Robots Go to War

Within 10 years, infantry soldiers will go into battle with autonomous robots close behind them. One day, they'll be fighting side-by-side

The squat, four-wheeled Robot driving itself through densely wooded terrain looks too macho to be cute, but it’s too small to be threatening (picture a cross between R2-D2 and a Jeep). “You start to associate personalities with each of them,” says Mark Del Giorno, of his ‘bots. But still, Del Giorno, the vice president for engineering at General Dynamics Robotic Systems, which built this machine for the Army, insists that he doesn’t anthropomorphize his robots: “You realize that the ‘personality’ comes from, say, the steering being a little loose. I guess I’m too close to the code to think of them as people.”

Nearly one million lines of code running through four onboard computers define the algorithms that allow XUV 12 (for”experimental unmanned vehicle”), as this robot is called, to navigate autonomously from point A to point B without clobbering a boulder, speeding off a cliff, or hammering a tree-which is exactly what it seems about to do right now, on a test loop at Fort Indiantown Gap, in central Pennsylvania. Suddenly the ‘bot darts to the right and jumps off the dirt trail it had been following.”This is good! This is good! Off is good!” Del Giorno enthuses from the back seat of the pickup truck he’s using as a chase vehicle. After studying the laptop computer balanced on his knees, he realizes that XUV 12 has left the road because it’s come up with a more direct route to its destination. Unfortunately, this new route runs directly through more trees. So the ‘bot slams on the brakes, dust pluming around its tires. Then . . . nothing.

Del Giorno is unfazed.”He’s gotten to the point where he’s saying, â€Hey, this is a dumb decision. I never should have gone this way,'” he explains, clearly anthropomorphizing the ‘droid in spite of himself. XUV 12 inches back and stops again.”OK, he’s taking another look at the situation,” Del Giorno says. The sensor pod mounted on top swivels right and left in disconcertingly lifelike fashion as the ladar (laser detection and ranging) system paints a 3-D image of the world directly in front of it. On his laptop Del Giorno examines what the robot is”seeing”-a color-coded map studded with trees represented by red no-go zones.”Now he’s trying to plan a hard-left turn,” Del Giorno says. Sure enough, the sensor pod rotates to the left, and XUV 12 snakes between a pair of trees. When it gets to the edge of the trail, it pauses again, troubled by the steep, slippery terrain.”Behave yourself!” Del Giorno mutters.”Don’t be a sissy!” On cue, the ‘bot climbs up on the trail and trundles off.

“Oh, that was awesome!” Charles Shoemaker says from the front seat. Shoemaker runs the Robotics Program Office at the Army Research Laboratory. He’s the military’s go-to guy for the robots known generically as unmanned ground vehicles, or UGVs, and he understands better than anybody just how difficult it is to create vehicles that think for themselves rather than being operated by humans via remote control.”It’s really, really hard,” he acknowledges.”But I’m convinced that we’re going to develop systems that work for a whole range of tactical missions, from patrolling a storage area to performing extremely challenging reconnaissance. And I don’t think it’s going to take another 20 years, either.”

Steps to Lethality

We could be at the dawn of a golden age of military UGVs. According to the 2001 Defense Authorization Act, one third of all operational ground vehicles are supposed to be unmanned by 2015. Therefore, the Department of Defense’s ambitious Future Combat Systems (FCS) program calls for the development of three semiautonomous UGVs that can perform certain tasks entirely on their own, and others-firing weapons is the biggie-only while being operated by a soldier. The runt of this FCS litter is the Small Unmanned Ground Vehicle. Weighing less than 30 pounds, it will be carried in a backpack and used like a disposable scout. It’s designed for high-risk missions in urban environments, such as searching through sewers and dealing with toxic chemicals. Bigger jobs are left to the Multifunctional Utility/Logistics and Equipment vehicle, which will travel with infantry units carrying supplies. The third FCS robot, the five-ton Armed Robotic Vehicle, will pack a machine gun, an automatic cannon and beyond-line-of-sight missiles-all remotely operated. In short, a bristling ‘bot with a nasty bite.

But the three FCS vehicles are just the tip of the military’s robotic iceberg. UGVs are uniquely suited to hauling giant loads of cargo in drone convoys. Ladar gives them a leg up on manned vehicles when operating at night and in rough terrain. Others are being designed to function as battlefield ambulances, to patrol supply depots, set up mobile communications links, and who knows, maybe even serve freshly frothed cappuccino with reveille. But will they ever actually kill on their own? Right now, the party line is that there will always be a human in the loop before lethal action is taken-that a robot will never decide on its own to fire a gun or cannon, or light a missile.

