Skip to Content

The Ascent of the Robotic Attack Jet

Building the planes is easy. Making them autonomous, and constructing airborne communications networks, is not.
March 1, 2005

Compared to many aeronautical curiosities that have taken wing at NASA’s Dryden Flight Research Center at California’s Edwards Air Force Base over the years, the latest military test stunts did not appear very remarkable. Last April, a low-slung aircraft, about the size of a sport utility vehicle but with batlike wings similar to those of the B-2 stealth bomber, took off, flew at 10,500 meters and then dropped a 110-kilogram inert precision bomb while zipping along at 700 kilometers per hour. Four months later, a pair of the aircraft took off and flew together. These were modest stunts, to be sure, except for this fact: the jets have no pilots. They are the future of warfare, the first working models of networked autonomous attack jets, and the U.S. Department of Defense would like to start building them by 2010.

Eventually such planes will be military mainstays. Of this, most observers are sure; it is simply a lot less expensive – and safer – to send machines into battle than to send people, who require food, sleep, training, and pay. Humans can only tolerate so much G-force and are prone to error; unmanned aircraft have the potential to be more dependable. Already, lone unmanned planes – with humans at the remote controls – are widely used for surveillance. But the next crop of planes will fly in coordinated groups, with more autonomy. They’ll tackle jobs such as attacking enemy air defenses, identifying new targets, and releasing precision bombs. “The long-range vision is that the president will wake up some day and decide he doesn’t like the cut of someone’s jib and send thither infinite numbers of myrmidons – robotic warriors – and that we could wage a war in which we wouldn’t put at risk our precious skins” is how John Pike, director of GlobalSecurity.org, a leading defense policy website, puts it.

Realizing this vision will require the creation of new airborne communications networks and a host of control systems that will make these jets more autonomous (though always under the ultimate control of a person) than anything built to date. These are the goals of a $4-billion, five-year program at the Defense Advanced Research Projects Agency (DARPA), the Pentagon’s advanced research arm. Though proposed Pentagon cuts are likely to push funding downward, the program is now DARPA’s largest. Under the program, the agency is paying aerospace behemoths Boeing and Northrop Grumman to develop distinct jets with common control systems; DARPA recently signed on the Johns Hopkins University Applied Physics Laboratory to help with the myriad networking, control, and processing problems. Because, while DARPA is ordering up new flying machines, it is also requiring something far more important: the electronic brains to make them work. The jets “are technologically advanced aircraft, to be sure, but the soul…lies in the command and control, sensor, and weapons systems that enable their operation, individually and collectively,” explains DARPA program director Michael Francis.

The brain – which is housed on board and within ground-control stations – must do some difficult new things. It must not only keep the plane aloft and on course, but enable groups of planes to fly in coördinated fashion. It must rapidly keep up with changing communication links as the jets slice through the atmosphere at 700 kilometers per hour or faster. It must help make preliminary targeting decisions and drop bombs. While the planes must include the latest networking bells and whistles, they mustn’t be too complex to use, so that a single controller – on the ground or in a manned jet flying with the unmanned ones – can effortlessly shepherd fleets of them.

Making each piece work – and adaptable to new missions, and applicable to different flying machines that may be built in the future – presents daunting software and control challenges. But if it works, it could transform how war is waged.

Successful Predator
In recent years, unmanned planes have proven themselves in war. For example, the Predator, a medium-altitude surveillance plane made by General Atomics, debuted in Bosnia and then served in Afghanistan and Iraq. The Global Hawk, made by Northrop Grumman, has been flying high-altitude reconnaissance missions for years. Meanwhile, Northrop has built and flown another unmanned prototype, called Pegasus, and shown that it could land on an aircraft carrier. But the Pentagon’s massive push for robotic attack planes began in earnest in 2003. That’s when the Pentagon set up the Northrop–Boeing competition and established a seven-year timetable to develop versions suitable for the air force and navy.

Boeing’s version is called X-45; a scaled-down prototype is what dropped the inert bomb last year, and a full-size model is under construction. Northrop’s version is called X-47; it builds on the Pegasus, and the next generation model is still under development. The X-45 is geared more to high-speed air force attacks, and the X-47 to naval reconnaissance and carrier landings. In both cases, the largest prototypes* are supposed to take their first test flights within two years. (Amid the current budgetary uncertainty, DARPA declined to make its researchers available for comment. Comments from DARPA officials in this story come from agency transcripts of presentations the officials made last year.)

It’s not yet clear how many of which version the Pentagon might eventually want to buy. In that important way, this effort differs from the intense, winner-take-all competition to build the F-35 Joint Strike Fighter, widely seen as the last manned fighter jet. Boeing lost the F-35 competition to Lockheed Martin in 2001. “It caused [Boeing] to skip a whole generation of fighter aircraft, after being the foremost fighter aircraft supplier,” says Paul Nisbet, an aerospace industry watcher at JSA Research in Newport, RI.

