Autonomous Cross-Country Navigation Using Stereo Vision

Abstract

This paper reports on an autonomous robot designed for operation in uncharted outdoor environments. To accomplish this we have developed a robot vehicle equipped with wide field-of-view stereo vision and high accuracy inertial navigation. The software resident onboard the vehicle uses binoculor cameras, mounted in a novel configuration, to produce dense range maps of the environment. This data is used to guide the vehicle around obstacles and the incremental map that is developed from range data taken from the moving vehicle is used to dynamically alter the route to the final destination. Additionally, we have investigated the possibility of merging range data from scanning lasers and stereo vision in order to achieve high performance and high bandwidth.