UPDATE 1-Robots to the rescue at international trials in Florida
By Zachary Fagenson
HOMESTEAD Dec 21 (Reuters) - As a squat, red-and-black robot nicknamed CHIMP gingerly pushed open a spring-loaded door a gust of wind swooped down onto the track at the Homestead-Miami Speedway and slammed the door shut, eliciting a collective sigh of disappointment from the audience.
The robot, developed by the Tartan Rescue team from the National Robotics Engineering Center at Carnegie Mellon University, was one of 17 competing in the U.S. military's Defense Advanced Research Projects Agency's (DARPA) Robotics Challenge.
The agency, which funded basic science research for now commonplace technologies like the Internet and global positioning satellites, hopes the competition will spur the development of robots that can work in places too dangerous for humans.
The challenge was launched in 2011 in response to the meltdown of Japan's Fukushima-Daiichi nuclear power plant after it was hit by a massive earthquake-spawned tsunami. Nearly 160,000 people were forced to flee the area.
The backup power systems needed to cool the plant's reactors failed and an emergency team from Tokyo Electric Power Company was unable to enter the damaged reactor building due to the intense radiation.
DARPA sent robots designed to disarm improvised explosive devices in Iraq to Japan, yet by the time workers were trained to use them it was too late to prevent a nuclear meltdown.
"What we realized was ... these robots couldn't do anything other than observe," said Gill Pratt, program manager for the DARPA Robotics Challenge. "What they needed was a robot to go into that reactor building and shut off the valves."
Hydrogen continued building in the days that followed, fueling a massive explosion.
During the two-day trials at a south Florida professional race car track, the platoon of robots faced obstacles designed to mimic the challenges following a disaster. Robots had to cut through a reinforced concrete wall, navigate debris-strewn terrain and locate and turn off leaking valves. Officials from DARPA also disrupted the link between robots and their operators, further simulating a disaster.
The eight teams with the highest scores will be awarded $1 million in funding to prepare for the final round in late 2014, where a winner will take $2 million.
While Carnegie-Mellon's CHIMP eventually opened the door, leading the field on Saturday was a two-legged robot from Japan's team SCHAFT, which finished first in the test, according to the DARPA Challenge website.
The Florida Institute for Human and Machine Cognition, based in Pensacola, Florida, took second place. Third went to Carnegie Mellon and CHIMP.
Successes in the challenges are about as common as failures. Many robots tumbled off an industrial ladder designed to test sight and balance.
"Murphy's law is very big in robotics," said Daniel Lee, a robotics professor at the University of Pennsylvania and program director for Team THOR, an agile, human-form robot, whose acronym stands for Tactical Hazardous Operations Robot. "It's very difficult to account for all of the uncertainties that you're going to face," he said.
A handful of teams, including ones from the Massachusetts Institute of Technology and Lockheed Martin, used a six-foot-two-inch, 330-pound humanoid robot named Atlas that DARPA contracted from Boston Dynamics, a company that was spun out of MIT in 1992 and recently acquired by Google.
A team from NASA's Johnson Space Center competed with a robot called Valkyrie covered in white plastic and vinyl, looking like a human wearing a robot suit.
Some robots looked highly mechanized, while others had four legs and resembled a dog.
"The goal is to make it comfortable for people to work with and to touch," said Christopher McQuin, NASA's chief engineer for hardware development.
After the final round next year Pratt said there are plans for another robotics challenge, possibly to be hosted in Japan.
For the next advance in robotics, he said, "the amount of intelligence inside the robots needs to be able to handle small tasks."
"We don't want to burden human operators with saying put your foot here, put your other foot here, put your hand there," he added.