A SWORDS unit undergoes testing. Image by QinetiQ (parent company of Foster-Miller).

You may not have heard about it, but history's first robot soldier has already come and gone.

The so-called SWORDS device was a weaponized variant of the popular TALON bomb disposal robot. A meter tall, remotely controlled, and toting a machine gun, SWORDS was designed as a replacement for human troops on the battlefield. After years and millions of dollars in development, the deployment earlier this year of the first three units to Iraq was greeted with heavy coverage by the mass media. But almost the moment that the US$230,000-apiece devices arrived on the ground, the SWORDS program was abandoned. All project funding was cut, and the robots have never been allowed to fire a single shot.

Soon, the media began to echo rumors of a potentially lethal malfunction in the field, though these stories have been vigorously denied by both the manufacturers and the Pentagon.

Now, months after the robots were pulled from the battlefield, there is still no official word on why.

With SWORDS being just the first of a wave of robot warriors poised to hit front lines around the world, there is good reason to worry why the program was so rapidly terminated.

Toward the end of the year 2000, Congress passed a bill requiring that, by 2010, one third of military ground vehicles and deep-strike aircraft be replaced by robotic devices.

This was part of a sweeping new defence plan known as Future Combat Systems—the Army's attempt to rebuild itself as a "21st century fighting force." Pentagon officials have been quoted referring to this effort as "the new Manhattan Project." At a value of US$127 billion dollars, it is the most expensive military contract in American history. By the end of the decade, it is expected to have driven the peacetime defence budget up to more than half a trillion dollars a year.

The cornerstone of the Future Combat Systems project is the ubiquitous use of advanced electronics, replacing humans wherever possible.

Colonel Bruce Jette, one of those in charge of rolling out the new systems on the ground, told Associated Press in 2003, "The best I can say at this point is the army is aggressively looking at applying robots in all future operations. . . . I think they would be useful in open battle."

With the current administration's warlike attitude and terror of body-bags, the attractiveness of this approach is obvious.

As Jette puts it, "I don't have a problem writing to iRobot, saying 'I'm sorry your robot died, can we get another?'"

Rodney A. Brooks, Director of the Computer Science and Artificial Intelligence Laboratory at MIT, agrees: "Anyone who's a decision-maker doesn't want American lives at risk. It's the same as 'should soldiers be given body armour?' It's a moral issue. And cost comes in."

It certainly does. With a recent Pentagon study putting the average cost of a single soldier over their lifespan at around US$4 million dollars, the New York Times has asked whether the project is motivated less by saving lives than by a desire to cut down on the number of expensive humans in dangerous places.

Not only will a mechanical device cost only about one-tenth of what a human doing the same job would, but it is also particularly well suited to modern forms of urban warfare. As Joe Dyer, head of iRobot's military robotics division, tactfully explains, "A robot can shoot second."

But some are wondering if the use of these remote-controlled systems reduces war to just another video game.

New Statesman quoted the developers of one military robot as saying that the controls had been modeled after the PlayStation 2 because "that's what these 18-19-year-old marines have been playing with pretty much all of their lives."

After driving around Lockheed Martin's prototype MULE drone with an Xbox 360 controller, Popular Mechanic's Erik Sofge felt he'd seen a "glimpse of what future warfare might look like—robotic, autonomous, and just a little bit chilling."

Examining the ways in which the Future Combat Systems project has distributed its funding, we can see that the replacement of conventional forces with robotic technology will take place in three waves.

The First Wave, which has been underway for quite some time, generally takes the form of small orders placed with current contractors to produce weaponized versions of present robotic technology. The resulting devices seem to be used mainly to test implementation and to prepare the way for the Second Wave devices.

The Second Wave consists of numerous large contracts for the creation of new, specialized robots, many relying on innovative technology. Scores of progressively more heavy-duty machines are due to begin service over the next seven years.

The Third Wave takes the form of grants to institutions and companies to conduct advanced research into areas of interest. The eventual goal is to produce a new kind of robot, much more like what one would expect to see in the movies. The research seems to focus on the development of devices that are largely autonomous.

The TALON robot (above) has proven to be durable in bomb disposal, and search and rescue. Image by QinetiQ (parent company of Foster-Miller).

SWORDS was a classic First Wave project—a weaponized variant of the TALON system.

