Time was when a uniformed man would close a metal gate, throw a switch, and intone, “Second floor—men's clothing, linens, power tools…” and the carload of people would glide upward. Now each passenger handles the job with a punch of a button and not a hint of white-knuckled hesitation. The first automatic elevator was installed by Otis Elevator Co. in 1924; the things became common in the 1950s.
And back in the day, every train had an “engineer” in the cab of the locomotive. Then robo-trains took over intra-airport service, and in the past decade they have appeared on subway lines in Copenhagen, Detroit, Tokyo, and other cities.
Quietly, automation has taken charge of many other life-and-death functions. It manages white-hot ribbons of steel that shoot through rolling mills. It guides lasers that sculpt the eye and scalpels that excise the prostate gland. It runs oceangoing freighters, the crews of which have shrunk by an order of magnitude in living memory. And, most obviously, it is mastering aerial warfare. Today, the U.S. military trains twice as many ground operators for its unmanned aerial vehicles (UAVs) as pilots for its military jets. Its UAVs started off by flying surveillance missions, then took on ground attack; now they are being readied to move cargo and evacuate wounded soldiers.
In the sphere of commercial flight, too, automation has thinned the cockpit crew from five to just the pilot and copilot, whose jobs it has greatly simplified. Do we even need those two? Many aviation experts think not. “A pilotless airliner is going to come; it's just a question of when,” said James Albaugh, the president and CEO of Boeing Commercial Airlines, in a talk he gave in August at the AIAA Modeling and Simulation Technologies Conference, in Portland, Ore. “You'll see it in freighters first, over water probably, landing very close to the shore.”
Later, when air-traffic control systems rise to the challenge, pilotless planes will carry stuff to your very doorstep. In the fullness of time, they'll carry you.
Still, UAVs have yet to find a place in even the humblest parts of the aviation business—surveying traffic jams, say, or snooping on celebrity weddings. Such work has not yet been approved for routine purposes, even when the aircraft is small and controlled by a human on the ground—a man-machine meld that keeps a pilot in the loop. Why is it taking so long for the pilot to go the way of the elevator man?
Technical obstacles can be cited: Fully automated planes can't yet visually identify nearby planes, and as for the remotely piloted ones, the civilian variety can't communicate with ground stations, because they haven't got enough bandwidth. “The name of your magazine is the underbelly of all the problems,” says Doug Davis, director of the unmanned aircraft program at New Mexico State University, in Las Cruces. "There's not enough frequency spectrum. We have to compete for it with all these mobile phones and other demands."
But technical problems are hardly the entire explanation. The military has proved this point time and again, by trusting automation to do things that few human pilots could do on a fair day and with a following wind. In a stealth aircraft, for instance, software fiddles from moment to moment with the flight-control surfaces; otherwise, the radar-eluding design would make the plane fall like a brick.
And for nearly two decades, automatic landing systems have been able to drop and stop a jet on the fog-shrouded deck of an aircraft carrier that's barely twice as wide and three times as long as the jet's wingspan—and the ship is moving. Meanwhile, the pilot sits in the cockpit, hands folded.
“Look, there's no harder job for a pilot than landing on an aircraft carrier,” says Missy Cummings, a former jet jockey for the U.S. Navy and now an associate professor of aeronautics and astronautics at MIT. “It's what Navy pilots have over those in the Air Force. And when I saw an F-18 land itself on an aircraft carrier, I knew my job was soon going to be over.” That was in 1994. Automation has gotten rather better since then.
In fact, the U.S. Army has now decided that if you can trust the life of a pilot to software, you can trust the life of a wounded soldier, too. It's funding R&D into robotic medical evacuation vehicles. One such system, being developed by Piasecki Aircraft Corp. of Essington, Pa., and Carnegie Mellon University, is being designed around a system of sensors and software meant to launch and land a full-size helicopter on cluttered, unmapped ground and also fly the vehicle at low altitudes. Similar systems are under development in other countries. Israel's Urban Aeronautics is experimenting with a pilotless medevac that uses four rotors shielded by cowling so that they won't get tangled in shrubbery.
