What's missing in this picture? The RQ-4A Global Hawk, shown here at Edwards Air Force Base, routinely flies itself with only computers for crew.

At precisely 0600 hours most mornings, Lt. Col. Michael J. Guidry's CD alarm clock snaps on, and he's up and headed to his job testing the most revolutionary airplane ever built. Using his truck's cruise control to keep him from hotdogging down the Mojave Desert roads, Guidry traverses Edwards Air Force Base, driving right by the spot where the original Right Stuff man himself, Chuck Yeager, landed the Bell X-1 after breaking the sound barrier in 1947, and arrives at the hangar where the 21st-century version of flying glory, the RQ-4A Global Hawk, waits for him. 
    Although Guidry's plane resembles the sleek 1950s U-2 spy plane in length, height, and weight, the plane of the future looks a lot more like a killer whale than a flying machine. It's not exactly fast either. Powered by a single conventional Rolls-Royce Allison jet engine that musters a modest 7,150 pounds of thrust, it can't outpace a common Boeing 737 jetliner. One look and it's obvious why Guidry and others affectionately refer to it as Shamu. Nonetheless, there's a powerful reason to be impressed by the Global Hawk. This whale flies blind. 
    Guidry doesn't board the aircraft. No one does. The Global Hawk has no cockpit. It flies itself. Guidry and his colleagues sit in a cramped camouflaged trailer called the Mission Control Element. They have no joysticks, throttles, or pedals. They don't even have a pilot's-eye view from the plane: The Global Hawk has no forward-facing camera. It makes its own decisions about how to fly. And it is successful enough at that to raise a larger question certain to become part of most commercial airline passengers' futures: Are pilots obsolete? And if they're no longer needed, will passengers accept flying without them? 
    "There is no question that we will be able to operate aircraft automatically," says R. John Hansman, professor of aeronautics and astronautics at the Massachusetts Institute of Technology. "We are already doing it." Today's commercial autopilots are so sophisticated that "if you taxi out to the center line in a 777 and tell it to take off, you don't have to touch the stick again until you start braking on the roll out" at the arrival airport. 
    The Global Hawk is the first completely "man-out-of-the-loop airplane," says Guidry, who still marvels at the ability of two onboard computers to take off, fly, and land with more grace than humans can call up. His blind whale has already proved so capable as a long-duration surveillance craft over Afghanistan that Pentagon planners are eager to complete the testing of two more-advanced pilotless fighter planes. 

Pilots like Mike Guidry join the Air Force to fly. Guidry flew F-15s as a weapons systems officer on 68 missions over Iraq. He was a finalist to become an astronaut four years ago, then failed an eye test. He might not seem like the right person to take on an assignment at the 452 Flight Test Squadron, where human crew members stay on the ground and the airplanes fly themselves. But Guidry boasts about it. "This plane won the Collier Award," he says, referring to the Nobel Prize of aeronautics, putting the Hawk in league with the B-52, the Apollo 11 mission to the moon, and, of course, Yeager's X-1. 
    In 1999 a Global Hawk became the first computer-operated unmanned plane to fly itself across the Pacific, 23 hours to Australia, where it "landed itself dead on the center line," says Guidry. (In 1946 two B-17 bombers with no humans aboard flew from Hawaii to California, but they were entirely radio-controlled by pilots on the ground.) The Hawk is capable of far more: 36 hours of continuous flight and a range of 13,500 nautical miles before it must land to refuel. 
    Three years ahead of schedule, it proved itself in combat last fall over Afghanistan, where it cruised at 345 miles per hour at 65,000 feet, relaying reconnaissance to officers on the ground via satellite. With optical, infrared, and radar sensors, it took images day or night, cloudy or clear. Although many performance specifics are classified, Guidry says that in 24 hours the aircraft could map the entire state of Illinois in enough detail to show individual people. 
    As Guidry takes a long walk around the aircraft's elegant 116-foot tapered wing, he arrives at a computer station on a trolley that is hooked up to onboard computers by a small jack port under the single tail-mounted engine. "When you start it up," he says, "the airplane runs a check of all its systems, just the way a pilot would do. You download the mission plan via computer. You say, 'We want you to image Kandahar and Tora Bora.' " Then, inside a windowless camouflaged control trailer, all Guidry has to do is "click the takeoff button" with a computer mouse, "and off it goes." 
    In flight the Global Hawk is piloted by two onboard computers, which "keep track of the plane's location, altitude, attitude, and posture of all its controls and spoilers," says Guidry. Using a complex set of algorithms, the computers decide "how to fly and steer and take pictures to accomplish the mission plan." The only thing humans do is plan the mission on desktop computers and monitor signals from the plane in flight. Display screens in the Mission Control trailer show the plane's position and course on a map and give classic cockpit instrument readings such as altitude, airspeed, and throttle settings. One monitor shows scrolling computer code. "That tells us what the computers on board the plane are thinking," says Robert Ettinger, the flight-test manager employed by Northrop Grumman, the aerospace company that built the Global Hawk. 
    Global Hawk's watchers, lounging in comfy office chairs in their camouflaged trailer, can jump in with the click of a mouse to change the plane's heading or altitude. And they can uplink another mission plan. "But really," says Guidry, "Global Hawk works best when you just sit back and let it fly itself." 
    The aircraft's self-control goes well beyond simply following the orders of a human-programmed mission plan. It can't think, but it can cope with unexpected events. Ettinger explains that the plane has its own automated emergency plans. Depending on the severity of the problem and where it is in its mission, it may decide to fly in a holding pattern and wait for consultation with ground controllers, return to base and land, or make an emergency landing. Ettinger, who flew 100 combat missions in an F-4 over Vietnam before becoming an F-16 test pilot, says, "The computer by far does a better job at making an emergency landing" than human pilots do. With four onboard GPS navigational systems, the plane can land itself on any runway in the world, with only a 50-foot margin of error. More impressive, Ettinger says, it can follow a flight path within inches to any airfield equipped with a specific antenna. "I show a video of the plane landing to Air Force pilots, and they just look at each other and say, 'Man, this thing does a lot better than I can do.'" 