Foster-Miller, the Boston-based makers of TALON, boast of their product's capabilities on their website, mentioning the simplicity with which it climbs stairs and its ability to travel safely through sand, snow, and depths of water up to 100 feet. They claim that "TALON is the fastest robot on the market today, easily keeping pace with a running soldier" and proudly recount a case from Iraq in which one of the units was blown off the roof of an Army Humvee as it crossed a bridge over a river. Not only did it survive the fall into the river unharmed, but was able to drive itself onto the bank for easy recovery.

There is no doubt that it is a highly reliable device: a TALON that was used on Ground Zero in the aftermath of the 9/11 attacks worked for 45 continuous days without requiring maintenance. The TALON was the only robot to do so (though, as a large robot, it was kept away from the more precarious areas).

Foster-Miller estimates that their products have carried out around 20,000 explosive-removal operations during the subsequent wars in Afghanistan and Iraq.

The affection that many troops feel for the robots has been widely commented upon. The Washington Post reported in 2007 that a number of bomb-clearing devices had been nominated for medals by their human colleagues. One, the property of the 737th Ordinance Company, was promoted by its fellow soldiers to the rank of Staff Sergeant. "Sgt. TALON" had also been the recipient of three Purple Hearts.

The TALON's demonstrated hardiness and reliability must have made it seem like the perfect candidate for weaponization.

The development of SWORDS (Special Weapons Observation Reconnaissance Detection System) consisted almost entirely of taking a standard TALON chassis and bolting on an off-the-shelf product called a "Telepresent Rapid Aiming Platform" (TRAP): a remote-controlled gun mount.

The weapons that SWORDS was designed to handle include a M16 rifle, a M249 machine gun, a larger M240 machine gun, a M82 Barrett rifle, a six-barreled 40 mm grenade launcher, various antitank RPGs (rocket-propelled grenades), and a 66 mm M202A1 FLASH (a rocket launcher that fires napalm-filled shells).

As deadly as this drone sounds, it is clear that it was never intended to be anything more than a stop-gap measure.

The MAARS device (above) is the larger successor to the SWORDS. Image by QinetiQ (parent company of Foster-Miller).

In addition to the SWORDS contract, Foster-Miller received an order for the development of the MAARS: a larger, deadlier machine. SWORDS was a quick and simple upgrade of existing technology, created so that the military could experiment with battlefield robots over the years it would take for the Second Wave MAARS to be completed.

The pullout of SWORDS was announced, without explanation, in April 2008.

A few days later, the Army's Program Executive Officer for Ground Forces, Kevin Fahey, gave a short talk at the RoboBusiness conference. During question time afterwards, an attendee casually asked him about the withdrawal of SWORDS.

Fahey, appearing uncomfortable, gave a response that was, according to all reports, "vague." He mentioned that, while it had never unexpectedly opened fire, there was at least one incident where "the gun started moving when it was not intended to move." He went on to say, apparently trying to explain why the military was treading so carefully in this area, "Once you've done something that's really bad, it can take 10 or 20 years to try it again."

These remarks were reported by Popular Mechanics in a short article that went online on April 9, 2008.

The Internet immediately began to bubble with countless articles like "US War Robots in Iraq 'Turned Guns' on Fleshy Comrades" and "Combat Robot Attempts Rebellion Against Human Masters in Iraq." The idea that a SWORDS malfunction had very nearly precipitated a major friendly-fire incident soon became a widely circulated story.

Foster-Miller was furious. Within days they had issued a media release that attempted to comprehensively refute this increasingly popular anecdote.

The company's media release claimed that the project had not lost funding; the military had never planned to purchase more than three of the prototypes—an assertion that ran completely counter to all of the previous publicity. Popular Mechanics noted that spending millions of dollars on development and then immediately losing interest is a perfect definition of a program being cancelled.

The press release was particularly adamant about the claims of erratic behavior, dismissing them as an "urban legend." The company insisted, "There have been no instances of uncommanded or unexpected movements by SWORDS during this period, whether in-theatre or elsewhere," but did admit, "A few years ago during the robot's development, there were three minor movement issues that were expected, identified and addressed during rigorous stateside testing." Reportedly, these issues proved to be caused by a loose wire, a solder break, and a motor shutting down to prevent overheating. All the causes were said to have been resolved, with measures being taken to ensure that they could not arise again.