Given such advances, pilotless commercial flight is overdue, argues Cummings. Civilian UAVs could easily and profitably be deployed to survey infrastructure and carry cargo, she points out. And there's no reason why software, alone or perhaps in conjunction with a quickly mobilized ground controller, couldn't take over a piloted plane should something happen to both the pilot and the copilot. Already, she notes, an airliner's software typically takes over flight seconds after takeoff, handles the landing—and most of what happens in between. The pilot just “babysits,” she says.
Of course, software that can meet only "most" of aviation's challenges would hardly satisfy the afraid-of-flying landlubber. That's why the pilot is still there, babysitting, until all the remaining kinks have been worked out. None of the problems are so bad as to prevent civilian pilotless planes from ever happening, but they are real, and they will have to be solved.
First is the problem of “sense and avoid”—the requirement that every airplane be able to spot other aircraft and obstacles and get out of their way. It's a hard job for a machine. “Regulators always require that the pilot be able to look out the window and see if he's about to have a collision,” says Rick Prosek, manager of the U.S. Federal Aviation Administration's (FAA) unmanned aircraft program office, which handles the regulations governing where in the civilian airspace UAVs may fly. “There are times when the technology doesn't give you the right answer,” he says, and a human should be there “to say, ‘Hey, I need to intervene.’”
Even UAVs that are remotely piloted need some sense-and-avoid capability, because there isn't enough bandwidth for sending lag-free, high-definition video to ground-based operators. Compare that with military UAVs, which have a network latency of less than half a second, because they link to their ground controllers half a world away via an owned-and-operated satellite system supplemented by other channels, such as undersea cables.
To go beyond remote control to truly autonomous flight, you'd need onboard software to interpret the data from the aircraft's cameras, radars, and other sensors and then to make good decisions. Autonomous planes would also need to play nicely with piloted planes, keeping their distance in the air, hewing to air-traffic controllers' directives, and avoiding fender-benders on the ground.
Here again the military has lessons for the civilian sector. Northrop Grumman has built some sense-and-avoid savvy into the unmanned helicopters and other UAVs it's developing for the U.S. Navy. The automation does, in three dimensions, pretty much what radar-enabled crash-avoidance systems now do in a few models of luxury cars (see IEEE Spectrum, “A Driver's Sixth Sense,” October 2011), and like those systems, it keeps a human being in the loop. The company's upcoming X-47B, a fighter-size unmanned combat air system (UCAS), will take autonomous flight one step further. The test will involve taking off from an aircraft carrier, using GPS to fly a predetermined route, and landing on the carrier under light supervision—a minder, somewhere onboard the ship, who stands ready to take control if necessary.
But that's not quite the same thing as flying in crowded airspace, where even a 1 percent error rate would be intolerably high. Even a tiny civilian UAV will need some sort of sense-and-avoid capability before regulators will let it share the air with people movers. In this case, at least, the danger is small enough and the potential payoff great enough that agencies like the FAA seem ready to allow such small UAVs to fly within carefully delineated spaces. "In 2013 we anticipate we'll publish a final rule that would allow small UAVs to fly in certain parts of the national air space," says the FAA's Prosek. "Then we could benefit from the data they'd collect for us."
Having negotiated a sense-and-avoid situation, a pilotless plane would also need enough brains to maneuver away from an uncomfortably close vehicle, notify air-traffic control of what was happening, and then land safely. These challenges, however, are more like flying by the book than by the seat of your pants, and flying by the book is what computers do best.
Take the Global Hawk, a UAV developed by Northrop Grumman for maritime surveillance operations. It's able to fly itself home and land on its own if it loses its satellite link with its ground station. “The plane ‘pings’ the ground station to maintain communications, called a heartbeat,” Cummings explains. “If the heartbeat's lost, the aircraft ‘knows’ and goes to self-repair mode, trying a second radio, checking circuit breakers, and so on. Then, if nothing works, the aircraft goes to its known profile, follows waypoints, and lands itself with GPS and radar.”
Such planes weren't designed to have the parts-per-million defect rate that you’d expect in a passenger plane. (Indeed, the Global Hawk still hasn't fully met the Pentagon's specs, which require that a tag team of several planes provide “near-continuous” surveillance over a particular place for up to 30 days.) But even the Global Hawk's level of autonomy would surely be welcome in an airliner as a last-ditch way of saving it if the pilot and copilot were killed or incapacitated.
