Ten years from now, we could be in for a new sort of traffic jam. In the near future, posits a new study led by two members of the Johns Hopkins Institute for Assured Autonomy, some 65,000 drones—weighing up to 55 pounds each, most of them programmed to operate without a pilot—will be “taking off, flying, or landing” per hour in what is so far largely uncontrolled lower U.S. airspace.
Drone technology is advancing so rapidly that the Federal Aviation Administration forecasts that by 2027 close to 1 million commercial drones in total will be in operation, doing everything from delivering our pizzas and packages to assisting in emergency response.
The FAA is developing a plan to manage the upsurge. But, as the Hopkins study notes, it will be the first time the agency provides the equivalent of air-traffic control in that airspace, which is 400 feet or below. And at the moment, it does not have the human capacity to handle the job. So co-authors Lanier Watkins, principal professional staff at the Johns Hopkins Applied Physics Laboratory, and Louis Whitcomb, professor of mechanical engineering at the Whiting School of Engineering, conducted a study to evaluate how we can ensure the safe control of drone traffic by replacing human-involved processes with autonomous decision-making.
Whitcomb makes clear that while no one is anticipating any kind of doomsday scenario, the level of traffic congestion will demand more coordination than humans can provide.
“It’s not that aircraft will be falling out of the skies,” he says. “But if we want to start using automated aircraft at scale, then the traffic management will have to be automated as well.”
Over the past several years, the tech industry, individual companies, and government agencies have invested heavily in testing drone technology that does not require pilot assistance. And some, like Amazon Prime Air and Zipline, have already launched operations with the delivery of medical supplies to remote areas. Zipline, co-founded by Keenan Wyrobek, Engr ’03, has “two bases in Rwanda that can reach anywhere in the country,” Whitcomb says. “That’s hundreds of deliveries per day, thousands per week.”
The study that Whitcomb and Watkins helmed piggybacks on work that Watkins’ APL team has conducted for 10-plus years. The new study, Watkins says, goes a few steps further, using a 3D simulator with noisy sensors that anticipates the autonomous aspect of future drones and applies AI algorithms (and AI assurance) to various inevitable and unexpected occurrences, such as mountainous terrain and skyscrapers, last-minute changes to flight schedules, and the sudden appearance of a flock of birds or a rogue drone.
“We also added a procedure to analyze the decisions made by the autonomy [or nonhuman] algorithms,” Watkins says. “All these things together — that’s where the novelty comes in.”
For every scenario, the team programmed algorithms to avoid collisions, a job too overwhelming for a human crew alone. But the efficacy of the regulatory system the FAA will eventually put in place demands a team effort. “It will not be human air-traffic controllers,” Whitcomb says. “It will be a computerized system that is not operated by the FAA. They envision commercial providers doing this. The idea is to automate traffic management and to do it at a scale that we can barely even conceive of today.”
While the FAA plan is still a “work in progress,” as Whitcomb puts it, both he and Watkins have faith in the agency, which already deals with continual air traffic increases.
“I’m sure there will be growing pains associated with this,” Whitcomb acknowledges. But air transport is a “giant engine of our economy,” he says, making the increase inevitable. The challenge for the FAA is maintaining balance. “It’s like medicine,” Whitcomb adds. “You want safety and you want efficacy.”