Animation by 731; Video: Getty Images

Traffic Jam at 400 Feet

NASA and the FAA are working to revolutionize air traffic control for the drone era.

“Please stay clear of the flight line,” warns Keith Hyde, director of U.S. operations for Wing. Safety comes first on these two fenced-off acres at the dead end of Welcome Street in Christiansburg, Va., where Wing has since 2019 been running the first North American drone delivery service. The drones are electric vertical takeoff and landing (eVTOL, pronounced “ev-tol”) aircraft, so instead of a runway, they park on a grid of landing pads that double as charging stations. Three dozen of the pads are arranged on a gravel patch the size of a basketball court, each topped with a QR code large enough for an incoming drone to scan and confirm its touchdown location.

Wing, owned by Alphabet Inc., has no competition for the skies over Christiansburg, a town of 22,000 not far from Virginia Tech, and it operates only in clear, windless weather. Its drones are made of light plastic and polystyrene but still weigh in at 10 pounds because of the controllers, lasers, cameras, and battery packs required to achieve their 12-mile round-trip range. This morning a dozen drones recharge, awaiting orders. The flight line is flanked by 11 shipping containers. The ones labeled C1, C2, and C3 are where the drones “sleep” during off hours. Containers C3, C4, C5, and C6 hold inventory from partners such as Walgreens, a local coffee shop, and an area Girl Scout troop, which relied on Wing to shore up flagging cookie sales during the pandemic.

Featured in Bloomberg Businessweek, July 25, 2022. Subscribe now. Illustration: Saratta Chuengsatiansup for Bloomberg Businessweek

On the perimeter, “merchant success associate” Folake Adeshina, who’s wearing a hard hat, an N95 mask, and a yellow safety vest, waits for an order. Her tablet dings, and she glances at the request: hot coffee. C6 is stocked with carafes, cups, cream, sugar, and stir sticks. Adeshina fills two cups with steaming hot brew. As she works, the pilot in command (PIC), a man identified only as P.J. who’s stationed in C11 behind a computer, chooses which drone will fulfill the mission. The system has already calculated an optimal flight plan, but the Federal Aviation Administration requires a “pilot” for the mission, along with an observer who’s surveying the operation from a nearby hill. “The PIC could probably be replaced with a decision algorithm,” Hyde says as P.J. smiles behind the window of his container. (Hyde is no longer working for Wing.)

It’s not P.J.’s fault that he’s human, nor that humans have caused most accidents throughout the history of air travel. The FAA has been trying to solve for the fallibility of pilots and air traffic controllers since the dawn of commercial aviation. In one study from 1965, the agency identified the vital characteristics of a successful air traffic controller. Among those criteria:

● Steady attention to work and conduct
● Ability to organize and make the most effective use of time, equipment, and information currently available
● Rate of continued improvement
● Emotional stability under pressure

Stalwart though they may be, P.J. and other humans in the aviation system can’t compete on those grounds with a finely tuned algorithm. So from Wing’s perspective, the sooner humans are only pouring the coffee, the safer and more profitable its operations will be.

For now, P.J.’s job is secure. The drone buzzes to life, its 12 rotors lifting it 20 feet from its QR code. It hovers over to C6, where Adeshina weighs the coffee order and tucks everything into a box, adding a small carton of water for ballast. A line with a red plastic hook unspools from the drone’s belly. Box meets hook, and the drone draws the package snug to its chest, rises to 180 feet, converts to cruising mode, and flies off at 70 mph. Somewhere in Christiansburg, the recipient might be tracking the coffee’s progress on Wing’s mobile app. Upon approaching its preapproved drop area, the drone will slow to a hover, descend to 20 feet, let down its hook, and set its cargo gingerly on the ground before buzzing home to recharge.

Lifting off for a systems check in Christiansburg, Va. Video: Cameron Davidson for Bloomberg Businessweek

This carefully controlled process is a crucial step toward building one of the trickiest elements of a successful drone delivery enterprise: public confidence. Crashes must become a statistical near-impossibility before the FAA will certify drones such as these to operate widely and before people will tolerate the increase in air traffic. Even the most zealous drone advocates admit that an ill-timed crash could set the industry back years.

