Multi-Perspective Planning -
Using Domain Constraints to Support the Coordinated Development of Plans
Austin Tate, Jeff Dalton, John Levine, Steve Polyak and Gerhard Wickler
Artificial Intelligence Applications
Institute
University of Edinburgh
80 South Bridge, Edinburgh EH1 1HN, UK
Tel: UK (+44) 131 650 2732
Fax: UK (+44) 131 650 6513
E-mail: oplan@ed.ac.uk
O-Plan WWW URL: http://www.aiai.ed.ac.uk/project/oplan/
An important planning capability is the generation and refinement of multiple
Courses of Action (COAs) to respond to a developing crisis requiring military
intervention. This proposal addresses two key areas of importance to the
military planning community which also pose significant challenges for the
AI planning community:
- generation of multiple qualitatively different courses of action
dependent upon alternative assumptions concerning the emerging crisis;
- support for mixed initiative plan development, manipulation and use
dependent upon different assumptions concerning the level of response
to be made and the levels of assets to be assigned.
These generic tasks are vital to support initial development of COAs and their
subsequent refinement, analysis, selection and use in crisis planning
situations for Non-combatant Evacuation Operations (NEOs) and Air Campaigns -
the target domains for the ARPA/Rome Laboratory Planning Initiative (ARPI).
The work addresses the development of plans from a number of
different perspectives. It will further a number of promising and
related developments in task specification, knowledge rich plan
representation, explicit workflow management and plan constraint manipulation
to coordinate user and system input to the planning process. The 4 main
technical themes of the project are:
- Task Assignment, Analysis and Elaboration;
- Domain, Plan and Task Representation;
- Mixed Initiative Plan Development;
- Manipulating Plans as Sets of Constraints
An innovative combination of these techniques will be demonstrated on a
series of realistic problems related to NEOs and Air Campaign Planning
using domain materials within the ARPI suited to research demonstration
and evaluation.
The existing O-Plan framework within the ARPI forms a basis for the
work. O-Plan can make use of domain constraint knowledge to direct
its search for plans. O-Plan has now reached ``critical mass'' where
experience being gained with realistic applications points the way towards an
approach which can address the two challenges described above.
The project includes working relationships and links with other
researchers in each of the 4 main technical themes. Additional ARPI-related
programme work concerned with domain-relevant evaluation and tier 2 project
involvement are allowed for via a responsive programmatic work package within
the project plan.
Page maintained by oplan@ed.ac.uk,
Last updated: Thu Apr 27 11:07:10 2000