Assuring Reliability of Autonomous Commercial Drones in the National Airspace

• Main Authors: Litton, Matthew L., Drusinsky, Doron, Michael, James B.
• Format: Conference Proceeding
• Language: English
• Subjects: airspace management; Atmospheric modeling; autonomous systems; Biological system modeling; Computational modeling; Data models; machine learning; Reliability; Sensors; Stakeholders; system safety; uncrewed aerial vehicles
• ISSN: 2577-0993
• DOI: 10.1109/RAMS51492.2024.10457690

The proliferation of uncrewed aerial vehicles, also known as drones, raises concerns about their safe and reliable integration with conventional aircraft. Today, drones are used commercially for wide-ranging applications such as infrastructure inspection, agriculture, and logistics, necessitating their further integration into the airspace ecosystem. Airspace management currently depends on human operators to perform deconfliction and emergency management, but the removal of human operators requires drones to possess autonomous decision-making capabilities for real-time conflict resolution and collision detection. We analyze data from U.S. military-developed detect-and-avoid algorithms to demonstrate the ability to automatically learn lightweight and effective models of collision detection. Then, we discuss how such models can be easily and cheaply developed for deployment on drones as part of autonomous detect-and-avoid systems. Such systems can provide assurances to regulators and other stakeholders about the reliable integration of drones into the national airspace system. (This research is sponsored by the U.S. Department of the Navy.)