Atlas-Ian, a 6-foot, 330-pound robot, gingerly picked up one leg at a time, ever so slowly traversing about 25 yards over three piles of cinder blocks. When Atlas crossed the finish line, cheers erupted from the entire team of IHMC Robotics of Pensacola.
It took the humanoid nearly all of the allotted 30 minutes to complete the task, a feat a human could have done in about a minute. But for a robot, this was a monumental accomplishment.
“We’re starting to come together and have robots do things that aren’t so farfetched anymore,” said Jerry Pratt, a research scientist at IHMC Robotics. “We’re getting to the point where it’s not a joke anymore.”
Traversing the terrain was one of eight tasks robots from around the world were attempting to complete as part of the two-day Robotics Challenge Trials that runs through Sunday at Homestead-Miami Speedway. The free trials run from from 8 a.m. to 6 p.m. and are open to the public.
While the robots were super slow, and at times, clumsy, the hope for the future is that these machines will develop into reliable first responders that can go into places too dangerous for humans.
To fast track the advancement of robots in this critical area, the U.S. Defense Advanced Research Projects Agency created the competition, doling out tens of millions of dollars to teams for research and development and offering a $2 million prize to the winner.
The idea for the competition was born following the 2011 Fukushima-Daiichi nuclear power plant disaster in Japan, said Gill Pratt, DARPA’s program director for the Robotics Challenge.
After an earthquake-created tsunami struck the Japanese plant, the backup power systems needed to cool the reactors failed. Workers went into the buildings to try to vent the gas that was accumulating inside the reactors, but the high levels of radiation forced them back.
DARPA quickly sent robots the company had designed to disarm improvised explosive devices in Iraq, Pratt said. But, by the time they arrived and plant personnel taught how to use them, it was too late to prevent the explosions, meltdown and radioactive releases.
But, even if the robots had arrived more quickly, they would not have been technologically advanced enough to perform many of the tasks needed in the nuclear plant’s chaotic environment.
“It spawned the idea that if robotics are almost good enough to handle these emergencies in real time, can we push the state of the art from inside the lab, where robots are used until now, to out in the field?” Pratt said.
Seventeen teams are competing in the trials, including ones from China, Hong Kong, Japan and South Korea.
International teams are welcome because the competition is to create robots for the Department of Defense’s non-military mission of humanitarian assistance and disaster response, Pratt said.
And after day one, team Schaft of Japan “is kicking butt,” according to one safety official.
Its bright blue robot, called S-One, punched its limb through the triangle shape it had just cut on the wall using a power tool. The 5-foot, 209-pound robot also hooked up a hose, walked through doors, turned valves and removed debris – slowly but deliberately – to easily take the lead with 18 out of a possible 20 points.
MIT’s entry, called Helios, was second with 12 points. No other team was in double digits.
Before the competition began, Pratt said he wouldn’t be surprised if all teams scored less than half of the possible 32 points.
“We’re just at the beginning of this,” he said. “The analysis I like to make is think of a 1-year-old, and I happen to have a 1-year-old. You can’t ask a 1-year-old to go mow your lawn, but they can kind of walk a little bit. They also fall down a lot. They can kind of pick up things. But they drop a lot.”
It’s not like in the movies, where robots make leaping tall buildings look easy or in factories, where robots can work extremely quickly with efficient precision.
Engineers and computer software gurus at this competition say getting a robot to move on its own, or simply open a doorknob or turn a valve is extremely challenging.
“Most robots in a factory today are bolted to the floor, literally,” said Seth Teller, one of the leaders of MIT’s team. “This thing is built to move its body to where the work has to be done.”
Factory robots work in a perfectly structured environment created by engineers. They are not operating in disaster zones.
MIT’s Helios demonstrated how deliberate a robot has to be in the real world. The movements were being controlled by two operators sitting in front of computers about 20 yards away in one of the garages where NASCAR teams usually set up at the speedway. They could not see the 6-foot Helios, but could see what the humanoid saw through cameras and a sensor that measures distance by illuminating a target with a laser and analyzing the reflected light.
“The operators will make decisions about where to go next and convey those decisions back to the robot in the form of commands,” Teller said. “The robot will then execute those commands using its own sensors.”
The robot has interior and inertia sensors to help guide it.
“Most factory robots don’t have much perception and blindly repeat the same action over and over again,” Teller said.
Because these robots are in their infancy, Pratt said the teams are allowed to use safety rope devices - like the belays used by mountain climbers - to prevent the multi-milllion dollar projects from falling and getting damaged. But they cannot be used to support the robots.
Those safety ropes were used often on the first day of the trials. It caught Hong Kong University’s robot while it was attempt to walk over the cinder blocks. One misstep and it began to fall.
“We were having some network issues,” said Chris Swetenham of team HKU. “And we had problems with the wind. The Atlas robots need to calibrate their motion units when the start up and it needs to be still. The wind made it difficult.”
The team did not have to contend with wind in the lab. “Maybe we should have designed it with sails so it could fly over it,” Swetenham joked.
For the trials, teams also are using umbilical chords for communication and power. To make the tasks realistic, Pratt said communications between the operator and robot will be degraded every other minute.
Now teams are allowed only as much computing power as they can get with 1000 watts of power, about the amount needed to operate a hair dryer. “The amount of computing power is a tiny fraction of 1 percent of the computing power of the human brain,” Pratt said.
The competition, which was announced in April 2012, began with more than 130 teams from eight countries. They included private businesses large and small, universities and government agencies. Teams could enter with an entire robot, or just the software.
After critical design reviews for the entire robot category, and a simulation competition for the software companies in June, the field was narrowed to 14 teams that would receive continued government funding for the trials. Four other teams are competing with private funding.
Six teams have received $3 million so far to build an entire robot, including the California-based NASA Jet Propulsion Lab. Its entry is Robosimian, which has four legs like an ape and does not use energy to maintain balance since it almost always has three legs on the ground like a tripod.
While Helios is noise-piercing loud, RoboSimian is extremely quiet, until it is commanded to do a task. “All of a sudden it wakes up and brings the rest of its systems up,” said Brett Kennedy, a principal investigator with the team. “It’s a little eerie to see a statue come to life.”
Another seven teams that performed the best in the software simulation are also competing with the hardware of Atlas robots, built by Boston Dynamics and funded by DARPA. In addition to use of the robot, those software-only teams have received up to $1.125 million.
At the end of the trials, eight of the 13 teams that already are receiving government funding will receive $1 million more to use for the finals in December of 2014. But all teams can compete in those finals and be eligible for the $2 million grand prize.