Surveillance is what the high-flying Global Hawk does best. It can gather detailed images— like this one (above) of an urban intersection— day or night in any weather. In the Mission Control trailer (below), operators keep track of the flight via monitors that show instrument readings such as altitude and airspeed.

The biggest surprise about the Global Hawk is that it contains no surprises at all. Its technology is "off the shelf," says Guidry. It simply combines the pre-existing technology of drones with the technology of existing autopilots. Radio-controlled pilotless drones have been used as target practice at least since the 1940s, and three-axis (pitch, yaw, roll) autopilots were first placed on commercial aircraft more than 60 years ago. 
    The Air Force, for example, has a new drone that is also celebrated for its surveillance work in Afghanistan. The RQ-1 (in Air Force notation, R stands for reconnaissance, Q stands for unmanned) is a delicate, 27-foot-long, 1,130-pound, upside-down flying spoon that carries the name of Predator. With a 101-horsepower engine and a maximum cruising speed of 135 miles per hour, it takes off from a regular runway and flies, like the Global Hawk, with no human aboard. But unlike the Global Hawk, the Predator requires a human pilot to be in the loop. Predator pilots sit in a remote trailer, but they have a full cockpit complete with joysticks and throttle controls. The cockpit screen shows a real-time pilot's-eye video image taken from the front of the aircraft. If the satellite link to the drone is interrupted, the Predator will circle until it re-establishes contact. Unlike the Global Hawk, the Predator has no autonomous ability to land itself. 
    Not that autonomous landing is cutting-edge technology, either. Most modern passenger aircraft are equipped with autopilots that can control a flight in zero forward visibility or without any help from the human crew. Aircraft so equipped are said to be Category III certified, a technology that dates back to the 1960s. "Our landings are very similar to a commercial Category III landing," Guidry says. 
    Pilots use the autoland function of Category III systems far more often than conditions require, says Joseph Baranowski, a retired 30-year-veteran commercial pilot. "Sometimes you're just tired," he says, "and it's easier to let the plane do it." Airlines don't publicize the frequency of autoland, says MIT professor John Hansman, but, "I would say chances are better than even that you've already done an autoland if you fly regularly." 

Elsewhere on Edwards Air Force Base, hidden from public view, stands an aircraft that promises to find the edge of the envelope and rip it open— the X-45A (X for experimental). à It is also known as the Air Force UCAV, for Unmanned Combat Air Vehicle. The X-45A completed high-speed autonomous taxi tests in April and successfully completed its first test flight in May. Just behind it in development is a version for the Navy, capable of operating from an aircraft carrier. Whereas the Global Hawk looks as if it could accommodate Guidry, if only someone would cut him a hatch, the Air Force UCAV's sleek 27-foot-long composite and aluminum pared design clearly has no room for a person. Boeing designers did leave room for a center-mounted 6,300-pound-thrust engine, small bays for computers and navigation systems, and large bays for ordnance. 
    "The UCAVs will go way beyond the Global Hawk," says program manager Col. Michael Leahy. Their autonomy will be so robust that it will take only one operator to oversee four of them at once. These quads, Leahy says, "will hunt in packs," flying in formation and adjusting their battle plans as they go. On a mission to take out enemy antiaircraft stations, for example, if one UCAV takes a hit and loses its sensors, the computers of all four planes will confer about tactics without help from any human and figure out the best way to fire the damaged plane's munitions, using the working sensors on the other three planes. 
    Self-flying, self-planning, pack-hunting UCAVs, expected to be operational by 2008, will deploy as the first wave against enemy air defenses when conditions are too risky for piloted planes. Eventually, engineers hope to build dogfighting UCAVs, a notion that thrills veteran top gun Ettinger. He says manned fighters reached their maneuverability limits decades ago because human pilots cannot tolerate much more than nine positive g's of force. Cruise missiles, by contrast, can make 20-g maneuvers, and so could a UCAV. "When it's fully developed," Ettinger says, "the UCAV will think better and move better than a human pilot and be a much better dogfighter." 
    Navy UCAVs will need greater sophistication to make carrier landings, long considered the most difficult flying challenge. David Mazur, program manager of the Navy's Pegasus UCAV, says the craft will use a Shipboard Relative GPS system, which provides the exact position of the aircraft relative to the tossing deck of the carrier within eight inches of precision. "In a heavy-sea state, it should do the landing better than a pilot," he says. In addition, the Navy UCAV will have to fit in among the beehive swarming of carrier traffic. "We want to operate all mixed up, human and nonhuman, without missing a beat," says Randy Secor, the Northrop Grumman Navy UCAV program manager. 
    One thing UCAVs won't do without human permission, at least for the time being, is drop bombs and fire missiles. "Someone has to have the moral obligation before that weapon is released," says Mazur. But putting a human back in the loop at a critical moment— when that surface-to-air missile station is about to blast the UCAV out of the sky— may turn out to be tricky. 