As significant as the technical hurdles are, however, by far the biggest impediment to pilotless flight lies in the mind. People who otherwise retain a friendly outlook toward futuristic technologies are quick to declare that they'd never board a plane run by software, which they know as the kludgy mess that makes their laptops freeze. But minds can be changed.
Each semester, Cummings asks her students at MIT whether they'd fly in a pilotless plane from Boston to Los Angeles. Two or three hands go up. Then she asks them how they'd feel if the fare was just 50 bucks. More than half the hands go up. Cummings's little experiment suggests that people's reservations about robo-flight aren't set in stone.
One factor that's often cited for keeping a pilot in charge is what's known as “shared fate.” That's the reassurance passengers get from knowing that the human in the cockpit wants to live just as much as they do. But shared fate is not the only way, or even the normal way, to ensure safe service. After all, restaurants don't employ food tasters to reassure diners, nor do losing defense lawyers join their clients in jail. It's usually enough for a professional to demonstrate sheer competence—the “right stuff” of aviator lore. And it's clear that automatic pilots—like those that land F-18s—now have a goodly amount of it.
Even Captain Chesley “Sully” Sullenberger, the pilot who deftly ditched his Airbus 320 airliner in New York City's Hudson River after its jet engines swallowed some geese, owed much to his onboard software—among other things, it managed the plane's angle of descent so as to avoid a stall.
But trusting software to safely shepherd hundreds of passengers across thousands of kilometers? A suspicious public isn't likely to buy into that vision, because safety is one of those things you can't have enough of. And how could a machine ever be as safe as a man-machine combo?
Yet the counterargument is also worth considering: Could the very collaboration of man and machine be causing human skills to wither? The FAA seems to think so. In a draft report cited by the Associated Press in July, the agency stated that pilots sometimes “abdicate too much responsibility to automated systems.” Automation encumbers pilots with too much help, and at some point the babysitter becomes the baby, hindering the software rather than helping it. This is the problem of “de-skilling,” and it is an argument for either using humans alone, or machines alone, but not putting them together (see IEEE Spectrum, “Automated to Death,” December 2009).
Consider the 2009 crash of Air France Flight 447 off the coast of Brazil. A report from the French investigators, released this summer, argued that the pilots hadn't been properly trained to handle a malfunction of the airspeed sensors system, the apparent proximate cause of the crash. Thanks to automation, which had taken over many of their routine tasks, some of their skills had atrophied. The report suggested that had the copilot (apparently the one in charge at the time) kept the plane level while trying to figure out what was going on with the airspeed, he could have mastered the situation. Instead, the copilot seems to have held the plane's nose up, causing the very stall he was trying to prevent. The French report [PDF] proposed that pilots get more experience in flying by the seat of their pants.
But it isn't easy for an airliner pilot to practice flying by the seat of his pants. Simulators hardly satisfy—isn't it the SimFlight world we're trying to get away from?—and even the occasional intervention in the plane's workings can't instill an intuitive feel for flying. And such interventions really are becoming occasional. Airliners—the Boeing 747, for instance—allow even landing to proceed hands free, says Davis, of New Mexico State. We are clearly in a particularly uneasy stage of the man-machine meld.
To win over the public, the autopilots of tomorrow will have to start today by exploiting niches where civilian pilots can't or won't work—just as was the case in the military. With time, the systems will improve and eventually fan out to conquer additional segments of the broader market.
Look for the following order of events. Today, small government-run UAVs are plying ocean routes, looking for pirates and lost sailors; next, companies will send the craft into the back country, along routes cleared for their passage by civilian regulators, to check on the state of pipelines and power lines. After that, UAVs will ferry valuable medical samples and packages. A doctor might, for example, put a vial of blood into a UAV and send it to the nearest teaching hospital for analysis; a courier service such as FedEx might fly important packages from Japan to California, using dedicated airfields on each country's shoreline, thus avoiding civilian air traffic altogether. Once they're demonstrated on the battlefield, robotic medevacs will graduate to civilian duty, rescuing people stranded by flood or fire. Maybe then, after seeing such rescues on television, the flying public will finally start to warm to the cold machine.