Furthermore, the company's release stated, "Any comments made after this timeframe about setbacks related to the robotics industry were hypothetical—never in response to some non-existent SWORDS incident."

The company pointed out that the robots had not technically been "[p]ulled-out" of conflict; they were still in Iraq—even if they weren't being used.

About the reasons for the Army's decision, the only hint given by the company was that "some technical issues still remain[.]"

Shortly afterwards, almost all information relating to SWORDS was removed from Foster-Miller's company website and replaced by a reiteration of their media statement in large, bold letters.

The uncharacteristic caution that the Army has apparently shown in dealing with the robots has been widely remarked upon.

The development of the V-22 Osprey aircraft has been completely unaffected by 30 people being killed during testing.

An RQ-1 Predator drone is marshalled by Senior Airman Robert Mascorro in Tallil Air Base, Iraq: January 20, 2008. U.S. Air Force photo by Staff Sgt. Suzanne M. Jenkins.

There is an even closer parallel: in the Balkans, in the early days of the RQ-1 Predator (an unmanned aerial vehicle, or UAV), there was an incident in which one of the drones suffered a major malfunction whilst in the air. Only some very fancy flying by its pilot prevented the aircraft—filled with around a tonne of highly explosive fuel—from crashing into a local primary school.

This near catastrophe didn't even slightly dent the Army's enthusiasm for the craft.

What could be these "technical issues" that have obviously worried the Pentagon so very much?

Examining the unit's history, we can see hints of trouble from early on.

On December 6, 2004, the Army News Service issued a story ("Armed robots to march into battle") announcing that SWORDS was fully operational and would be deployed to Iraq early the next year.

This claim was also made through other sources, including Associated Press, BBC World, and Defence Review, all reporting that eighteen of the machines were about to reach the battlefield.

In early 2005, the press were informed of a "media opportunity" to observe the robot in action. This was to take place on March 1, 2005, at Picatinny Arsenal in New Jersey.

The demonstration was abruptly cancelled due to "logistics issues."

Everything suddenly went very quiet.

In 2006, the Army conducted a "safety confirmation" on the system. It passed, but still everything remained quiet until early 2008's sudden media blitz as the three units finally made it to Iraq, where the program abruptly died.

All of this suggests a long-running problem.

But the eventual deployment was designed as a media event: why would the Army risk serious embarrassment by not merely using, but hyping, a device that they had grave doubts about?

The only clue that has ever been officially given as to the nature of these mysterious "technical issues" can be found in Foster-Miller's promotional brochure for their forthcoming MAARS robot. This glossy handout makes the claim that "compared to its SWORDS predecessor" the MAARS "has significant improvements in command and control, situational awareness, manoeuvrability, mobility, lethality and safety."

These statements are obviously intentionally vague, but they indirectly cover the three main suspects for the SWORDS robot's sudden demise: fragility of the weapon mount (thus, poor "lethality" in sustained firefights); limitations in vision and aiming ("situational awareness"); and issues of radio interference ("command and control").

One of the first media write-ups on SWORDS appeared in Defence Review in March 2003. The writer, David Crane, was curious about how the military planned to prevent the device from being turned into scrap metal the moment it encountered enemy fire. This is a valid concern: the vast majority of the unit's cost lies in the TRAP apparatus, and while the TALON chassis is heavily armored, designed to withstand bomb blasts and to be easily repairable, the complex and delicate TRAP is certainly not.

The negative publicity value of having such expensive and heavily promoted machines destroyed before they could bring down a single enemy soldier would be considerable.

Did the Army finally decide that it would be prudent to wait until something more lethal was developed?

Another possibility is suggested by the remarks of Colonel Barry Shoop, Deputy Head of the Electrical Engineering and Computer Science Department at West Point. He feels that the reason the development of remote-controlled ground vehicles has lagged behind that of aerial devices is because of the technological difficulties involved in close-quartered, urban battles: "The technical challenges are greater, think of the kind of image and graphics processing you need to make positive identification to use lethal force. That's inhibiting."

In 2006, National Defense quoted bomb disposal experts as complaining that TALON's vision was inferior to that of its rivals, and could cause difficulties, especially at night.