If key regulatory approvals click into place, though, the global market for commercial drone services could increase from $20.8 billion in 2021 to about $500 billion by 2028, according to an analysis by Grand View Research Inc. Designers are preparing unmanned aircraft to assist with long- and short-range package delivery, agriculture, infrastructure inspection, fire and police services, and more. Passenger flights might not be far behind.

The companies jockeying for airspace promise that their drones can help decongest cities, decarbonize transportation, and make it unnecessary to build and maintain roads in undeveloped areas. Yet despite advances in electric propulsion and battery technology, the rosiest projections for drone flight have been slow to materialize. Almost 10 years have passed since Jeff Bezos demonstrated Inc.’s drone delivery project on 60 Minutes, but the technology remains at the proof of concept stage, at least in the US. The hardware, software, and infrastructure demands are complex enough, but the most intense challenge might well be managing the quantum leap in air traffic. Today about 200,000 planes and helicopters are in active use in the US, with around 5,400 flying during peak hours; the potential number of drones that could take to the sky is exponentially higher.

With an eye to this future, NASA and the FAA have been collaborating since 2017 to research and write new rules for US airspace. The aim is to accommodate millions of unmanned drone operations below 400 feet, as well as next-generation light passenger aircraft that mix human and algorithmic piloting at altitudes as high as 5,000 feet. The project includes a dizzying array of specifications that cover airframes, sensors, communications hardware and bandwidth, vertiports (i.e., airports for drones), and, most important, traffic. The aircraft will need rules that humans and computers alike can follow intuitively and that help determine who’s accountable for accidents.

The drone network will require unprecedented cooperation between government and industry. True believers worry that regulators are lagging behind advances in aviation technology, pointing to the strides other countries are making toward integrating drones into their economies. Some critics, on the other hand, are concerned that industry is pushing regulators to move too fast, underestimating the difficulty of safely introducing millions of new aircraft to the national airspace. Done wrong, lives could be lost. Done right, the system will win the public’s trust, create massive new industries, and mark the biggest evolution in U.S. transportation since the interstate highway system.

Simulated flight path onscreen at Ames

“People forget that the ‘A’ in NASA stands for ‘aeronautics’ ”

Photographer: Balazs Gardi for Bloomberg Businessweek

In 2010, Mark Moore, a veteran NASA engineer, published a white paper showing that distributed electrical propulsion—which replaces bulky combustion engines with smaller, highly efficient electric motors—could transform how aircraft take off, maneuver, and land. “Back in those days, everyone looked at these electric motors and controllers and batteries and said, ‘They’re way too heavy,’  ” Moore recalls. But the skeptics were running numbers based on tired assumptions about distance, weight, and payload. Most traditional passenger and cargo aircraft are designed to carry the maximum amount of weight the farthest possible distance. That design in turn determines the infrastructure required on the ground: regional, national, and international hubs built on routes planned and managed by human air traffic controllers.

Moore’s vision for eVTOL aircraft was smaller payloads flying shorter distances. In the decade after his paper was published, advances in electric motors, battery capacity, and computing power made that vision viable. In 2017 he left NASA to advise Uber Elevate on its air taxi ambitions. “Once you develop this new technology set, you completely change the rules for how aircraft are designed,” he says.

This new technology set also means changing the rules for how air traffic is managed. Today most congestion occurs 10,000 feet above ground within a few nautical miles of major airports, in what’s called Class B airspace. With the help of air traffic controllers, pilots carefully share this airspace until they reach the relative safety of Class A airspace (above 18,000 feet) and ultimately cruising altitude (generally between 33,000 feet and 42,000 feet), where there’s far less potential to harm people or property or to collide with another aircraft.

In the world of eVTOL aircraft, by contrast, thousands of delivery drones could be flying under 400 feet in Class G airspace, a small, currently uncontrolled stratum. Right above that, hundreds of larger passenger drones, some autonomous, some piloted, could be operating in Class B airspace, zipping along routes typically occupied by helicopters. Pilots such as P.J. could be supervising dozens of these operations at once, with computers handling most moment-to-moment decisions.

