by DAVID AXE
On Oct. 27, 2004, a suicide bomber riding a motorcycle blew himself up alongside a U.S. Army flatbed truck in Balad, in north-central Iraq. The blast killed the truck’s driver, Staff Sgt. Jerome Lemon, from the South Carolina-based 1052nd Transportation Company.
Nearly five years later, at a sandy outdoor range at the Massachusetts Institute of Technology, an unassuming orange-and-black-painted forklift approached and lifted a pallet of mock munitions, as an audience of Army officers looked on. It might have looked like any day at any austere supply depot, but for one thing: the forklift had no driver.
While separated by years and thousands miles, there’s a direct link between Lemon’s tragic death and the robotic forklift’s quiet feat. From warehouses to highways to supply depots, the Pentagon is working hard to replace human logisticians like Lemon with machines that cannot be killed. After several years of intensive development, the first supply bots are just beginning to crawl and fly towards battlefields in Iraq and Afghanistan.
The “convoy war”
Lemon was just one of several hundred U.S. military logisticians to die in 2004 and 2005, during a phase of the Iraq war one might call the “convoy war.” For at least two years, Iraqi insurgents focused their attacks on America’s long, delicate supply lines. To keep front-line troops in beans and bullets, the Pentagon had deployed around 20,000 trucks and hundreds of helicopters, together accounting for hundreds of supply sorties and convoys every day carrying thousands of tons of supplies from Kuwaiti and Iraqi airports and seaports. These logistical forces were lightly armored, poorly protected and predictable in their movements, making them particularly vulnerable to attack.
The Pentagon reacted swiftly to protect its logistical flanks. The Army and Marines ordered more than 10,000 heavily armored Mine-Resistant Ambush-Protected trucks to escort the convoys, developed radio jammers to block bomb detonators and boosted the number of helicopters. Despite the cost, the Air Force added C-130 and C-17 airlifters to replace the more vulnerable ground convoys on certain routes. “We’ll do what it takes,” said then Air Force Chief of Staff General John Jumper.
But these were stopgap measures for a long-term problem. To more permanently harden war-zone supply lines, at an acceptable cost, the Pentagon has turned to unmanned systems. “If we can reduce the number of people who are driving convoys in a place like Iraq or Afghanistan, we would definitely reduce the infrastructure to take care of those people,” Tony Tether, then director of the Defense Advanced Research Projects Agency, said in 2007. Darpa took the lead in the unmanned logistics effort, with a series of programs and competitions meant to seed industry with the ideas and technologies that might produce robotic logisticians.
Those efforts are finally bearing fruit. Darpa’s much-heralded Urban Challenge robot race in Victorville, California, in December 2007, in particular, spun off hardware and software that is feeding numerous practical logistical applications, including a robotic autopilot for cargo trucks that is almost production-ready.
This Convoy Active Safety Technology (CAST) truck autopilot, co-developed by the Army’s Tank-Automotive Research Development and Engineering Center and Colorado-based tech firm Perceptek (now owned by Lockheed Martin) — and based on software developed for Urban Challenge — has demonstrated increasingly reliable “hands-off” driving for Army and Marine supply trucks. CAST uses simple sensors and data-links to allow trucks to follow each other, while their drivers focus on scanning for attackers. CAST made its latest appearance at an Army “robot rodeo” at Fort Hood, Texas, in September.
Meanwhile, the Marine Corps and Special Operations Command are driving demand for a range of simple Unmanned Aerial Vehicles (UAVs) capable of carrying small cargo loads to isolated combat outposts. This month the Air Force finally joined the Marines and SOCOM in developing cargo UAVs. The SOCOM cargo bots, including the Snow Goose bot from Canadian company Mist Mobility Integrated Systems Technology, are already in service. In August, the Marines contracted with Boeing and Lockheed for demonstrations of their respective A160T and K-MAX unmanned cargo helicopters. The Air Force is eyeing a 2011 demonstration for a cargo UAV capable of hauling up to 3,000 pounds over a 500-mile distance.
And in June at MIT, a 25-person team led by Dr. Seth Teller demonstrated their autonomous forklift, two years in development with Army funding. A forklift might, at first glance, seem less impressive than self-driving convoys and robotic cargo aircraft, but in reality Teller’s forklift represents several major advancements that should feed back into other unmanned-logistics efforts, bringing them closer to their combat debuts.
The thinking forklift
“The focus my lab is to develop the ability for machines to have a representation of their environment, in order to move and work purposefully in the world — the same way people that build up a mental model of the world, where things are connected in space,” Teller told Unmanned Systems.
His robo-forklift is a standard commercial model, weighing three tons, with a 3,000-pound lifting ability. The robot is highly autonomous, requiring only general instructions from a remote human operator, such as “move that pallet over there.” Interface is via a simple application on a Nokia smart-phone, or by voice command. The bot feels its way around using a LIDAR and optical sensors. Its computer algorithms are derived from MIT’s Urban Challenge racer.
Where the forklift differs from other robots, is in its ability to safely work in a complex environment alongside human beings, without any extra infrastructure devoted solely to the bot, Teller said. Earlier robotic forklifts traveled along special lanes inside a warehouse or workspace; they did not truly mix with human beings, for they weren’t responsive enough to avoid harming them. Teller’s bot is smart enough to sense and avoid people, while deftly manipulating pallets in what Teller calls a “semi-structured environment,” such as an ad-hoc supply depot established on short notice close to a battle zone. All it needs, to get started, is a “guided tour” of the facility, so that it can internally map out the area, noting landmarks and obstacles with its sensors. This tour basically entails a person driving the forklift around for a few minutes.
The MIT forklift’s flexibility and reliability make it useful for real-world, even combat, applications — although it might take up to two years to make the design production-ready, Teller said. In the meantime, he added, technologies refined for the forklift might find their way into other unmanned logistical systems. The result, in a few years, could be front-line supply lines that seamlessly mix an ever-shrinking number of people with an ever-growing number of machines, thus exposing fewer human beings to attack — and all without sacrificing the capacity and responsiveness of today’s manned logistical systems.
“We think lot of our techniques are broadly applicable to other robots,” Teller said. The speech interface, for one. “Another piece is the narrated guided tour,” Teller added. “There’s a major disconnect now between the languages people speak and those robots speak. Now people take pains to express an environment in terms a robot can understand — usually GPS coordinates. But people don’t think in GPS terms, they think in terms of landmarks. So we want to make robots that come all the way to people and do the bidding of people on human terms. Our robot does that.”
And so, too, might the generation of logistical robots emerging from industry and university labs, propelled by brutal lessons of Iraq’s bloody supply lines.