Real-time stereo and motion integration for navigation

Our purpose in this paper is to describe new techniques for high-resolution range sensing and dynamic vehicle localization. An emphasis of the work is rapid processing, with data redundancy, rather than complex analysis, being used to ensure high quality range estimates. Two aspects of the work are presented here. The first is the use of spatiotemporal consistency filtering to verify range estimates, and the second is the development of a fast registration method that dynamically solves for position and orientation of the platform from stereo range and motion tracking estimates. These methods will enable us to integrate range information over time into a consistent local 3D map.