“We often talk about the 'save little Johnny' scenario, where no human-operated aircraft wants to go out in a terrible storm, so you send out robotic aircraft to save little Johnny, alone in the ocean,” said Rodney Walker, in an interview conducted last summer. Regrettably, Walker, a professor of electrical and electronic engineering and of aerospace engineering at the Queensland University of Technology, in Brisbane, Australia, died as this article was being prepared for publication.
Walker's team, however, continues to work on his more immediately useful project—to use pilotless planes to survey power lines and find early signs of encroaching brush that might take the lines down in a storm. (It was apparently a wayward branch in Ohio that triggered the 2003 blackout in North America.) The idea is perfect for the empty parts of Queensland, a province with nearly three times the space and just a fifth of the population of Texas.
“Our energy distributors are desperately looking for new ways to surveil power lines,” Walker told IEEE Spectrum in June. “They're taking the automated system we developed, but as a flight-assist system, and putting it into piloted aircraft for two to three years to prove the guidance and the automation. The pilot is largely hands off, but his presence allows us to meet regulations.” He added that it would probably take no more than a decade to dispense with the pilot and fly the plane, perhaps autonomously and “certainly by remote control.” Even then, such UAVs wouldn't be able to save little Johnny, but you could send out scores of them to sweep the seas for the tyke and then call in a piloted plane to pull him from the water.
Such baby steps in automation will eventually collide with another trend in aviation: the decades-long “decrewing” of airliners. Back in the day, the standard cockpit contained a full cast of characters, including the flight engineer, the navigator, and the radio operator. Technology has replaced those functions, and those people, one after the other.
The copilot is next in line. “Instead of a two-person cockpit, we'll see one person only, and software would serve as a backup,” says New Mexico State's Davis. “And it'll come only after a certain period of performance, when the insurance industry is willing to accept the risk.”
Of course, money is the necessary fuel for any pilotless air business, and so far, nobody in the industry has spied a huge pile of it waiting to be made. The field still needs a killer application to motivate moneyed interests to lobby the government and win over the flying public.
“The FAA isn't jumping ahead of things, because there hasn't been an extreme business case for doing so,” says Davis. “No killer biz application has stood up and said, ‘Here I am.’”
Cutting pilots’ and copilots’ salaries won't save much money: They matter little when spread over a jumbo jet's worth of passengers. Perhaps more significant, though hardly overwhelming, are the indirect costs of having to schedule flights so that fresh, rested crew members are available. Here the savings would be greater because you'd be able to minimize an airliner's downtime.
However, any true killer app would probably dispense with the existing business model altogether and replace it with an entirely new model, one that might start in a small niche and grow to take over commercial aviation. That's how the PC made its way from a glorified word processor to a general-purpose tool for small businesses and, later, for medium-size corporations. At some point it began to eat at the bottom of the mainframe market, in classic “disruptive” fashion.
For robo-flight, the killer app could start off with local air service. In many places just getting to the airport or to the main airport hub is the hardest leg of the trip. If small robo-planes could get you there, air travel would become vastly more attractive.
“Look at five-seat airplanes,” says David Vos, senior director of unmanned aircraft systems and control technologies for Rockwell Collins, in Cedar Rapids, Iowa. “You could make a business out of a [pilotless] taxi service from one small airport to another, and it'd be a whole new world.”
Just when we might see that world is still the great question. Back in 2002, two illustrious technogeeks bet US $2000 on whether full-fledged pilotless airliners would fly routinely by 2030. And they placed the bet (on the website Longbets.org) before much of the huge progress in military UAVs had even been demonstrated.
Craig Mundie, chief research and strategy officer at Microsoft, thought such an outcome likely. He imagined “arriving at a methodology for system design that yields as much dependability, on an everyday basis, as the triple-redundant computer that flew guys to the moon.” Eric Schmidt, now the executive chairman of Google, argued that “the FAA changes so slowly that if this were even all possible, the adoption and certification would all take at least 50 years.”
That would be true, perhaps, if the United States were the only force in civilian aviation. But some other country, with lots of space and fewer people, may very well decide that commercial pilotless flight makes sense now. Australia, are you listening?
This article originally appeared in print as “When will software have the right stuff?”
About the Author
For more about the author, see the Back Story, “Flying the Robotic Skies.”