During testing, SWORDS proved capable of repeatedly hitting targets up to two kilometers away—an impressive aim. The test results also indicated, however, that the machine was "not as accurate" if movement was involved.

Statements about the tests do not make clear how great a decrease in accuracy resulted.

Was it problems with vision and aiming that made the Army nervous about pulling SWORDS's trigger?

But the most likely culprit for the failure of drones on the ground, despite their success in the air, lies in a fundamental difference between the two systems.

Predator drones are controlled by a line-of-sight radio link: a directional beam emanating from a ground station. To lose contact with this, they would need to fly over the horizon or have something large pass between them and the transmitter. If they do lose contact for whatever reason, they are equipped to use a satellite link instead.

Both line-of-sight and satellite transmission systems are unlikely to be affected by jamming or radio interference.

On the ground, things aren't as simple.

Foster-Miller often like to point out that, over open terrain, SWORDS can be controlled from a distance of more than a kilometer. If there are objects between the operator and the command unit, then "this distance decreases."

The company's statements never indicate by exactly how much this distance can be expected to decrease.

This is a serious concern; most cities are composed largely of hardened concrete walls, which can block RF (radio frequency) signals entirely, so it is possible that contact with the units will be lost almost immediately.

But things are even more complicated than that.

One of the most popular weapons used against coalition troops is the radio-activated roadside bomb; in an attempt to prevent these from being detonated, the U.S. Army uses heavy radio-jamming. Jammers are included in all convoys, and small personal versions are carried by most soldiers. Of the latter devices, there are reportedly well over 10,000 in Iraq.

Radio control in this environment is, to say the least, challenging.

The Army's former bomb-disposal robot, the Vanguard, has been largely scrapped due to its total inability to operate under these conditions.

How are the newer machines coping? Not surprisingly, neither the military nor manufacturers have been terribly talkative about their machines' limitations.

However, bomb-disposal teams have complained to Foster-Miller about the TALON's lack of a signal strength indicator. This would seem to suggest that transmission problems have been encountered.

Other sources also suggest that such problems they are also likely to be common.

In 2001, NASA funded the Mars on Earth project. This was an attempt to test the feasibility of a proposed manned expedition to Mars by having teams of scientists simulate the mission in extreme environments on Earth. During this project, various "declassified military robots" stood in for external probes: one of these was a TALON. It was found that none of the robots could be controlled from inside the simulated Mars Habitat because of "radio interference issues."

In 2006, Laurence Livermore National Laboratories developed a bomb disposal robot that was controlled by a long cable. The head of the project said that they had started work on this "after hearing complaints from soldiers that robots operated by wireless controllers were unreliable and subject to radio interference."

In 2007, the Department of Homeland Security decided to develop performance standards for robots likely to be used for search-and-rescue operations. They put a number of robots, including a TALON, through simulated missions to examine their current capabilities.

Radio interference was found to have effects described as "serious-to-catastrophic" in ten of the fourteen units tested—though to avoid embarrassing contractors (and hence the government), the affected robots' brand names were never released.

Radio interference problems are worrying because of the possibility that, in addition to severely reducing the robots' range, this noise may affect how they respond to instructions, or even be misinterpreted as legitimate commands.

The nightmare scenario would be a drone misinterpreting random signals as a command to open fire.

Fortunately, steps have been taken to protect against this happening. The TRAP system uses 256-bit encryption and requires a 120-bit authentication sequence before it can be activated—measures that make random firing highly unlikely.

The TALON itself, however, is a much simpler beast. It is possible that interference could cause it to move erratically and unexpectedly—not a good thing when one is using a machine gun/grenade launcher/antitank missile mounted on top of it. Problems of this sort would, of course, be greatly reduced if similar encryption methods were used for all commands—but this would also greatly limit the robot's range, since encrypted signals must be received fully and strongly to be acted upon.

But even this would not be foolproof: the Army has also never explained what measures would be put in place to prevent nearby units from interfering with one another. If identical encryption protocols were being used by multiple units, the accidental firing of a gun would be a very real possibility if each robot misinterpreted the order as its own.

As mentioned, the deployment of SWORDS was announced in 2004, but then suddenly delayed for three years. The reason for this is a mystery.