The human touch is a rarity for the Global Hawk and occurs only when it's time for routine maintenance.

Formation flying, dogfighting, carrier ops: If demanding military flying can be done handily by autonomous aircraft, why not autonomous airliners carrying hundreds of passengers? People already trust their lives to simple automated systems such as traffic lights and light rail. 
    Beyond reliability concerns, such as what happens if the two control computers fail while passengers are at 35,000 feet, engineers worry that computers can't improvise very well. "Machines need information," says Hansman. "If I can write the rule, the machine can do it. But if I can't write the rule ahead of time, machines don't do well." Humans, on the other hand, can think on their feet. In July 1989, for example, on United Airlines Flight 232, something went wrong that no one had predicted: The tail engine of the DC-10 threw its fan rotor, slicing through both main and backup hydraulic lines connecting the pilot to control surfaces such as the rudder and the ailerons. The plane banked hard right and would have bored into a cornfield, killing all 296 people on board, but the captain and the flight crew invented a solution. By adjusting the two wing-mounted engine throttles one at a time, the pilots were able to steer the plane crudely with differential thrust— like steering a rowboat by pulling on one oar or the other. They managed to crash land at the Sioux City, Iowa, airport and save 184 people aboard. 
    If an autopilot alone had been flying that DC-10, no one would have survived, but engineers at McDonnell Douglas subsequently programmed autopilots to steer a rudderless plane by using throttles. That improvement was never included in commercial autopilots because of cost and because total hydraulic failure was considered too unlikely to worry about. And therein lies an interesting trade-off between costs in the airline industry. If airlines could pour pilot-salary money into computer-pilot improvements, aircraft manufacturers might be able to endow the computer brain of an autonomous aircraft with a flying lesson from every accident in the history of flight. This would make it capable of every good move a human pilot thought up in an emergency and, more important, make it incapable of human error, found to be the cause of 70 percent of fatal crashes. Novel emergencies will no doubt arise, and autonomous planes will surely crash. But Hansman says it should take less than a decade for autonomous aircraft to become highly visible replacements, having surpassed piloted ones in efficiency and reliability in most instances. Then information technology will trump improvisation in flying, just as Deep Blue trumped Garry Kasparov in chess. 
    "It's not a question of 'Can we do it?' " says Hansman. It's a question of public acceptance. "While a machine may have a higher probability of a correct decision, if the consequence of an error is death, it is not clear to me that we are ready to trust the machine over the human." 
    Meanwhile, Lieutenant Colonel Guidry keeps the faith: He would allow his wife and daughters to fly anywhere on the Global Hawk— if there were room for them. 


To pilot or not to pilot . . .

Pilots tend to think autonomous aircraft are a terrible idea. "You have to let us fly the airplanes," says veteran captain Joseph Baranowski. "Already, on the Airbus planes, the computer will override the pilot's reactions. It won't let you go into a 60-degree angle of bank. But what if I need a 60-degree angle of bank to avoid a building? The computer pilot would crash you." Baranowski concedes that pilot error contributes to the majority of airplane accidents. "I'm sure the engineers think their [autonomous aircraft] death rates would be less than ours. But don't kid yourself— acceptable losses are acceptable losses. The industry thinks that way." 
    Oddly enough, Baranowski would rather see the industry go to full automation, with no pilots, than to let current autopilot technologies continue to dull piloting skills. "This generation of airplanes is designed to be programmed, not flown," he says. Milton Painter, former director of flight standards at Southwest Airlines, agrees: "We're scared to death of the fact that automation dulls the pilot's skills." Southwest, primarily a short-hop carrier, doesn't have a strict automated-navigation-use policy. "I ride jump seat in other airlines, and when you see a pilot who's heavily into automation, you see that their situational awareness is diminished. I believe strongly in the pilot and pilot skills. When one of the systems fails, you need a pilot in the plane."