The FAA saw trouble looming in 2005, when unmanned recreational drones first entered the national airspace. It would be the agency’s responsibility, in conjunction with local, state, and other federal policymakers, to determine the precise regulations and infrastructure required for the safe operation of these proliferating vehicles. Items up for debate: whether drones followed existing transportation routes such as railroad tracks and highways; whether they’d be free to travel through private airspace; what times, altitudes, and decibel levels they’d be required to observe. The sheer number of potential drone flights also introduced privacy and surveillance concerns. Would the aircraft be allowed to capture still images, video, or audio? And how long could that data be stored?

Integrating drones with the air traffic control system required vast aviation expertise and computational prowess. Fortunately, that was the signature blend of the NASA Ames Research Center, a moonbase-inspired campus sandwiched between Moffett Field and the Googleplex in Silicon Valley. “People forget that the ‘A’ in NASA stands for ‘Aeronautics,’  ” says Joey Rios, chief technologist in the aviation systems division at Ames. Engineers at this complex of wind tunnels, laboratories, and supercomputers have pushed every boundary of flight since it was established in 1939 to hone military aircraft. They’ve fail-tested war blimps and fighter planes, subjected astronauts to blackout G-forces in experimental aircraft, and designed the parachutes that land rovers on Mars.

Rios, a computer scientist by training, joined NASA’s Unmanned Traffic Management project in 2014, a year before the agency formally teamed up with the FAA to work on the problem. Much of this group’s work involves crunching real-time and theoretical flight data to create models and 3D simulations. “It’s like an untouched ball of clay,” Rios says. “Everyone has a view of what it should look like when we’re done. But we have to go through that process of peeling away the parts and actually coming away with something beautiful at the end.”

For drones to be commercially useful, they must first be certified for flight beyond their operators’ visual line of sight, which requires proof that they can fly safely even if they lose power, drop connectivity, or encounter something weird. In 2016, NASA began testing the aircraft in increasingly dense, uncontrolled airspace at six sites throughout the US. First, test pilots flew drones in rural areas, in anticipation of their assisting with agriculture, infrastructure inspection, and fire suppression and other dangerous jobs. With the airspace almost empty, Rios and his colleagues focused on designing basic rules for things such as scheduling or establishing geographical boundaries, trying to make the guidelines flexible enough to accommodate more complex environments. “One of the hardest things was making sure we weren’t making design decisions that pruned off potentially very valuable operational architectures in the future,” he says.

From there, testing progressed to exurbs and small towns, where risks to life and property were greater, and then to the suburban areas where drone delivery could become commonplace. That was when the technical challenges began intersecting with questions of trust. “Now you’re getting into more of public perception,” Rios says. “You know, ‘There’s drones flying over my head.’ How do you begin to assure the system and the vehicles themselves such that folks can have confidence that it’s going to be safe?”

Joseph Rios and other engineers at the NASA Ames Research Center in Mountain View, Calif.
Joseph Rios and other engineers at the NASA Ames Research Center in Mountain View, Calif. Photographer: Balazs Gardi for Bloomberg Businessweek

Between rounds of testing, he and the team fed flight data to the “hyperwall,” a supercomputer-powered visualization system that allows researchers at Ames to simulate flight plans in 3D environments. They’ve organized the sky into color-coded virtual highways and lanes—yellow, red, blue, and green—each with particular considerations and rules for altitude, velocity, and right of way. The Class G airspace under 400 feet, where the bulk of drone traffic will be, is far from commercial traffic, but it’s close to some sensitive infrastructure. Power transmission lines and bridges are vulnerable to being hit. On school grounds, security and privacy are obviously of utmost concern. Prison officials worry that drones could drop contraband over walls. Between 400 feet and 5,000 feet, the altitude ideal for passenger drones and air taxis, most aircraft would begin by using existing helicopter routes, expanding only when the system is proven to be safe. All the lower-altitude traffic still needs to cooperate with existing air traffic flying to and from Class B airspace above 10,000 feet.

