Redesigning the Sky (WIRED)

Air Repair

Nearly all US flight delays can be traced to the snarl of jets over New York City. How do you squeeze more efficiency out of an archaic air traffic control system? Redesign the sky.

(link) (photo by Jeffrey Milstein)

Inbound JFK. The turns start while you’re still in the clouds. Engines howling, flaps down, the plane lurches and dives, jerky as a taxi in Midtown. Seatback upright and tray table locked, you’re oblivious to the crowded flight paths around you. But the air above New York City is mapped: a dense and nuanced geography nearly as complicated as the city below.

More than 2 million flights pass over the city every year, most traveling to and from the metropolitan area’s three busiest airports: John F. Kennedy, Newark, and LaGuardia. And all that traffic squeezes through a network of aerial routes first laid out for the mail planes of the 1920s. Aircraft are tracked by antiquated, ground-based radar and guided by verbal instructions issued over simplex radios, technology that predates the pocket calculator. The system is extremely safe—no commercial flight has been in a midair collision over the US in 22 years—but, because the Federal Aviation Administration treats each plane as if it were a 2,000-foot-tall, 6- by 6-mile block lumbering through the troposphere, New York is running out of air.

This is a nightmare for New York travelers; delays affect about a third of the area’s flights. The problem also ripples out to create a bigger logjam: Because so many aircraft pass through New York’s airspace, three-quarters of all holdups nationwide can be traced back to that tangled swath of East Coast sky.

Six years ago, Congress green-lit a plan to solve this problem. The Century of Aviation Reauthorization Act calls for a new system, dubbed NextGen, that uses GPS to create a sort of real-time social network in the skies. In theory, it should give pilots the data they need to route themselves—minus the huge safety cushions.

But NextGen needs some serious hardware: roughly $300,000 in new avionics equipment for every cockpit. That’s a lot of peanuts for the struggling airlines. Add to the tab nearly 800 new federally funded ground stations to relay each plane’s location and trajectory to every other plane in the sky and—by the time NextGen finally launches in 2025—the price tag could reach $42 billion. In the meantime, the New York-area skies have seen a huge traffic bump over the past two decades—including a 48 percent increase between 1994 and 2004. So the FAA has set out to coax new efficiency from old technology.

To help reorganize this airspace, the FAA called on Mitre, a Beltway R&D firm that works exclusively for the government. Mitre’s scientists and mathematicians, in cooperation with some of the region’s air traffic controllers, are completely rethinking the flow of aircraft in and out of New York City. Current flight patterns evolved like a rabbit warren, with additions tacked on to an existing architecture. As airports grew busier and airplanes started flying higher and faster, that architecture became increasingly inefficient. The plan, the unfortunately named New York/New Jersey/Philadelphia Metropolitan Area Airspace Redesign, aims to bring order to the air.

Think of it as a redrawn map of the roadways in the sky. While planes used to chug in and out of the city on a few packed roads, the redesign spreads out the aircraft by adding new arrival posts (exit ramps), departure gates (on-ramps), and takeoff headings (streets leading up to the intercity highways). But the biggest move will be making the space for all these additions. Mitre’s proposal is to extend the boundaries of this airborne city into a 31,180-square-mile area that stretches from Philadelphia to Albany to Montauk.

The FAA started implementing the first part of the plan—the new takeoff headings—in December 2007 and should have the full strategy in place by 2012. By then the agencies hope to have reduced delays in New York by an average of three minutes per flight. And in a system as interconnected as the US air traffic network, those few minutes could quickly cascade into hours.

The nine runways at Kennedy, LaGuardia, and Newark—which together would form the nation’s busiest airport—are roughly parallel to one another to accommodate prevailing wind conditions. The result is way too many planes flying in the same general direction. The controllers can override FAA safety regs when the skies are clear and calm, inching planes closer together and relying on the pilots’ sharp eyes to avoid catastrophe. But when it rains and the pilots can’t see each other, the multi-mile buffers return, the airspace overflows, and the line of traffic clogs the skies clear back to LAX. The redesign does not change the official safety separations—that’s for NextGen—but it does try to use every last wisp of sky.

The first plan Mitre tested erased the current flight paths and placed the airports in an idealized box, where arrivals came in at the corners and departures went out over the sides. It didn’t work. All the traffic clustered around one corner and jammed up.

The first plan Mitre tested erased the current flight paths and placed the airports in an idealized box, where arrivals came in at the corners and departures went out over the sides. It didn’t work. All the traffic clustered around one corner and jammed up.

