Exploitation of Geographic Information Systems for Vehicular Destination Prediction
Abstract
This research proposes that vehicles in an urban setting subject to persistent Intelligence Surveillance and Reconnaissance will exhibit attributes that make it possible to predict their future position within a time-horizon. GeoSpatial Information Systems obtained from municipal, commercial, or hyperspectral sources may be used to model an urban grid and to make use of graph-theoretic search algorithms that can prune the future state-space of the vehicle's immediate environment. The results are representational polygons inside the GIS environment that constrict around the subject's likely destinations in the urban grid as the stream of observations continue to drive the process. These polygons predict where the subject may travel within the time-horizon, and also where they may not travel, given the assumption that they are moving along an optimal path between two points. It demonstrates a 81 % success rate for predictions carried out during experimentation in the Dayton Ohio area. It further demonstrates a 97 % improvement over predictions made with general motion models for moving vehicles.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2009
- Accession Number
- ADA499514
Entities
People
- Richard T. Muster
Organizations
- Air Force Institute of Technology