Supersizing robotic system capabilities;
Army's collaborative effort with industry, academia break open new space for autonomous warfare
September 27, 2011
- Military, industry and academia are combining resources to upgrade fielded combat robots.
- Researchers are investigating ways to get robots to essentially and seamlessly integrate into military and civilian societies.
- Modeling human behavior in robotic systems isn't easy.
Robots on the battlefield remain among the most exciting advancements of technology's meet-up with warfare since the machine gun, but as with any other modern tech luxury, these once-fascinating Soldier must haves' need for more capabilities has been supersized.
Military, industry and academia are combining resources to upgrade fielded combat robots through proofs of concept they've designed under the Army Research Laboratory-managed Robotics Collaborative Technology Alliance, which formed in 2010. This is so teleoperated systems like Talons and iRobot Packbot can become autonomous, meaning, "a Soldier doesn't have to use a joystick to direct a robot to a location. He can treat it like another Soldier and tell the robot to go down the road and peer into a window. It would be as a good as a Soldier but if the robot gets shot, at least we don't lose a Soldier," said Dr. Jonathan Bornstein, ARL's collaborative alliance manager in Aberdeen, Md.
Already, a team of researchers who first met as part of this collaborative alliance combined technological ingenuity to morph the Talon, a man-portable tracked vehicle, into a new autonomous system for small robots, equipped with customized sensors, an in-house INS/GPS unit, a 360-degree camera system and a 360 LADAR scanner. Together, these allowed for greater obstacle detection and 3-D mapping. The new system also included upgraded power distribution boards, e-stop system, Ethernet radios, control computers and the code for running the system. It was showcased at the MAGIC 2010 competition in Australia, and placed third across the globe. Work on the autonomous Talon will continue under the Robotics CTA, Bornstein said, to enable further development of autonomous mobility technology for small unmanned ground vehicle systems.
Robot autonomy can be a touchy subject though, because many solutions, although unmanned, are actually teleoperated, even from distances as far away from military operations as Las Vegas like the Predator was during the Bosnian conflict. To arrive at a truly autonomous state, Bornstein said, robots will need to be given "basic behaviors much like we do a trained dog; we'd be doing good then."
That means, robots will need to be programmed to pperceive, understand and easily adapt to dynamic, unknown and changing environments and scenarios, independently plan and execute military missions, and learn from prior experiences like people and share common understanding with Soldiers, and eventually, other robots. Researchers are investigating ways to get robots to, essentially, seamlessly integrate into military and civilian societies beyond learning and reasoning, but also to react with near-human dexterity to do what's currently impossible like turn a doorknob or pick up a needle or maneuver through three-dimensional environments like dark caves that have rugged dry terrain, mud and water.
Modeling human behavior in robotic systems isn't easy. Some experts say translating the world into symbolic coding schemes is frankly close to untenable. But ARL engineers specializing in human factors integration are leading military efforts in cognitive robotics research to take ground-based robotic systems beyond architectures that support a robot's working memory to architectures that will support perception and long-term memory.
Here's the difference: with working memory, robots could play a game of chess or complete any other task-related function but to group that with long-term memory and perceptual systems, robots could play chess and think about the world around them to consider memories about things that are rarely ever forgotten like George Washington was a president or cats are mammals, or even more difficult to program, that a particular door - because of its size, shape, color, thickness, fixtures, etc. - is one that's never been seen before, explained Troy D. Kelley, team leader of the cognitive robotics and modeling research area.
"We can't program robots for everything; they'll have to learn from experience just like humans so we have to program them to have adaptive learning," Bornstein said. "A Soldier has to be adaptable through training and intuition to fulfill operations. Technology can satisfy this in a robotic system."