To complicate matters, companies are experimenting with a variety of sizes, shapes, and design features. NASA and the FAA have been working closely with Boeing, Airbus, Amazon, FedEx, police chiefs, firefighters, and other public and private partners. They’ve conducted thousands of hours of flights, passing along data on connectivity rates, landing accuracy, flight performance, and more, to help government agencies and manufacturers refine their regulations and designs.

The ultimate test for these aircraft was the “urban canyon” of a densely populated metropolis, where hundreds of drones could theoretically be delivering sushi, aiding firefighters, and shooting wedding videos simultaneously. Would household Wi-Fi devices interfere with aircraft signals? How would drones respond to the wind tunnels and updrafts between high-rises? With more demand on bandwidth, would they retain their connectivity? The testing began in Reno, Nev., where researchers first simulated flights by strapping sensors to the bed of a pickup truck and driving below the routes. Only when they were confident it was safe did the drones take to the air.

One of the foundational tasks of air traffic management is known as strategic deconfliction—essentially, planning routes in space and time so vehicles don’t collide. Traditionally, this problem is solved by air traffic controllers with help from radar and software. Then, once the aircraft are aloft, they have to be prepared for tactical deconfliction, in which pilots avoid a collision, also assisted by radar and software. For designers of unmanned aircraft, tactical deconfliction is particularly vexing. It’s one thing to build a vehicle that can identify its own position in relation to the ground. It’s even more difficult to design sensors that can identify other small aircraft in an uncontrolled environment with wind, rain, birds, kites, and other variables. “We found that aircraft operating even within 300 feet of each other were experiencing dramatically different weather,” Rios says.

Strategic and tactical deconfliction become vastly more complicated as air traffic density increases and users summon aircraft on demand. Frat-house burrito deliveries would need to make way for Grandpa’s prescription refill, which would need to hold for a kidney en route for transplantation. In certain emergency scenarios—medical evacuations, electrical storms, terror threats—all drones would need to be grounded. The most challenging circumstances to plan for, though, are everyday “off-nominal scenarios”: unexpected reroutes, missed approaches, obstructions on the vertiport, unstable landings, uncooperative vehicles in the airspace. Such scenarios are difficult for human pilots to negotiate in open airspace; they become potentially catastrophic in congested space, where a mistake can incite a chain reaction of failures or collisions.

To evaluate the risks, researchers with the Mid-Atlantic Aviation Partnership, a collaboration among leading universities and the FAA, have slammed drones into buildings, cars, and dummies at a test site on the outskirts of Christiansburg. They’re measuring the kinetic energy on impact, trying to determine, among other things, at what speed a drone of a given weight might penetrate a wall, shatter a windshield, lacerate skin, or inflict a severe injury. The data will inform standards for aircraft variables such as weight, construction materials, battery capacity, maximum speed, altitude, range, and payload. These standards will in turn influence the infrastructure needs of the entire system.

Rios Photographer: Balazs Gardi for Bloomberg Businessweek

While Rios and his colleagues continue gathering and modeling the data, smaller nations including Finland and the United Arab Emirates have already approved commercial drone flights. The UK launched its first commercial drone corridor in 2021 with plans to expand it into a “drone superhighway” by 2024, linking towns along a 164-mile span. And earlier this year Chinese regulators outlined certification requirements for autonomous flight systems. The US is behind in some respects, but the federated network it’s planning would be built to scale up fast. Much like today’s cellphone network providers, it would encompass a web of smaller private networks, which would share vital information in real time so aircraft could plan routes, detect and avoid one another, and prioritize traffic. Operators and network service providers would share enough flight data to keep the sky safe, but not so much that they jeopardize privacy or competition.

If the network proves safe for cargo deliveries, drone companies and investors hope, people will soon feel comfortable becoming the cargo themselves, hopping in bigger, fancier aircraft to get across town. They’re pouring in billions of dollars in anticipation of that day.

