Business Week November 12, 2001
Planes that Know What to Bomb
Robots Take Wing
By Stanley Holmes
For a fighter pilot, no mission is more dangerous than a Wild Weasel. To flush out hidden surface-to-air missiles in the early days of a war, the first jets into enemy airspace advertise their presence while trailing fighters lock on to the SAM targets and take them out. But in Afghanistan, the U.S. had a trick up its sleeve: In place of fighter pilots, it sent in small, remote-controlled Predator drones -- originally designed for reconnaissance but now armed with Hellfire antitank missiles.
The Predator ploy is just a crude foreshadowing of what lies ahead. By the end of the decade, the military could be sending the first true robotic warplanes into battle. These autonomous weapons-on-wings would sniff out hidden enemy air defenses before human-piloted fighters or bombers ventured into enemy airspace, deliver up to 3,000 pounds of smart bombs and missiles, and even take on enemy fighter jets. Called unmanned combat air vehicles, or UCAVs, their presence is likely to redefine the role of warplanes -- and even warfare itself -- by giving machines more responsibility for attacking enemy targets. For six years, the U.S. military has employed drones called Predators -- and more recently, Global Hawks -- to spy on enemy forces. And for at least two decades, it has relied on cruise missiles. These are flying bombs that hug the ground to avoid radar as they near a pre-programmed target, then use smart imaging systems to lock on to the target and make necessary final-course corrections.
But UCAVs take the technology to new heights: For the first time, military leaders will send into combat machines that, in theory, could choose their own targets, decide whether to destroy them, and figure out when to return to base. Many military experts consider this next wave of technology inevitable. Ever-improving silicon smarts will make it both feasible and affordable -- and airplane designers are itching to create fighters that can outmaneuver any human pilot. Already, fighters such as the F-16 can turn so sharply that pilots would black out without special flight suits that maintain blood flow to the brain.
The technology has a dark side, however: While UCAV prototypes such as the X-45A are impressive feats of engineering, they also reflect the military's increasing reliance on machines to analyze and act on the information that it gathers -- a trend that disturbs some experts. ''I worry that human intelligence analysts will become so overwhelmed by data that they'll grow excessively reliant and overly confident of computers picking the targets,'' says John Pike, a defense analyst for Globalsecurity.org.
Despite such worries, most defense experts predict that UCAVs will eventually supplant piloted fighter jets, including Joint Strike Fighters. These are advanced first-strike aircraft that can evade radar, which the military plans to fund to the tune of $ 200 billion over the next 25 years. UCAVs boast at least three advantages over JSFs: They can fly extended missions for more than 24 hours, where the limiting factor often is pilot endurance. And at just $ 10 million to $ 15 million a plane, they'll be a third as expensive as a new JSF fighter and cost just two-thirds as much to operate and maintain. And, of course, no U.S. pilots are put in harm's way. ''UCAVs are potentially the most potent long-term competition for the Joint Strike Fighter,'' says a recent report by Merrill Lynch & Co. analyst Byron Callan, who predicts that UCAVs could cut into future orders for JSFs.
UCAVs are an outgrowth of the military's longstanding focus on ''information superiority.'' This concept hinges on dramatic advances in information technology, including ''faster processing, multiple satellites, and fast linkups,'' says Boeing Co. Phantom Works President George K. Muellner, a former Air Force lieutenant general who directs the planemaker's UCAV work. Eventually, Muellner predicts, every element of warfare will be linked to a communications web that can be visually displayed on monitors at a command-and-control center, either airborne or on the ground. Smart machines will analyze and distribute the information in real time to UCAVS, tanks on the ground, or manned fighter jets.
