For unmanned aircraft systems to safely fly autonomously in increasingly crowded airspace, they must be able to accurately detect and avoid obstacles like trees, power lines, and, critically, other aircraft.
“This is the core problem with unmanned aircraft systems: the ability to see and avoid other aircraft,” said Mark Blanks, the director of the Virginia Tech Mid-Atlantic Aviation Partnership.
The aircraft will rely on control software to detect and avoid obstacles, and that software must be rigorously tested.
“Until someone demonstrates that their algorithms can meet or exceed human capabilities, unmanned aircraft won’t be able to fly beyond an operator’s visual line of sight, and that limits the applications,” said Craig Woolsey, a professor of aerospace and ocean engineering in the College of Engineering and an expert in control systems. “The first phase is collecting realistic data that the community can use to test their algorithms.”
Acquiring that data was the goal of research flights conducted this year at Virginia Tech.
The flights were part of a collaboration involving the Mid-Atlantic Aviation Partnership, Woolsey, and Karl Warnick, a professor of electrical and computer engineering at Brigham Young University.
The team is working to create a database analogous to the ones used in computer vision research, where huge repositories of labeled images are used as test sets for visual-recognition software. This database will provide researchers with information on what potential obstacles, like a telephone pole or a small quadcopter, would look like to an aircraft’s sensors.
“We want to gather all this data in advance so that we can enable people to test their algorithms before they get implemented,” Woolsey said.
To collect that data, the team outfitted a 35-pound fixed-wing unmanned aircraft with an optical camera on each wing and a unique radar system designed in Warnick’s lab.
Flying over a rural test range near the Blacksburg campus, the aircraft collected images and radar signals of both fixed objects and other aircraft from a variety of distances and angles.
Using two types of sensors providing complementary information should allow aircraft control software to more accurately assess a potential hazard.
Optical cameras offer high enough resolution to classify or identify objects, but require significant processing power that drains the aircraft’s battery; they are also less effective under certain lighting and weather conditions. Radar is more robust, and by using a phased array, can provide both the distance to an obstacle and its coordinates. Warnick’s system, which weighs less than half a pound, is the first phased-array radar light enough to be carried by a small unmanned aircraft.
The sensor data from the test flights will be processed and stored in a publicly available database, along with the GPS coordinates corresponding to each data point — showing not only what the sensors detected but also how far away it was.
This data will allow researchers — including Woolsey’s group — to develop and refine their algorithms for aircraft control systems. “We want to start testing our own ideas against images in the database,” Woolsey said.
The project is funded by the National Science Foundation through the Center for Unmanned Aircraft Systems, an Industry/University Cooperative Research Center led by Brigham Young University; the center uses university research to tackle fundamental challenges preventing the integration of unmanned aircraft systems in the national airspace. Woolsey is the director of Virginia Tech’s arm of the center.
“This is an exciting project, because having a database will be a key enabler for unmanned aircraft systems integration,” Blanks said. “It will be phenomenally impactful.”
The Virginia Tech Mid-Atlantic Aviation Partnership, which is headquartered at the Institute for Critical Technology and Applied Science, runs one of only six national test sites for unmanned aircraft systems designated by the Federal Aviation Administration.
The test site’s pioneering work across disciplines, from agriculture to journalism to emergency management, has positioned the university as a leader in unmanned aircraft systems research. Current research topics include flight beyond visual line of sight, flight operations over people, unmanned aircraft system airworthiness certification, air traffic management, remote sensing and payload development support, and airspace integration.