When cars exit Hollister Municipal Airport, a two-runway public facility in California’s Central Valley that serves as a base for firefighting missions, they’re met with a sign: PILOTS: DID YOU CLOSE YOUR FLIGHT PLAN?

Wisk Aero LLC, a joint project of Boeing Co. and the Larry Page aviation venture Kitty Hawk Corp., is trying to get around the age-old problem of pilots forgetting to file their paperwork by developing an artificial-intelligence-driven flight control system that’s reliable enough to be trusted with passengers. On the opposite side of the airfield, in a hangar surrounded by vegetable fields, Wisk is conducting test flights of Cora, an autonomous air taxi designed to fly safely and tirelessly—and always remember its flight plans. “Over 80% of aviation accidents are human error,” says Wisk Chief Executive Officer Gary Gysin on the morning of a Cora test flight. That figure seems damning of human pilots, though it’s also the case that many fateful decisions are inextricably linked to environmental or technological factors. No matter. Gysin, the former CEO of Liquid Robotics Oil & Gas, a Boeing subsidiary focused on autonomous undersea mining, exudes confidence in machines: “We know that self-flying will be safer.”

Whereas competitors such as Archer Aviation Inc. are seeking to launch services that will involve human pilots at first, Wisk is betting that an AI-driven product will ultimately be safer and more profitable. In its world, a (human) greeter at the vertiport will help you into your seat for a safety briefing, but once you’re up in the air, it will be just you, your fellow passengers, and a headset connected to ground control as the aircraft flies you to your destination. The company signed an agreement last year with aircraft-sharing service Blade Urban Air Mobility Inc. to provide and operate 30 autonomous air taxis on its US routes. The goal is to deploy them in late 2024—part of a larger plan to manufacture a fleet capable of operating 14 million flights annually in 20 cities worldwide. First, though, Wisk will need to submit its sixth-generation, four-seat vehicle to the FAA for certification, with the aim of having the country’s first operational autonomous passenger aircraft.

Until regulatory standards for such vehicles have been published, it’s testing, testing, and more testing, feeding Cora terabytes of flight data so it can learn to avoid the kinds of mishaps that have dogged autonomous driving efforts on the ground. On this day, Gysin and his team will oversee a flight from across an airstrip dotted with roaming chickens. Back in the hangar, chief test pilot Raymond Schreiner, the FAA-mandated human in command, will take Cora through a series of exercises designed to hone midair turns. Or, rather, Schreiner will watch Cora take itself through a series of exercises.

“From a social licensing perspective, is this acceptable? Is the noise level OK? Do you mind seeing these things in the sky?”

Video: Balazs Gardi for Bloomberg Businessweek

On the tarmac, the canary-yellow multicopter the size of a taxicab, equipped with 12 white rotors, spins to life. Cora lifts off with a whir and hovers at 33 feet, then the rotors retract, and it zooms down the airstrip powered by a tail propeller. According to the FAA, Cora can’t fly beyond the sightline of its humans, so Wisk staffers chase it haplessly down the runway in a Nissan Leaf. Before reaching Highway 101, where earthbound commuters go about their business, Cora stops, hovers, and executes a delicate 180-degree turn. “Those pedal turns are really hard,” Gysin says. “We have experienced pilots who say it’s really difficult to keep it so stable.”

Cora whizzes back to its starting point to repeat the exercise, and the Leaf pulls a U-turn, struggling to catch up. Over the years, the team has taken it through various failure scenarios, pushing the hardware and software to their limits and testing the drone’s emergency parachute. Later this week, Schreiner will take Cora through pirouette and roller-coaster maneuvers, “to really work the fans.” Every exercise makes the computer more responsive to changes and the aircraft less likely to crash. Ultimately the company will need to test Cora’s visibility, connectivity, and reliability in every conceivable flight scenario, but before it can do that, it will need new directives from the FAA. In the meantime, Wisk is doing additional testing in New Zealand, where looser regulations allow it to fly longer, more complex missions beyond the sight of a human observer.

