Image-based Navigation for Cloud Avoidance (INCA)

Unmanned aerial systems (UAS) such as Global Hawk must exercise due regard for the safety of other aircraft by performing sense and avoid, providing an “equivalent level of safety”, compared to manned aircraft. However, the Air Force frequently needs to fly UAS missions covertly using emissions control procedures, precluding the use of transponder-based systems or radar for collision avoidance. Most passive sensors, however, do not have the ability to “see through” clouds and other visual obscurants, necessitating flights to remain within visual meteorological conditions (VMC). To address this need, we propose a system based on passive sensing to detect and avoid clouds, called Image-based Navigation for Cloud Avoidance (INCA). INCA performs cloud detection in either visible or long wave infrared imagery, using a combination of motion, color, and texture cues. To enable navigation, the system goes beyond simple 2D image detection, and estimates 3D range information using structure from motion. When the current flight path violates VMC minima, a path replanning algorithm based on A* uses the 3D information to compute alternate flight paths. A closed-loop evaluation framework incorporating a flight simulator tests cloud detection performance over a range of operating conditions and tests the efficacy of avoidance maneuvers.