But some observers sense that robots’ lack of emotions will eventually be taken advantage of.”Part of the process of creating soldiers is disinhibiting people from killing,” says John Pike, director of, a military-policy think tank.”Robots have no such inhibition. They will kill without pity.”

Trial by Fire

Twenty years ago, the Defense Advanced Research Projects Agency (Darpa) inaugurated an Autonomous Land Vehicle (ALV) program. Right idea, wrong era. Then, the only way to harness enough computing power for autonomous operation was to stuff half of a schoolbus-size vehicle with Silicon Graphics workstations. Even so, the ALV couldn’t go more than a few miles an hour without leaving the road or smacking into something. Clearly, this wasn’t what the Department of Defense had in mind. Unmanned aerial_vehicles (UAVs) were taking off, but UGVs appeared to be about as dead as the also-rans in a _Battlebots telecast.

The robotics world looks a lot rosier these days. The accuracy of commercial-grade GPS navigation has improved from 100 feet to about 30-or, in some military applications, only a few feet. New sensors generate denser images that are approaching one million pixels in resolution, rather than a mere 100,000. This, in turn, allows ‘bots to”see” with reasonable clarity for 50 to 100 yards. Meanwhile, more elegant algorithms enable faster onboard computers to consider thousands of routes as often as 10 times a second. Roll all this into a single package, and you’ve got a UGV that can safely achieve speeds of 35 miles an hour on obstacle-laden cross-country jaunts; 55 mph is possible on well-marked highways.

The first generation of small, crude remotely-operated (as opposed to autonomous) UGVs has already acquitted itself on the battlefield. The Talon, a 100-pound treaded ‘bot built by defense contractor Foster-Miller, was first used in Bosnia in 2000. At the moment, there are several hundred Talons and slightly smaller PackBots, built by iRobot (of Roomba robotic vacuum cleaner fame), on the ground in Iraq and Afghanistan. Both carry cameras and articulated arms to deal with roadside bombs and booby traps. The Army has tested a Talon variant that can be equipped with any one of four weapons ranging from an M16 rifle to a medium-weight machine gun, although it hasn’t yet blessed it for combat.”The operator sees a bore-sighted image with a crosshairs,” says Bob Quinn, Foster-Miller’s general manager for Talon operations.”It’s exactly like a videogame-except it’s real.”

But the real goal for UGVs is autonomous operation with no human input whatsoever and no dependence on unreliable wireless linkages. This is, after all, standard operating procedure for many UAVs. You would think it would be easy to apply this technology to ground ops. But you wouldn’t have thought so if you’d been in the California desert in March 2004, during the inaugural running of the Darpa Grand Challenge. Created to spark grassroots UGV development, the first Grand Challenge came off as something of a Three Stooges affair. None of the entrants came anywhere near completing the 142-mile cross-country course, and most of them barely made it past the starting line. The Oshkosh Truck Corporation showed up with an autonomous version of its six-wheeled Marine Corps truck dubbed the TerraMax, but the bright yellow beast was terminally flummoxed by a software glitch after just one mile.

What a difference a year of development made. Many of the participating teams returned with vastly improved systems that made the best of the learning experience that the first race really was. During prerace testing for the second Darpa Grand Challenge, the TerraMax hightailed it at 35 mph through the scrub near Barstow, California, with its steering wheel magically moving back and forth as if it were being manipulated by the Invisible Man. During the actual race, staged in October, the truck was one of five vehicles to complete the 131-mile desert course. The $2-million prize was claimed by an autonomous Volkswagen Touareg SUV, called Stanley in honor of its alma mater, Stanford University.”Yes, we were competing against each other,” says Sebastian Thrun, leader of the triumphant Stanford Racing Team (and a 2005 Popular Science Brilliant 10 winner).”But all of us had to defeat nature. The real victory is that five vehicles finished. I think this will prove to be a pivotal moment in the history of transportation.”

