An autonomous ground explorer utilizing a vision-based approach to indoor navigation

We introduce a vision system based approach to autonomous navigation and mapping in an indoor environment. Our goal is to eventually use a common processing system that relies on using airborne and ground vehicles. This paper presents our initial results for a ground vehicle and demonstrates the potential of the vision based system for a solution that combines airborne vehicles with ground vehicles. The Autonomous Ground Explorer (AGE) described here serves as a test vehicle to validate the algorithms that will be integrated into our aerial vehicle that is under development. The ground explorer uses only vision (three basic web cameras augmented with simple laser pointers) to efficiently move within building corridors. It can estimate distance in three directions by evaluating the location of laser points within its fields of vision. The ground explorer navigates hallways and makes directional decisions based on the following processes: tracking the corridor's visual vanishing point, anticipating intersections using odometry and corner edge detection, identifying and classifying intersections based on distance measurements and vanishing points, and completing controlled turns based on a confidence factor in its location on a calculated route within a simplified topological map. Experimental results show the effectiveness of our approach and the promise of our combined ground/aerial system concept.