Wired's Noah Schachtman, however, has no doubts as to why: "The reason: safety. Talons [sic]—like all computers, like all machines—have a tendency to act extremely weird every once in a while."

Days before the SWORDS project was finally cancelled, Schachtman interviewed some of its leaders. The system was then on its way to Iraq, and they were flushed with what they believed was the eventual success of their program.

Apparently, Schachtman received admissions that, in the past, there had been problems with the way the unit responded to commands. Orders would either be ignored completely or acted on only after a delay. The lag between the button being pressed and the robot responding was sometimes as great as eight seconds—an amount that could be potentially lethal during a battle situation.

This was found to be due to transmission problems.

Schachtman learned that the three units being sent to Iraq had been retooled to prevent this from happening: "Now, the SWORDS won't act on a command unless it's received right away."

Michael Zecca, the program manager, also said that "Kill Switches" (devices that instantly shut the system down) had been added to the robots: "So now we can kill the unit if it goes crazy."

Did the Army receive their new machines on the radio-clogged battlefield, relieved that all of those worrying signal problems had been resolved . . . and then discover that they hadn't been?

The clear transmission of signals through an environment such as Iraq is a technical nightmare, one that has yet to be resolved.

It is not known how the makers of the Second Wave devices plan to deal with this serious difficulty.

Nor is it a purely academic concern any longer; on June 4, 2008, the first MAARS unit was shipped to Iraq for testing in the field. Interestingly, unlike with SWORDS, this deployment has been kept largely quiet from the mass media.

The Warrior will come in weaponized and nonweaponized variants (the latter being seen above). Image by iRobot.

Foster-Miller's main rival, iRobot, is also expected to have its armed Warrior drone on the battlefield within months.

The huge problem of reliable radio-control will inevitably put even more impetus behind the development of the Third Wave autonomous machines.

The Pentagon is already doing whatever it can to speed these devices' delivery.

One of the Pentagon's initiatives is the DARPA Grand Challenge—a race for robot cars with a million-dollar prize.

In the latest race, competing robots were required to navigate along the roads of an abandoned town, picking out their routes and obeying all traffic rules.

There were a number of successful entries.

Allowing a robot to decide whether to take the bridge or the tunnel is one thing, but will they ever be trusted to fire at will?

John Tirpak, editor of Air Force Magazine, believes so. He says that once they "establish a track record of reliability in finding the right targets and employing weapons properly, the machines will be trusted to do even that."

One idea that has apparently excited top military officials is that of automating current enemy-location systems. These systems compare the times when a gunshot is detected by various microphones in different positions. By this means, it is possible to triangulate the location of the shooter with extreme accuracy. The idea of adding an automated gun mount and allowing it to instantly exact "massive retribution" seems to be very attractive to the military.

Many find these concepts disturbing and ask who can held responsible for "war crimes" committed by malfunctioning machines.

In 2005, the New York Times said, "As the first lethal robots head for Iraq, the role of the robot soldier as a killing machine has barely been debated. The history of warfare suggests that every new technological leap—the longbow, the tank, the atomic bomb—outraces the strategy and doctrine to control it."

The political ramifications of the robotization of the military have also been pondered by some. Robert Finkelstein, president of the military contractor Robotic Technology, wonders, "If you could invade other countries bloodlessly, would this lead to a greater temptation to invade?"

New Statesman was even more forthright, saying that the entire concept "helps perpetuate the biggest and most dangerous myth of all, which is that technical and military solutions can somehow magic away resistance to George W. Bush's geopolitical project."

But the effort is forging ahead.

George Johnson, the leader of the robotics program at the Pentagon's Joint Forces Command Centre, recently told reporters asking about autonomous robots, "The American Military will have these kinds of robots. It's not a question of if, it's a question of when."

With the largest military project in American history underway, and the issue currently so sensitive, it will probably be many years before the full story of SWORDS is finally known.

Perhaps, ultimately, it doesn't matter what killed the first robot soldier—an entire army will be arriving to replace it soon enough.

Select Bibliography

Army-Guide.com. "First MAARS Robot Shipped to U.S. Military." June 2008. http://www.army-guide.com/eng/article/article_1077.html.

BBC News. "US plans 'robot troops' for Iraq." 23 January 2005. http://news.bbc.co.uk/2/hi/americas/4199935.stm.

