Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!newsfeed.rice.edu!bcm.tmc.edu!news.msfc.nasa.gov!newsfeed.internetmci.com!tank.news.pipex.net!pipex!peer-news.britain.eu.net!newsfeed.ed.ac.uk!ainews!louisep
From: louisep@aisb.ed.ac.uk (Louise Pryor)
Subject: CFP: Plan execution
Message-ID: <DKM512.IIz@aisb.ed.ac.uk>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Organization: Dept of Artificial Intelligence, Edinburgh University, Scotland
X-Newsreader: TIN [version 1.2 PL2]
Date: Wed, 3 Jan 1996 16:12:37 GMT
Lines: 106


		       AAAI 1996 Fall Symposium

		 Plan execution: Problems and issues

This symposium is part of the 1996 Fall Symposium Series of the
American Association for Artificial Intelligence, to be held in Boston
on 9th - 11th November, 1996. Information about the Symposium Series
is available at URL
http://www.aaai.org/Symposia/Fall/1996/fssparticipation-96.html 

Submissions are due by 15th April 1996 - see below for details.



			Call for Participation

Traditionally, work on planning has concentrated on how plans can be
constructed, but as the planning community has begun to address more
complex, real-world problems, the issues surrounding the execution of
plans have come to the fore.  Systems are being built that construct
plans that are used in domains ranging from robot navigation to image
processing and information retrieval.  Such systems must often be able
to handle actions with duration, simultaneous execution of actions,
plans with conditionals and loops, and plan failure.  These new
demands on an agent require considerable extensions to the classical
model of plans as simple sequences of actions guaranteed to achieve
their goals.

In this symposium we shall address problems associated with executing
plans in real-world domains.  Such domains have many, if not all, of
the following characteristics:

-   Complexity: neither the plan construction system nor the plan
    execution system (if different) can have complete information; 
-   Dynamism: the world can change independently of the plan being
    executed (whether through the actions of other agents or through
    exogenous events);
-   Uncertainty: the results of performing an action often cannot be 
    predicted with certainty;
-   Interruptibility: actions may last over appreciable durations, and
    may be interrupted during their execution;
-   Concurrency: actions and events may occur simultaneously;
-   Changing Objectives: new goals can arise and old goals can become 
    unimportant as time passes;
-   Goal Variability: goals may vary along a spectrum from
    maintenance goals (keep the value of G as close to V as you can)
    to achievement goals (make G true then terminate).

We shall look at the plan execution problem from the point of view of
the system executing the plan by considering two interdependent
aspects: the nature of executable plans and how they should be
executed.

First, does the nature of a plan change when one takes into acccount
the possibility of feedback, failure, and recovery?  How should such a
plan be represented to facilitate its execution?  What information
should it include? Should plans be sketchy or detailed?  An important
issue is how the answers to these questions vary with the capabilities
of the system executing the plan.

Second, we shall consider how plans should be executed.  How should
plan failure be recognized and how should the recovery or repair
process proceed?  Unexpected changes in the world may be either
adverse or benevolent: how can a system exploit run-time opportunities
to improve its performance?  Once again, how does the approach to
these issues depend on the form that the plans take?

Further information on this symposium is available at URL
http://www.dai.ed.ac.uk/staff/personal_pages/louisep/PEsymp/.

Submission Information
----------------------

Potential participants should submit either an extended abstract of
up to 5000 words or a position paper of up to 2 pages.

All extended abstracts, whether describing a working system or more
theoretical in nature, should answer the following questions:

1. Which of the domain characteristics listed above were considered
   especially significant in the work described?
2. Which other characteristics does your domain have, or does your
   theory account for?
3. What are the limitations of your system or theory? Which of the
   domain characteristics would cause problems for it?

In addition, extended abstracts should include worked examples set in
a simulated or actual real-world domain.

Position papers should summarise an approach to an issue or issues in
plan execution, explaining which domain characteristics are considered
especially relevant.

Electronic submissions only will be accepted. Email ASCII (position
papers only) or postscript files to the symposium chair, Louise Pryor
louisep@aisb.ed.ac.uk.


Organising committee
--------------------
R. James Firby, University of Chicago, firby@cs.uchicago.edu; Steve
Hanks, University of Washington, hanks@cs.washington.edu; Louise Pryor
(chair), University of Edinburgh, louisep@aisb.ed.ac.uk; Sam Steel,
University of Essex, sam@essex.ac.uk. 

