Image-Based Navigation.

This report describes add-on work involving the Passive Navigation Study an investigation to determine if a small, low-flying vehicle can be navigated using passively sensed images of the ground. The seminal work was carried out under contract no. F33615-78-C-1612 from 91578 to 121580, during which a stereo bootstrapping approach for position determination was developed. A landmark subsystem was employed to correct the calculated position of the vehicle after a number of bootstrap iterations. An error analysis of the approach and suggestions for further work were included. In this report, an attempt is made to integrate a set of landmark detector schemes with mission task information to locate the vehicle with respect to known ground coordinates. The navigator can be thought of as a program that processes various subsystem information and makes intelligent decisions to determine vehicle position. An approach is outlined which attempts to simulate a human pilot who navigates by visual correction. In this approach the autonomous aircraft have redundant and independent navigational subsystems and a landmark database. Techniques from the field of artificial intelligence are used to model the subset of human intelligence that pinpoints the vehicles location.