The mother of all traffic jams could be on the horizon, and researchers at the Johns Hopkins Applied Physics Laboratory (APL) in Laurel, Maryland, are working to create a crossing guard that’s up to the challenge.

The Federal Aviation Administration (FAA) estimates that by 2024, there will be more than two million commercial and recreational drones — also known as unmanned aerial vehicles, or UAVs — flying in the National Airspace System. Lanier Watkins, a senior cybersecurity research scientist in APL’s Asymmetric Operations Sector, and his team propose addressing this boom by deploying a set of autonomous algorithms capable of preserving airspace safety and directing traffic on this unprecedented scale.

Their initial research was funded by the Johns Hopkins Institute for Assured Autonomy, run jointly by APL and the Whiting School of Engineering.

The low-altitude airspace presents an opportunity, enabling routine autonomous operations such as search and rescue, precision agriculture, critical infrastructure inspection and medical delivery, according to Josh Silbermann, APL project manager of Airborne Collision Avoidance System X for small UAVs (ACAS sXu). But that opportunity is not without obstacles.

“Constructing a low-altitude traffic management framework in the spirit of our traditional air traffic control system is a challenge. The various networked components, which may have been independently designed, must work safely together,” said Silbermann. “Their interactions must be assured to avoid failure modes, especially as the individual decision processes become increasingly complex. Lanier’s team is taking this long view toward a more complex, dense and autonomous UTM [Unmanned Aerial System (UAS) Traffic Management] ecosystem.”

This work is an offshoot of APL’s longstanding collaboration with the FAA on surveillance and collision-avoidance standards, which date back multiple decades, primarily in creating standards for large manned aircraft. More recently, that work has extended to creating systems, such as ACAS sXu, designed to facilitate the entry of increasing numbers of unmanned craft into airspace traditionally dominated by manned craft. Systems populated entirely by UAVs present an even greater challenge, according to Silbermann — particularly as commercial entities enter the space, each with their own proprietary UAV designs and navigation systems.

“All these systems haven’t been designed together, and as they start to be used together, there may be some emergent behavior that’s problematic,” said Silbermann. “You don’t have to view the solution in terms of autonomous systems — the FAA is not taking that viewpoint — but we’re trying to take the long view. If this concept is going to be successful in industrial applications, that will require the kind of autonomous systems Lanier envisions.”

Watkins and his colleagues — Silbermann, computer scientist Sebastian Zanlongo, systems engineer Randy Sleight, software engineer Tyler Young and data scientist Nick Sarfaraz — published their research on this approach to UTM and presented it virtually at the Institute of Electrical and Electronics Engineers (IEEE) International Conference on Communications in June. Louis Whitcomb, a specialist in robotics systems and a professor at the Johns Hopkins University Whiting School of Engineering, where Watkins chairs the computer science and cybersecurity programs, is co-principal investigator on the project.

Their research has implications for many other domains beyond the present context.

“Instead of UAVs at airports, we could apply this to ships arriving in and leaving a port; to a battle scenario with ships, submarine vehicles and aerial vehicles, where you have to distinguish between friend and foe; or a battlefield with swarms of drones in the air,” Watkins said. “We can extend this work to any situation where there’s a high volume of traffic that has to be precisely managed in real time.”

Another key piece of the work moving forward is building additional levels of safety into the system, such as the ability to validate decisions and make on-the-fly adjustments to compensate for malfunctioning components.

“In a real system, there’s a lot of imperfection — drones may stop conforming to flight plans; GPS may return incorrect coordinates,” Watkins said. “There’s a lot of noise in a real-world system that our model has to be able to account for.”