Understanding and Exploiting Hierarchy
Abstract
The project made important progress in the three key areas it set out to address: information content, information flow and information processing, showing that explicit modeling of information, its content and the ways it changes, can provide a powerful means of handling large distributed problems. Among these, the use of generic task description templates greatly improves the agent-tasking process by making explicit the constraints and dependencies between tasks. Such task models allow algorithms to understand potential tradeoffs and identify ways tasks can be modified to suit the changing environment. The technologies and ideas developed during the project have been successfully applied to problems in mission planning and ISR management. In particular, the DEOS system developed under the project offers faster, more flexible solutions than those available using current technologies. Research on the information-processing aspects of process management highlighted several new approaches, particularly exploiting phase transitions. These are naturally occurring "computational cliffs" in problems that represent the point where problems transition from being manageable to be being very difficult to solve. Many important problems fall in this transition region, making the potential payoff of this work very high.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 2002
- Accession Number
- ADA408557
Entities
People
- Brian Drabble
- David W. Etherington
- Matthew L. Ginsberg
Organizations
- University of Oregon