These Bots Might Save Your Life Someday
But the DARPA Robotics Challenge has another purpose: teaching us to love our mechanical friends.
--
In 2011 a magnitude 9.0 earthquake sent a tsunami toward the coast of Japan, where 46-foot-high waves overtook the walls of the Fukushima nuclear power plant. Overseen by corrupt and overconfident officials, the plant entered a series of preventable failures—both the primary and backup power sources failed, taking down the cooling system and allowing the reactor to grow dangerously hot.
A team of human workers entered the site to vent the accumulating hydrogen gas. But as radiation levels grew, they had to turn back. Robots were badly needed to take over where the human workers left off. Instead, the gas continued to build, until the plant exploded.
Back in 2001, Japan had built several nuclear disaster robots. But the program was discontinued the following year, after a task force found them to be limited and unnecessary. Workers at power plants didn’t like them, either, as the robots were a constant reminder that someday they might be needed.
After the Fukushima disaster, robots built by iRobot and the Japanese Chiba Institute of Technology entered the leaking plant. But all they could do was monitor radiation levels and take photographs. They were not equipped to stop the leaks, and they couldn’t have vented the gas.
Now people can agree it’s scarier not to have robots. All we need are ones up to the task. And the best way to make that happen is to pit the 24 best bots in the world against each other in a show of dexterity and guile. What’s interesting is not which bot wins — but how the world can learn from their collective existence.
Last weekend, the Fairplex in Pomona, Calif., came alive with lights and tents that almost could have been mistaken for the L.A. County Fair, which takes place there annually. Kids dashed between exhibition booths or waited impatiently in line to meet famous faces. A roar arose now and then from the crowds seated around the horse racing track.
But it wasn’t horses that were racing around the track. It was robots, carefully placing one foot in front of the other as they worked their way across a mock-disaster scene. They were in the DARPA Robotics Challenge (DRC) finals, where many of the best robots in the world had to drive and exit a car, open a door, close a valve, cut through a wall, walk over a pile of rubble and climb a set of stairs, all in less than 45 minutes.
In 2012, DARPA revealed the DRC, a three-year competition that asked entrants to build a disaster-response robot that could have entered the Fukushima plant that day and vented the gas. “Fukushima was really a great inspiration for us because we don’t know what the next disaster is going to be,” DRC lead Gil Pratt said at a May press conference. “But we know that we have to develop technology to help us to address these kinds of disasters.”
Nuclear disasters are extremely uncommon, so much so that in the early 2000s both government officials and private companies had convinced themselves that developing disaster robots was not a priority. “There’s not really a tremendous need except in rare times,” said Tony Stentz, who led a team called Tartan Rescue. “That’s a difficult demand for a company to enter the space and provide that. Something like the challenge really incentivizes people to develop these technologies anyway.”
DARPA infused the competition with significant funding, including nearly $100 million for the robots’ development and an $11 million contract awarded to Boston Dynamics to build Atlas robots, which teams could use instead of building their own hardware. The winning team was awarded $2 million. Second and third place took home $1 million and $500,000, respectively.
Spurred on by the financial incentive, 24 teams set out to build an intelligent, mobile robot. Their members came from industry and universities, from the U.S., South Korea and other countries across the world. And they dreamed up 24 unique solutions for researchers across the world to see. Between the first trials held in 2013 and the final competition last weekend, teams were able to see winning strategies and incorporate them into their own robot, according to Brett Kennedy, the team lead for the NASA Jet Propulsion Lab. That same exchange of ideas will continue based on the performances at the finals.
“In 20 years we might look back and think this is the robot Woodstock,” Kennedy said. “As far as I know, this is the largest collection of robots doing more or less the same thing the world has ever seen.”
DARPA ran its first Grand Challenge in 2004, when autonomous cars competed to drive a 150-mile course. None of the vehicles completed the challenge. However, the next year five were successful. By 2007, autonomous cars were sophisticated enough to complete a new course in an urban setting.
Now, in 2015, self-driving cars are almost within reach for all of us. It took 10 years for them to go from failing a desert course after the 7th mile to flawlessly traversing Silicon Valley’s suburban streets.
That same evolution is already apparent when you look back at the 2013 DRC trials. The robots went from completing each event in half an hour to executing the entire course in 45 minutes. Three robots completed all eight tasks, and four more scored on at least seven of them. In 2013, only three robots scored more than half the available points.
If we look forward another 10 years, DARPA officials and competitors think multi-talented robots will continue on the same path as autonomous cars. And we could see them out in the field even sooner. Stentz, whose robot CHIMP placed third in the DRC, said CHIMP could be ready to go into a disaster zone after a single rebuild. Its form and software are already good enough — the robot just needs to be made resistant to radiation and water.
Stentz says that no single component of the robots marks a breakthrough. The real innovation is in the collection of so many abilities into a single system. He sees their descendants responding to all sorts of disasters — whether fires or explosions or mine collapses. They could also find a place in manufacturing plants, where they could quickly adapt to producing new types of items.
Private companies are already taking notice. The highest-profile example is Google, which acquired Boston Dynamics and SCHAFT, the Japanese company behind the robot that won the 2013 DRC trials (and later dropped out of the competition). Google now owns the technology for two of the most capable disaster response robots in the world. It has the money to send them out when the world needs them most, but disaster response is most certainly not its final plan for the technology.
Kennedy’s team has also licensed its Robosimian robot, which placed fifth in the competition, to Motiv Space Systems. The Bay Area company already has plans for a version built for researchers, and is investigating commercial applications.
“Originally conceived for disaster response, RoboSimian is an efficient robotic platform for a wide variety of applications: serving as everything from a service robot, to a reconnaissance scout, to a field technician,” the Motiv website now reads.
At the very least, the competition was an intense course for the world in the current state of robotics. The 45 minutes the robots had to complete the course felt like ages, as they agonized over every single decision. They stared at stairs for minutes at a time and tried 10 different ways to open a door handle. They fell over — lots — and generally convinced the world that the robot uprising is far, far away.
“What this entire competition can do is show the world what practical robotics looks like and allow them to get away from the C3PO version — the robot can do anything — or the Terminator version, where the robot is too scary to let near humanity anyway,” Kennedy said.
To Pratt, that’s the legacy of the DRC. After IHMC Robotics’ robot whisked over the rubble almost effortlessly and prepared to climb the stairs, the crowd hushed, save for an occasional whoop. Everyone wanted the robot to succeed, to accomplish its absolute maximum potential. When it fell, they gasped and remarked on its admirable performance.
“There’s some incredible untapped affinity between people and robots that we saw,” Pratt said at a press conference. “Every time a robot scored a point, the crowd cheered, and they cheered despite the fact that the event was one robot going through eight simple tasks and taking nearly an hour to do it.”
With each DARPA competition comes a little bit more exposure to robots, and a little bit more understanding of what they can do and what they can’t. Observers learn that robots can be collaborators — not harbingers of trouble. They are allies.
Follow Backchannel: Twitter | Facebook
Are grand challenges the best way to develop robots? Do you agree that robots will be our partners in addressing disasters? Or are you worried that they may become the disasters? We invite you to continue the conversation by responding below.
