

 

Session:  Distributed Task Assignment and Scheduling

What principled techniques from distributed task definition, assignment, and scheduling are suited to defining and solving the distributed plan/schedule management problem? 

 

Deciding Task Schedules for Temporal Planning via Auctions

Alexander Babanov and Maria Gini

 

A Multiagent Task Associated MDP (MTAMDP) Approach

Pierrick Plamondon, Brahim Chaib-draa, and Abder Rezak Benaskeur

 

Discussant(s): Debra Walker

Position Paper: The Role of Deliverable Specification in Automated Process Planning

 

Notes:

 

Each session lasts 90 minutes and involves 2 presentations of no more than 25 minutes each (including time for questions) for full papers appearing in the working notes and the remainder of the time is devoted to discussion led by an author of a position paper.  (Note that invited presentation by Tom might last more than 25 minutes, TBD.)

 

The session’s discussant is encouraged to work with the session’s presenters in forming and framing the discussion.  The question associated with each session is only a starting point for where the concepts and results being presented could lead.

 

The participants in a session have freedom to arrange the timing as they see fit, such as the order of the two presentations, whether some discussion (or context setting) by the discussants will occur before and/or between the presentations, etc.

 

By default, after the two presentations, all authors of the presented papers and all of the discussants who are present could sit up front in a “panel” format, moderated by one of the discussants.

 

To a large extent, the sessions attempt to be designed around a question (generally a variation of a question from  the CFP).  The papers in the session have something to say about that question.  This is not to say that the papers don’t address other questions too, nor that other papers don’t address this question.  That is why we are leaving lots of time for discussion, so that others who have insights into how to address this kind of question have a chance to voice them.

 

Because the discussant(s) has(have) a position paper that might augment or contrast with the presentations, the discussant(s) can feel free to articulate and briefly justify alternative positions, referencing points made in the position paper. Thus, the discussant(s) can use their position paper as one vehicle to promote discussion among the participants, but should not give a formal presentation of the paper.