DARPA seeks a robot to respond to Fukushima-style disasters
WASHINGTON — The Pentagon on Tuesday put out the call for robots that can handle “dangerous, degraded, human-engineered environments” without missing a beat.
Although that sounds a lot like a description of most U.S. combat troops, the Defense Advanced Research Projects Agency, or DARPA, is actually looking for robots to operate in a disaster like the meltdown at Japan’s Fukushima Daiichi power plant last year.
The first team to produce a robot that can meet the requirements of DARPA’s Robotics Challenge will receive $2 million. Meanwhile, the agency is ready to hand out up to $34 million in research money to developers.
The winning robot will have to accomplish a series of tasks currently beyond the ability of any machine. Plans call for it to drive a vehicle before getting out and negotiating rubble-strewn ground at a mock disaster site. Then it has to clear debris from a doorway, enter a building, climb a ladder and cross a catwalk. Next, it must smash through a concrete wall with a tool, and finally, close a valve on a leaky pipe and swap out a component like a cooling pump.
“The key to successfully completing this challenge requires adaptable robots with the ability to use available human tools, from hand tools to vehicles,” said DARPA program manager Gill Pratt.
The announcement doesn’t specify a human-like robot, but negotiating objects shaped for human bodies may require it. A Defense Department push in that direction would be very different from most DOD robots, which have been mainly vehicles like aerial drones or small-treaded devices used commonly by bomb disposal units.
Anthropomorphic robots have great advantages, said Greg Trafton, a Naval Research Lab scientist who oversaw the development of a firefighting robot that, thanks to its shape, is able to use off-the-shelf gear for humans. A robot that looks like us is also just easier and more efficient to use, he said.
“A humanoid robot that doesn’t understand what you’re saying can get a confused facial expression,” he said. “It doesn’t have to be taught to people as much, because it’s not a new interface.”