The company is also working there on building social acceptance for its services, surveying locals about their fears and aspirations for the new technology and conducting public flight demonstrations. It’s even hosting field trips for Maori schoolchildren at its hangar, aiming to stir the imaginations of future engineers. “It’s a blueprint for how we have to do this in any city and any country where we launch,” Gysin says. “From a social licensing perspective, is this acceptable? Is the noise level OK? Do you mind seeing these things in the sky?”

To shore up public confidence in the US, NASA and the FAA are planning a national campaign consisting of industry working groups, flight tests and demonstrations, and community engagement events. Throughout, NASA will gather opinion data that policymakers and standards bodies can combine with technical requirements to set local, state, and federal rules. “These vehicles are flying at lower altitudes and landing in areas that aircraft typically have not landed before,” says Shivanjli Sharma, an aerospace engineer who leads Advanced Air Mobility research at NASA. “Noise is about more than just frequency and decibel level. Those pieces are really important for us to understand at a much deeper level before we can integrate these operations more tightly into our everyday lives.”

A Wing delivery takes flight in Christiansburg.
A Wing delivery takes flight in Christiansburg. Photographer:  Cameron Davidson for Bloomberg Businessweek

Of course, the core component for social acceptance is safety. It seems inevitable that drones will crash, like computers crash, like phones crash, like planes crash. And disaster can come in many forms. Drones can drop weapons and traffic drugs. They can fly over personal airspace and violate civil liberties. They can weaken already frayed social nets by pushing a contactless, on-demand culture that draws people further into their bubbles. However pervasive drones become, the system that guides them will need to be flexible enough to accommodate futures as yet unimagined, or else it will be doomed to the same dangers, inefficiencies, and inequities of our ground transportation networks.

In Christiansburg, Wing’s pilot project turned into a lifeline for residents in the early weeks of the pandemic, and after Covid-19 restrictions eased, an affinity for drone delivery remained. A peer-reviewed study from Virginia Tech showed that 89% of respondents in Christiansburg had used the service or planned to soon, even as many identified a common constellation of concerns: noise, privacy, potential job loss—and, of course, crashes. Three-quarters of those who said they were bothered by the noise still liked the idea of the service.

Christiansburg Middle School librarian Kelly Passek became one of Wing’s most loyal clients during the pandemic, using its drones to deliver library books to homebound kids—something that continued after the lockdowns. “Some kids get a book a day,” she says.

On a bright spring morning near the end of the 2021 school year, she leads her seventh and eighth grade tech and education students on a virtual field trip to the Wing nest. On the grid of a videoconference call, an operations associate shows the children around while Wing marketing rep Jacob Demmitt talks about the company and its origin story. He tells them about an elderly couple in Christiansburg who’d been having difficulty getting out to shop because one person had suffered a broken leg and the other had poor vision in one eye. Wing was there to help.

“We believe the sky is this vast resource that is untapped,” Demmitt says, “and if we use the sky responsibly, we can help society.”

Passek demonstrates how the students might order goods in the future, choosing an Adult Snack Pack from Walgreens: cheddar popcorn, a Snickers bar, a Frappuccino.

A Wing delivery arrives at its destination.
A Wing delivery arrives at its destination. Photographer:  Cameron Davidson for Bloomberg Businessweek

“How many orders do you get per day?” a student asks.

“We want to keep our trade secrets to ourselves, so we don’t help our competition,” Demmitt says.

On the videoconference there’s a rare delay in order processing because P.J. is due for a break. Soon a fresh pilot takes the seat, and the Wing rep resumes his play-by-play. The tablet dings, and the merchant success associate weighs the order.

“How hard was the coding?” another student asks.

“Very hard,” Demmitt replies.

At last the aircraft departs its nest, three miles away. The students pepper the Wing rep with questions. Passek steps out into the yard to wait for the drop. You can’t hear it yet, but it’s coming across the cloudless sky.

A student wonders, “What happens when the drone crashes into something?”

Demmitt answers without hesitation: “That’s never happened.”

More On Bloomberg