Boeing’s initial pair of scaled-down X-45s have already proved themselves in several initial demonstrations. In 2003, Boe­ing passed one milestone, showing how the plane’s ground controllers could coordinate flight plans with conventional air-traffic controllers and modify the X-45’s flight plan as needed. Then, in 2004, Boeing’s X-45s demonstrated a few more tricks – deploying inert bombs and, critically, demonstrating that its ground controllers could hand off the wireless yoke to another station nearly 1,400 kilometers away while the plane was in the air. Finally, Boeing showed that a single ground controller could control two X-45s.

And Boeing has another – perhaps more important – ace in the hole. The Pentagon already considers Boeing its “lead systems integrator” for a development project called Future Combat Systems. This mega­project is supposed to yield 18 kinds of sensor-riddled combat vehicles and the ad­vanced communications technologies to link soldiers with vehicles, planes, robots, and each other.

This program is also likely to get scaled back as part of a new round of Pentagon cuts; the new emphasis will be on adding technology to existing vehicles. Still, “both of these programs are talking about putting robots on the battlefield,” says Pike. “Boeing has looked at it and basically said, It’s the future. They are the lead company for robots on the ground battlefield, and they’ve staked out a pretty tall position for aerial robots.” But Boeing and Northrop recognize that the current program isn’t about who can build the best plane. “Before, we were looking at building the best platform,” says William Body, a Boeing manager for business development at the company’s R&D outpost, Phantom Works, based in Saint Louis, MO. “Now we are looking at creating the system-of-systems. We’ll have unmanned planes, we’ll have core technologies. But the endgame here is a network-centric endgame.”

Nitty-Gritty Things
The first and most obvious challenge is how to enable increasingly autonomous operation. This may seem like a problem nearly solved. After all, for years even the most mundane commercial jets have included autopilot features that maintain trim and course during long flights and that can also perform essentially automated takeoffs and landings.

Importantly, though, ordinary planes still, of course, have pilots sitting in the cockpits. Pilots make countless decisions to handle little breakdowns on the plane – and decide whether or not it’s appropriate to engage automated systems at all. “It’s not so easy when you don’t have a pilot in there to take care of mishaps, faults, failures, and all that jazz,” says Eric Feron, an aeronautical engineer at MIT who is not connected to the current DARPA program. “It’s unbelievable how much the human is able to act as the glue between the technological gaps. The human covers so many nitty-gritty things, from frequency switching to target acquisition and recognition.”

Then there is the problem of constructing what amounts to an Internet in the atmosphere. On the ground, mobile communications networks are fast expanding, thanks to cellular and Wi-Fi networks. But when you get up to 10,500 meters at speeds of 700 kilometers per hour or faster, new challenges arise. To pick one technical example: today’s airborne radio links incur one bit error in every 10,000 bits sent. That’s far too unreliable for an airborne Internet. In fact, it’s 100 times worse than what’s needed for the ground-based Internet to provide even minimal service, says Dave Kenyon, an information architect at the air force’s Electronic Systems Center in Bedford, MA. The center is developing satellite-based networks that will be used by all kinds of military planes, including future unmanned planes.

But even when satellites are used, the fact remains that jets cover great distances, and that communication links will thus regularly break. “From a networking perspective, the frequent making and breaking of links will require new or improved network routing protocols,” Kenyon says.

In other words, the unmanned planes will require new ways for information to change communication pathways on the fly – literally. “We will not always have perfect communication and, in fact, will always have some form of latency,” says Paul Waugh, a DARPA deputy director of the X-47 program. “Thus, the system, in all its parts, demands some level of autonomy, which means we will need smart platforms, smart sensors, and smart data processing.” The plane needs to think for itself, at least during the gaps. “We recognize that we have entered perhaps the richest, deepest part of the information revolution that deals with mobile, wireless computing,” Waugh says.

To further reduce strain on the communications networks, the planes must be designed to do as much work on board as possible. For example, after collecting images of targets, a plane must do much of the processing and filtering, sending only the most relevant images back to the human controllers. “The lines of code [for flying the plane] are minuscule compared to the lines of code required for mission planning, sensor management, and getting aircraft to fly together as a team,” says Rick Ludwig, the business development manager for Northrop’s program.

Eric Feron is developing a key system that is generic to all unmanned aircraft: the human–machine interface. In the future, unmanned jets might be controlled by a pilot in a single manned fighter jet. Feron is working on a natural language interface, so that the pilot can “talk back and forth with the [unmanned jet] as if it were just another person,” Feron says.

But that’s only part of the task; once the spoken commands are conveyed, those commands must be translated into a set of electronic and mechanical actions. Feron is also writing software that sorts and prioritizes commands and turns them into instructions that the machine can act on. Last June this command-and-control part of the system was successfully tested – using typed commands – on surrogate aircraft.

Above all, the software – which DARPA calls the “common operating system” – must be adaptable. The Northrop and Boeing versions are supposed to connect to one another and to other military systems – including those only yet envisioned. In various corners of academic, corporate, and military labs, autonomous helicopters, desk-sized robotic planes, and even insect-sized, flapping-wing aircraft are in various stages of development. If the networking and control systems are worked out, any future aircraft could make use of them. Once the airborne networks are as reliable as the land-based Internet, the myrmidons can take any form that pleases the Pentagon.

David Talbot is Technology Review’s chief correspondent.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.