So the team tried routing aircraft over the Atlantic. That didn’t work either: Flight times were longer, the patterns were more complex, and the number of planes the airspace could accommodate decreased. “We don’t have the luxury of saying all the arrivals are going down the Long Island Sound and all the departures are going out over the ocean,” says Steve Kelley, the former New York controller overseeing the redesign for the FAA. “We’d handle about two airplanes an hour.”

These test patterns, which worked poorly enough in ideal conditions, really fell apart when the simulators cranked up the intensity, adding bogeymen like inclement weather (of which the real New York has plenty). The team needed to factor in a controller’s prerogative to make adjustments on the fly. The problem was a lot more complicated than just drawing new lines in the air. So, naturally, Mitre brought in its nuclear physicist.

Joe Hoffman came to Mitre in 1990 to work on command-and-control systems (what about them, precisely, he declines to say). When “nuclear war stopped being so popular,” Hoffman says, he transferred to the wing of the building where the necessary security clearance isn’t quite so high. As the redesign’s chief strategic thinker, his first step was to figure out how to mathematically express the way planes move through New York’s airspace. It wasn’t so difficult: He’d been working with similar equations for years.

Airplanes in flight mimic (to a point) electrons whizzing around in their subatomic orbits. “The mathematics relate,” he says. While a moving object in the terrestrial world can be tracked with four variables—latitude, longitude, speed, and time—an airplane soaring along a flight path adds a fifth—altitude. In Hoffman’s sky—and in the math he uses to describe it—not all of these variables are equal; each one has to be weighted differently.

His calculations showed that some variables could be changed with fewer negative consequences than others. Planes have to keep a certain distance from one another, so latitude and longitude are rigid. Time, not surprisingly, is inelastic. The easiest variables to change are speed and altitude, but if you slow a plane down on the wrong road, you cause a traffic jam. The system needed more roads. Instead of lining up birds one behind another along the same trajectories, he needed to spread them out, maximizing the airspace.

To do this, Mitre recommended integrating the jurisdiction of the New York Tracon with other regional control centers—expanding the low-altitude zone in which all arriving and departing aircraft fly. The biggest backups in the current system happen when a flight transitions from the high-speed, long-distance “en route” highways to the slower, local “terminal” roads—the ones that the Tracon controls. The redesign creates a kind of airborne suburbia, paving the skies far out into what was the countryside. The idea is that the controllers can get planes off the intercity highways sooner, keeping them clear for through-traffic.

Ultimately, they’ll be able to get more planes on the ground per hour by interlacing their runway approaches. And more widely spread patterns for outbound flights will prevent bottlenecks and allow more planes to take off every hour.

The aim of each of these tweaks is to shave off a few seconds here and there to ultimately hit that three-minute goal. Mitre sees it as a classic systems engineering scenario: Little changes snowball into big effects. But the question is, can the controllers handle the extra territory and fuller screens? Mitre put them in a giant simulation facility to find out.

Mitre headquarters in McLean, Virginia, isn’t far from the CIA. It shares a parking lot with Northrop Grumman and an architectural sensibility with Dunder Mifflin. The exterior is tan and glass and crenellated with cameras. The interior is a study in grays. When you check in, the receptionist will inquire, “Is this a classified meeting?” Nearby, a sign warns: No un-encrypted Mitre laptops beyond this point.

Founded in 1958 as an offshoot of MIT’s Lincoln Laboratory, Mitre worked on an air defense network that was a model for Arpanet, the predecessor to the Internet. In the 1960s it figured out how to merge the nation’s airports into a connected web that would become known as the National Airspace System. Since then, it has helped design flight patterns above Los Angeles and Chicago. But neither of these skyscapes was nearly as daunting as New York’s.

Mitre’s primary tool for testing whether its New York plan will work is the Air Traffic Management Laboratory. A sprawling windowless simulation facility, it has a full-size 737 cockpit, a mock airport tower, and an army of radar screens just like in a typical control tower. Technicians can simulate any event possible in the FAA’s world—from a routine takeoff to a crash landing. Visitors—usually angry airline reps—watch the action piped into a conference room.

Hoffman stands inside the darkened lab. A half-dozen controllers gaze at pizza-box-sized monitors filled with pulsing green blips, imaginary planes flying around a fictitious New York City sky. These guys are the real thing: golf-shirted, bug-eyed FAA vets down from the city for a week of experiments and lonely beers in a strange town. With radio triggers gripped firmly in their left hands, they use their rights to punch at keypads, entering the verbal commands they’ve just issued to the pilots passing through their sectors.

