Route Planning Using Pattern Classification and Search Techniques.
Abstract
A pattern classification system was constructed for the air interdiction mission route planning domain. The system accepts, as input, a list of factors associated with a point which may be overflown enroute to a target. It then returns classification of the desirability of overflying that point. Pilots and weapons system officers from the 89th Tactical Fighter Squadron were used as domain experts. They indicated in a survey what combinations of factors were preferable to fly over. This information was then used as a knowledge base for the pattern classifier. The pattern classification system is coupled with a search based route planner. A crude map scenario is created with an overlay grid of node points. These points represent the search space of the area to be overflown enroute to a target. Two heuristic search strategies are used in determining a route: hill climbing and algorithm A. At each node in the search, a list of possible next nodes is generated. Each of these nodes is fed to the pattern classifier for classification of the desirability of overflying it. The best next node is then determined by considering each node's classification in conjunction with the estimated distance costs associated with traversing that node. The route planner proceeds in selecting nodes and expanding paths until it finds a suitable route to the target. Keywords: Artificial intelligence, Expert systems; Pattern recognition. (Theses).
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1986
- Accession Number
- ADA177664
Entities
People
- Douglas M. Rouse
Organizations
- Air Force Institute of Technology