A critical capability needed by unmanned aircraft systems (UAS), both in military and in commercial use, is to sense and avoid other aircraft. Such a capability constitutes a last line of safety when air traffic control is unable to keep aircraft away from each other. This is partly because UAS cannot practically be constrained to assigned corridors but also because these aircraft must fit into current airspace operations which do not require general aviation to be controlled at the level that would provide deconfliction. Specifically, we are interested in meeting the need for sense and avoid at a size, weight, power, and cost (SWAP-C) scale applicable to ORBs. These vehicles will need to operate in general air traffic scenarios where full-scale vehicles (e.g., helicopters and general aviation fixed-wing aircraft) are present. While some component technologies (radar sensors and computer vision systems) have been proposed, no commercial solution is available at this time. In Phase I we validated the hypothesis that a combination of radar and computer vision can meet the safety and SWAP-C requirements for UAS to operate among existing traffic in the national airspace. We note that while recent small SWAP radars measure range to targets well, they are not sufficiently precise in determining direction. They are also prone to false positives from terrain. On the other hand, computer vision detects targets in air traffic with very good angular direction but in many cases is unable to determine range to the target. In Phase I we built several sensor payloads and collected data in multiple flight tests with representative air traffic, ranges, and closing speeds. We showed that the complementary behaviors of the radar and vision sensing modalities improved overall precision. Additionally, the tests confirmed that their dissimilarities help the system eliminate background detection clutter normally present in each individual modality. The system operated up to distances and types of aircraft directly applicable to ORBs. In Phase II we will further improve the methodology by using radars with longer detection ranges and scale optical fields of view for vision input. We will work closely with our Department of the Air Force stakeholder, the Air Force Research Lab Sensor Exploitation Applications Branch, to further define requirements, conduct flight testing, and report results. The core of the Phase II effort will be to further develop and adapt the detection, data fusion, and state estimation algorithms. In Phase I the Federal Aviation Administration ACAS X avoidance policy was utilized to analyze detection results and observe correct avoidance advisory. In Phase II we will demonstrate that this policy can be used for avoidance of the relevant aircraft without losing guarantees. The technical results of Phase II will have wide applicability in keeping ORBs and other aircraft safe from collisions with uncooperative aircraft in the national airspace.