Stanley averaged 19 mph over the race and hit a top speed of 38 mph. It was equipped with five ladar sensors that reached an average of 85 feet and a single camera with a range of 260 feet. The biggest challenge the bot faced during the race was a flock of birds that confused it by temporarily occupying a lakebed it was crossing. Nevertheless, Thrun is convinced that a version could be readied for military-convoy duty almost immediately and that civilian highway applications are just a matter of time.”If nothing else,” he says,”the Grand Challenge showed people that it’s possible.”

Ladar and stereo cameras proved the most effective sensors for the Grand Challenge entrants; the jury is out on which is more promising. Other UGVs feature systems that detect color, heat and texture. Eventually, the goal is to be able to differentiate not only between wispy and solid, and living and inanimate, but between friend and foe.”If it’s a child, you want to stop,” says Donald Verhoff, Oshkosh’s executive vice president of technology.”If it’s a guy with an RPG-7 [grenade launcher], you want to run him over.”

But better sensors create more data to process, which increases the time it takes to make decisions, which forces the ‘bot to slow down. This places a premium on software that”stovepipes” directly to the correct solution rather than working through an endless series of possibilities. The most practical way to do this is to code up algorithms that establish a rules-based hierarchy covering every situation that might arise. Obviously, this requires a boatload of if-this-then-that commands. And even so, UGVs are bound to encounter circumstances that weren’t anticipated by the programmers.

At Darpa, Larry Jackel, who manages three of the agency’s ground- robotics programs, is spearheading a project to program UGVs with behavioral cues that will allow them to respond appropriately to unexpected situations and learn from experience.”We believe the problem is too complex to write an algorithm to traverse every piece of topography,” Jackel says.”A robot’s got to be like a child, who learns to crawl, then walk, and then run.” But this approach remains highly experimental, and it seems unlikely to be found on the next generation of military UGVs.

Money and Ethics

General Dynamics Robotics Systems president Scott Myers walks briskly through what appears to be Santa’s workshop for robotics geeks. Here in the cavernous GDRS facility near Baltimore, technicians are assembling a fleet of experimental UGVs. Myers paternally pats the head-well, the sensor pod-of a warehouse patrol ‘droid, but he lingers the longest around the Tactical Autonomous Combat-Chassis, a jeep-like ‘bot that’s the hot rod of the UGV world.

Equipped with a 150-horsepower Volkswagen turbo-diesel engine and a suspension inspired by off-road racing vehicles, the TAC-C can scoot cross-country at speeds up to 80 mph. But what makes it really special is the state-of-the-art perception suite-ladar as well as stereo, color and infrared cameras, all working together to take an 400,000-pixel snapshot of the world 30 times a second.”We can do 60 percent of what human [scouts] can do now, and will be able to do 90 percent by 2010,” Myers says.”These vehicles are going to be a lot cheaper than you think, and they’re going to be a hell of a lot cheaper than replacing a human.”

Military UGVs make so much sense for so many missions that they seem to be a no-brainer.”I’ll take one of these things any day,” says Sergeant First Class Ralph Brewer, who served in a tank in Iraq, during a break in testing a UGV. “I can’t believe we don’t have them already.” But everybody in the Army isn’t as enthusiastic as Brewer. With its tradition of, well, tradition, the military doesn’t often rush to embrace newfangled ideas, especially ones that will profoundly change the way it operates. Also, Congress is wary of devoting money to a technology that hasn’t yet produced many tangible results. Ultimately, of course, a weapons system is a political issue as much as it is a military one.”It’s not a question of technology,” Myers says.”It’s a question of money.”

And ethics. There are no plans to develop fully-autonomous robot tanks that can fire lethal weapons without any human input. So in the near term, at least, the creepy specter of hunter-killer cyborgs recycled from The Terminator remains a dystopian science-fiction clich. But the technology that allows UGVs to make life-or-death decisions on their own is coming. The concern is whether we have the political will to allow our machines to do our killing for us.”It isn’t going to happen without scrutiny,” says Peter Danielson, director of the Centre for Applied Ethics at the University of British Columbia.”But the impulse to get people out of the line of fire-to protect your troops-puts pressure on you to move farther and farther away from the enemy.”

Despite all the lip service paid to keeping a human in the loop,’s Pike is convinced that autonomous lethal robots make so much sense that they’re inevitable.”They’re going to sneak up on us, just like UAVs did,” he says.”They’re going to do more and more of the toting. They’re going to more and more of the surveilling. And when they start fighting, no organized force could stand against them.”

