An Investigation of Discovery-Based Learning in the Route Planning Domain
Abstract
This thesis presents MAVERICK, a Discovery-Based learning (DBL) system designed to learn maneuvers in the route planning domain. DBL was originally designed to learn in domains for which little domain knowledge exists. This thesis proposes using it in domains for which knowledge exists, but the acquisition of this knowledge is difficult or time-consuming because of the knowledge acquisition bottleneck. The operation of the DBL process in MAVERICK was investigated to determine the potential utility of such a system to a real- world Air Force problem in the domain of aircraft route planning. MAVERICK worked well in its limited domain, and demonstrated several positive aspects of the DBL process, specifically robustness, flexibility, and graceful degradation. Some negative aspects of the process were also encountered during this research; MAVERICK demonstrated a pronounced tendency towards unpredictability, both in its operation and its development. This likely precludes DBL from application to critical systems, however the positive aspects suggest DBL can have potential utility to other systems.... Artificial intelligence, Machine learning, Discovery-Based learning, Route planning, Inductive inference.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1992
- Accession Number
- ADA259141
Entities
People
- Freeman A. Kilpatrick Jr.
Organizations
- Air Force Institute of Technology