Vision-Based Motion Sensing for Small Unmanned Ground Vehicles

Unmanned ground vehicles (UGVs) will play an important role in the nation’s next-generation ground force. As these systems become increasingly autonomous there is increasing need for precise awareness of their absolute position and velocity. Autonomous navigation systems rely on such data to be effective. Unfortunately, today’s position and velocity sensing technologies have significant drawbacks when applied to UGVs – encoders can be thrown by slippage, GPS by signal jamming, LIDARS are highly emissive, and IMUs by inherent sensor drift and expense. Visual Odometry (VO) is the process of estimating position and velocity information based on camera sensor data, and overcomes many of the drawbacks of other methods. VO has its own drawbacks, however, and these must be overcome if it is to become a major component in future UGV sensing and navigation systems. In this project, a robust, modular VO capability will be developed. It will overcome difficult VO problems by a combination of techniques, including fusion with IMU data. The module, titled “VOOM”, will be provide precision data and be useful on a wide variety of UGVs from small to large and at a wide range of speeds.