How They Work

Need a robot to fly from A to B? No problem. Need one to charge through the mountains, dodging trees, rocks and ditches? Well . . .

Two big head-scratchers unique to autonomous vehicles are perception and path-planning. Robots must see well enough to recognize obstacles and be smart enough to avoid them. A robot has to be able to tell the difference between, say, a harmless bush and a half-ton boulder. Even trickier is a negative obstacle. Is that depression in the distance a dip in the road or an impassable ditch?

Information processing, therefore, is critical. Most autonomous systems use sensors that scan the environment to create a 3-D model of the world. Pattern-recognition algorithms identify obstacles and project them onto a 2-D map that color-codes the world according to how suitable it is for locomotion (green for “all systems go,” red for “danger, Will Robinson”). GPS and a complementary inertial navigation system, which tracks movements based on speed and turns, locate the robot and its destination on this map. Path-planning algorithms then plot the best route for the ´bot while taking hundreds of preprogrammed rules into consideration, such as: Stay on the road if possible. Avoid grades higher than 10 percent. And never-repeat, never-drive off a cliff.

Automating the Army: Step 1

Unmanned Convoys


The first military use of fully-autonomous ground vehicles will be in basic convoy operations. Large vehicles will follow known roadways to Air Force, Army or Marine Corps bases, carrying fuel, supplies and ammunition. Although convoy missions will typically be on paved roads, even the earliest autonomous trucks will need full off-road capability in case their roads become impassable.


The Oshkosh Truck Corporation’s TerraMax [above] is poised for such missions. In October it was one of five vehicles to successfully complete the 132-mile Darpa Grand Challenge, a desert race through the Mojave Desert. Equipped with ladar and stereo cameras, it uses synthetic vision and advanced navigation technology to maneuver through the most challenging off-road environments.

Automating the Army: Step 2

Battlefield Support

Advanced autonomous capabilities will bring robots into combat, keeping soldiers supplied and protected

Once autonomous vehicles can be trained to get safely from point A to point B, they will be given greater responsibilities, in actual combat environments. As sensor technology advances, battlefield support will fall to smaller jeep-like vehicles.


The Army’s Multifunctional Utility/Logistics and Equipment (MULE) vehicle will help pave the way for the vehicle shown here. The 2.5-ton MULE will carry 2,000 pounds of equipment for infantry squads. It can also be configured as a remote weapons platform or mine sweeper.

Automating the Army: Step 3

Battle Tank

Training robots to roll, tumble . . . and kill

A fully-autonomous weaponized unmanned ground vehicle-one that can distinguish enemy combatants or vehicles on its own and attack them without direct human command-remains decades away, but analysts are convinced that it will arrive. Such a vehicle will need to be able to differentiate among friendly forces, enemy combatants, and civilians with unquestionable reliability. It will also require the ability to act from mission objectives, combat tactics, and all military protocols and rules of engagement.

A group at Carnegie Mellon University is developing a platform for an unmanned vehicle, Spinner, that could evolve into a weaponized tool. Now in testing [top], Spinner uses long-travel suspension to negotiate brutal terrain. Other visions of the Spinner [above] can recover quickly if tossed upside-down-it will simply push its wheels to what was the top of the vehicle, rotate its gun to the bottom, and then drive off as if nothing had happened.

Automating the Army: Step 4

Going on Foot

Wheels are well and good, but only legged robots will conquer the other half of the Earth’s terrain


Legged robots will be able to dash through the woods and scramble over rough terrain that would stymie most wheeled or tracked ‘droids. The challenges are immense. Legged robots must master basic locomotion skills such as maintaining balance and traversing broken terrain.

Boston Dynamics is working on a variety of biologically inspired robots. Two of its four-legged ‘bots-BigDog and LittleDog-could become a soldier’s best friends.”Wheeled and tracked vehicles can get to half of the Earth’s surface,” says company president Marc Raibert.”But people and animals can get just about anywhere on foot.”

In Big Dog’s case, a gasoline engine powers the hydraulic system that actuates articulated limbs. A gyroscope and other sensors help the onboard computer plot each step. The ‘bot relies on feel to keep its balance-if a leg hits the ground earlier than anticipated, the computer assumes that it’s either hitting a rock or heading uphill, and BigDog adjusts its gait accordingly. Researchers are also creating robots that mimic geckos, cockroaches, fish and even the slithering movement of snakes.