As the commercial use of drones continues to increase, drone traffic in low-altitude areas below 400 feet is expected to grow significantly in the next few years. Experts predict that by 2027, the United States will have nearly 1 million commercial drone systems used for express delivery, traffic monitoring, emergency rescue and other tasks. The influx of drones into low altitudes will pose serious challenges to aviation safety.
In response to the surge in drone traffic, a research team led by Lanier Watkins and Louis Whitcomb of the Assurance Automation Institute at Hopkins University used artificial intelligence to build a system model to more safely direct drone traffic through a certain degree of automated decision-making, replacing some processes that require manual participation. Their research results were published in the journal Computer.
"We wanted to see if different AI approaches could safely handle the expected scale of drone operations, and it turned out to be feasible," Watkins said. The team used autonomous algorithms to enhance the safety and scalability of drone operations in areas below 400 feet. To verify the safety of drone traffic, the team evaluated the impact of autonomous algorithms in simulated three-dimensional airspace. Their earlier research found that collision avoidance algorithms significantly reduced accident rates. After adding the strategic conflict resolution algorithm, the accident rate is further reduced by controlling traffic time to avoid collisions, and airspace accidents are almost eliminated.
To make the simulation more realistic, the team also added two features to the system. The first is a "noise sensor" to simulate the unpredictability of the real environment and improve the system's adaptability; the second is a "fuzzy inference system" to calculate the risk level of each drone based on multiple factors such as the distance between the drone and obstacles. Watkins and Whitcomb said these methods enable the system to make automated decisions to prevent collisions.
Professor Whitcomb said: "Our research considered a variety of variables, including scenarios where a 'rogue drone' deviates from its intended route. The results are very promising." The team plans to make the simulated environment more comprehensive and realistic by introducing dynamic obstacles such as weather.
Watkins said the research builds on more than 20 years of research by the Hopkins Physics Applications Laboratory focused on improving the safety of the U.S. national airspace system. With the rapid development of commercial drones, using AI and simulation to provide decision support for their traffic management and achieve efficient and safe operation of drone systems is an important direction of current research.