TOYON RESEARCH CORPORATION — Department of Defense SBIR Phase I: N191-003

Description

The success and ubiquity of the Global Positioning System (GPS) have created a single point of failure for applications requiring precise time and position. While inertial measurement unit (IMU) technology continues to improve, the performance of micro-electro-mechanical systems (MEMS) technology still has not achieved the type of stability required of navigation-grade instruments necessary for autonomous aviation. Vision-based navigation aiding offers an attractive alternative to GPS that is immune to both jamming and accidental interference. However, traditional approaches to vision-based navigation aiding rely on extraction and matching of keypoints between image frames or on computation of optical flow. Those techniques work well over, e.g., vegetation-covered land, but neither technique is robust while flying over 'texture deserts', such as water, snow, or sand. We propose the use of alternative vision-based techniques, including calculating heading directly from skylight polarization of visible light, and attaining celestial navigation fixes during daytime or nighttime via InGaAs-based SWIR sensing of the sky. Finally, because vision alone cannot provide an all-weather solution like GPS, we additionally discuss exploitation of signals-of-opportunity (SoOp). These alternative vision-based approaches, along with information gleaned from SoOp, are optimally combined in a multi-sensor fusion framework to maintain the accuracy of the navigation solution.