But the pilots are not real cockpit jockeys. They’re lab techs trained in air traffic lingo, plugged into headsets, and sitting in front of another bank of monitors in a nearby room. When a controller issues a command, the “pilot” logs it into a program that looks like a spreadsheet. The simulation responds with a corresponding blip on the radar screen.

Today controllers are testing new flight paths above Robbinsville, New Jersey, an area about 50 miles south of Manhattan that currently falls under JFK’s departure path and could soon see traffic from Newark and LaGuardia. The Mitre folks speculate that they can improve the flow of departures out of New York by sending them farther south or west before they peel off toward flyover country. Much of the Garden State’s sky is essentially a giant merge (poor state can’t catch a break) where the westward flows from each NYC airport climb toward cruising altitude. The challenge is to make sure each sector is safely handling as many planes as possible. Or rather, as is humanly possible: For all the high tech toys at Mitre, the success of the redesign depends entirely on the processing power of the guys in the golf shirts—their ability to mentally keep track of the additional planes in their sectors. Mitre’s redesign remains a decidedly analog affair.

The controllers direct planes around the simulated airspace for 45 minutes. Then engineers debrief them about their experience: Did the planes’ transitions out of the Tracon feel right? Were the widened departure gates improving the flow? Could crossing traffic flows be properly “deconflicted,” as the controllers put it? Mitre officially thinks it was a success. Hoffman has his own method of gauging controllers’ stress levels: Check the angles their spines make with the seats of their chairs. At 100-plus degrees—leaning back—the work is easy; straight up, things are getting interesting; once they cross the 90-degree threshold and begin to perch forward, the sky is roiling chaos. Most of the controllers at the simulation never crossed the 90-degree mark.

Three hundred twenty-seven feet above Newark Liberty International Airport, in the windowed cab of the control tower, Hoffman’s stress-assessment metric proves utterly useless: Louis Caggiano likes to stand as he works the hot seat, with departing flights lined up 15-deep on the taxiway. It’s a hazy morning in July, and inbound aircraft are hard to pick out on the horizon, appearing first on a radar screen that hangs from an adjustable track. Each of Caggiano’s outbound flights corresponds to a paper strip in a plastic sleeve. He holds a stack in his hand and fiddles with them as if they were poker chips. After giving the verbal command that sends a plane lumbering down the runway, he time-codes the flight strip with a machine that looks like an electric stapler, then drops the card into a slot in the counter. Bending down like a boxer ducking a punch to get a higher view of the sky, he visually confirms that the plane is up. Then Caggiano launches the next one in line—46 departures that hour.

It may not look like much, but this is the first piece of the redesign in action. Controllers will start testing the new departure gates and arrival posts over the next two years, but they’ve been using the modified takeoff headings for months. All the planes leaving Newark used to depart along fairly similar courses, but the redesign’s “dispersal” headings aim to reduce delays by fanning flights out across the sky. According to FAA rules, planes can stack up on the tarmac as long as they’re not actually touching. But the instant they’re airborne, they must be 3 miles apart—unless they’re moving away from each other. Dispersal headings satisfy that directive and thus increase capacity, allowing the next plane to be cleared for takeoff right away and saving another few seconds that Mitre predicts will add up to hours over the course of a year.

Caggiano isn’t buying it. “They don’t work,” he says with a brevity that suits the simplex radio plugged into his ear. In the lab, the second plane is always ready to go when the first one leaves the ground. But the pilots at Newark aren’t robots, and they don’t move as quickly as the simulation. The airborne flights may be banking according to the new plan, but often the next takeoff isn’t even positioned at the runway centerline yet. Even so, the FAA’s analysis shows that dispersal headings alone are increasing the number of takeoffs by an average of two per hour. (A typical rush hour at Newark sees 40-something departures.) And this is only the first piece of the redesign. Patience, Hoffman says. “The controllers are working on a scale of minutes,” he says. “We’re looking at a scale of years.”

The project may look futile to in-the-moment controllers and plane-spotting activists, but the FAA predicts a savings of 9.7 minutes for planes leaving LaGuardia and 1.3 minutes for planes arriving at Kennedy; they’ll shave 7.3 and 7.1 minutes off Newark’s arrivals and departures, respectively. And if Mitre’s models are right, these efficiencies should ripple out across the system, freeing up runways and air routes from Los Angeles to Denver and back to New York. Whether or not you believe those minutes will really make a difference, one thing is certain: “The airspace is the airspace,” the FAA’s Kelley says. “No one’s going to give us more of it. We just have to use it better.”