Crane, David. "Robo-Soldier Ready for Combat Deployment to Iraq for Urban Warfare/CI Ops." Defense Review, 3 March 2003. http://www.defensereview.com/modules.php?name=News&file=article&sid=704.

Crane, David. "Armed/Weaponized Infantry Robots for Urban Warfare and Counterinsurgency Ops" Defense Review, 13 December 2004. http://www.defensereview.com/modules.php?name=News&file=print&sid=657.

Foster-Miller.com (website of Foster-Miller, Inc.).

Garreau, Joel. "Bots on The Ground: In the Field of Battle (or Even Above It), Robots Are a Soldier's Best Friend." Washington Post, 6 May 2007. http://www.washingtonpost.com/wp-dyn/content/article/2007/05/05/AR2007050501009_pf.html.

Graham, Stephan. "America's robot army." New Statesman, 12 June 2006. http://www.newstatesman.com/200606120018.

iRobot.com (website of iRobot Corporation).

Jewell, Lorie. "Armed Robots to March into Battle." Army News Service, 6 December 2004. http://www.defenselink.mil/transformation/articles/2004-12/ta120604c.html.

Lee, Pascal. "Field Report: July 28, 2001—Haughton Crater." MarsOnEarth.org. http://www.marsonearth.org/interactive/reports/2001/HMP-sr-072801.html.

Magnuson, Stew. "Bomb Disposal Teams Deliver Blunt Talk on Robots." National Defense, July 2006. http://www.nationaldefensemagazine.org/archive/2006/July/Pages/BombDisposal2922.aspx.

Merle, Renae. "Fighting Roadside Bombs: Low-Tech, High-Tech, Toy Box." Washington Post, 29 July 2006. http://www.washingtonpost.com/wp-dyn/content/article/2006/07/28/AR2006072801462.html.

Micire, Mark. Analysis of the Robotic-Assisted Search and Rescue Response to the World Trade Center Disaster (Master's thesis). University of South Florida, 2002. http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=2DD607CEEC65DF011086BBE9140AA371?doi=10.1.1.2.4551&rep=rep1&type=pdf.

Nist.gov. "Department of Homelands Security: Urban Search and Rescue Performance Standards" (workshop series). http://www.isd.mel.nist.gov/US&R_Robot_Standards/.

Pope, Justin. "Looking to Iraq, Military Robots Focus on Lessons of Afghanistan." Associated Press, 23 January 2003. http://www.space.com/businesstechnology/technology/robot_wars_030123.html.

Schachtman, Noah. "More Robot Grunts Ready for Duty." Wired, 12 January 2004. http://www.wired.com/science/discoveries/news/2004/12/65885.

Schachtman, Noah. "Israel's Killer 'Bot: Safe Enough for War?" Wired, 8 March 2007. http://blog.wired.com/defense/2007/03/inlate2004_th.html.

Schachtman, Noah. "First Armed Robots on Patrol in Iraq (Updated)." Wired, 2 August 2007. http://blog.wired.com/defense/2007/08/httpwwwnational.html.

ScienceDaily. "Conflicting Signals Can Confuse Rescue Robots." 3 March 2007. http://www.sciencedaily.com/releases/2007/03/070302110944.htm.

Smith, Tanalee. "Robots Draw Rough Duty as US Combat-Tests Them In Afghanistan." Associated Press, 30 July 2002. http://www.space.com/businesstechnology/technology/afghan_robots_020730.html.

Sofge, Erik. "America's Robot Army: Are Unmanned Fighters Ready for Combat?" Popular Mechanics, March 2008. http://www.popularmechanics.com/technology/military_law/4252643.html.

Sofge, Erik. "Non-Answer on Armed Robot Pullout From Iraq Reveals Fragile Bot Industry." Popular Mechanics, 9 April 2008. http://www.popularmechanics.com/blogs/technology_news/4258103.html.

Sofge, Erik. "The Inside Story of the SWORDS Armed Robot 'Pullout' from Iraq: Update." Popular Mechanics, 15 April 2008. http://www.popularmechanics.com/blogs/technology_news/4258963.html.

Weiner, Tim. "New Model Army Soldier Rolls Closer to Battle." New York Times, 16 February 2005. http://www.nytimes.com/2005/02/16/technology/16robots.html.