Under this framework, machines will play a bigger role in integrating the information than they do today. But even though UCAVs will have the capability to make their own decisions, a human will give the final O.K. to drop bombs, according to an Air Force brigadier general who heads the unmanned air vehicle program. ''It's not our intent to have robots flying in the air automatically dropping weapons without human control,'' he says. HOW SMART? Such statements don't allay all the fears swirling around the UCAV concept. One sore point with critics such as Human Rights Watch in Washington is the planes' ability to identify targets and make last-minute corrections if the situation on the ground changes. That's a challenge even for human pilots -- witness U.S. warplanes that have twice, tragically, dropped bombs on an International Red Cross building in Kabul. ''The more combatants are removed from one another, the greater is our concern,'' says Joost Hiltermann, executive director of the arms division for Human Rights Watch.
There's reason for concern. Crucial identification systems, such as automatic target recognition, have yet to prove their reliability. Using a combination of sensors, high-speed algorithms, and artificial intelligence, these systems are supposed to imbue UCAVs with enough smarts to distinguish a school bus from a tank or a weapons depot from a food shelter. But Glenn Buchan, a senior defense analyst for Rand Corp. and author of a classified study on UCAVs, is skeptical. Such technology, he says, has ''notoriously underachieved'' in simulations of combat conditions. And given the hurdles in getting computers to distinguish real targets from false ones, reliability flaws will continue for at least a decade. ''It's an extraordinarily challenging and controversial issue,'' Buchan says. ''We're not very far along.''
This view is endorsed by Peter G. Neumann, principal scientist at SRI International Computer, a Palo Alto (Calif.) nonprofit research center and author of Computer-Related Risks. Routine banking software has come a long way in terms of security and reliability, but creating dependable systems for autonomous aircraft is more difficult. ''One challenge is writing reasonably correct software code, and one is testing the [weapons] system under all sorts of scenarios,'' Neumann says. ''There are all kinds of false negatives and false positives'' in recognizing targets.
In October, the U.S. Air Force raised the stakes in this debate. For the first time in history, it conducted remote-controlled bombing raids, penetrating deep inside Afghanistan. The agent in the attack was a Predator surveillance drone equipped with Hellfire missiles that were fired at Taliban targets by operators directing the mission from U.S. bases thousands of miles away. (The military won't comment on these raids.)
Built by General Atomics Inc. of San Diego, the Predator is designed to circle over a target at about 20,000 feet while its cameras record everything that moves. It has been used for intelligence-gathering in Kosovo and in no-fly zones in Iraq. Earlier this year, in a test at Nellis Air Force base in Nevada, a Predator with a Hellfire missile successfully blew up a tank.
Afghanistan is also the testing ground for a much more advanced, jet-powered surveillance drone called the Global Hawk. It circles at about 65,000 feet, making it harder to shoot down than the low-flying Predators. It can also stay aloft for 24 hours -- or the equivalent of five round-trips from New York to Los Angeles -- without stopping or refueling, and it can circle in one area for up to 12 hours. All the while, the plane takes infrared or radar pictures and even analyzes heat signatures, thus enabling intelligence specialists to see not only what is on the ground but also what was there. So far, Northrop has produced only four Global Hawks, but it could sell up to 60 of them by the end of the decade.
Because of UCAVs' inherent advantages over these surveillance drones, the Air Force awarded Boeing a $ 135 million contract to produce just two prototypes, the X-45s. The Y-shaped planes are 27 feet long, with a wingspan of nearly 34 feet, and should take their maiden voyages early next year. The Navy is also interested in these planes, which will be capable of flying, identifying, tracking, and hitting targets on their own. In an actual conflict, human ''operators'' -- each monitoring four or five UCAVs at a time -- will confirm and authorize targets.
When the planes aren't flying, they'll sit in crates with their wings removed, ready to be assembled and prepped for take-off in just one hour. If the testing goes well, Boeing could turn the development contract into a full-scale production run. Muellner says the Air Force could order more than a hundred X-45s over the next decade, with the first delivery anticipated in 2008.
That's far too soon for comfort in the view of Human Rights Watch and techno-skeptics such as Peter Neumann. But neither holds much hope of stalling the UCAV programs. Economics, and the hope of cutting casualties, are in alignment. That means robot fighters are sure to play a role in the future of war.
Copyright 2001 The McGraw-Hill Companies, Inc.