Next June, several robotics teams will converge on a testing ground at the Fairplex facility in Pomona, California, to compete in the American Idol of robot contests, the DARPA Robotics Challenge finals.
It’s a massive effort to rapidly advance the field of search and rescue machines and owes much of its inspiration to the devastating earthquake and tsunami that hit Japan in 2011. In fact, Japan, South Korea, and the European Union will all be sponsoring teams to compete in the effort.
Why is Japan competing in a U.S. robot competition? On its face, the dangerous situation presented by the Fukushima Daiichi nuclear disaster seemed perfectly suited to Japanese technical expertise. But when a wave took out the backup generators at the power plant, triggering a massive meltdown event, Japan — a country famous for advanced robot technology — had no machines capable of operating in the dangerous environment.
The Massachusetts-based iRobot company donated four robots (two PackBot 510s and two Warrior 710s) to assist. The robots were able to provide video from inside the power plant, but were unable to execute any of the important human functions to slow the meltdown or manage plant operations.
The facility has since seen a parade of robots go through, but in the critical hours when the emergency first erupted, the mother of all robot tests, the robots failed.
DARPA program manager Gill Pratt took members of several of the competing teams to Japan recently, an experience that he called “tremendously moving.”
He said the obstacle course for this final portion of the Robotics Challenge will be significantly tougher than what the teams had to contend with during the trials phase. For one thing, they’ll have less than hour to complete eight different, disaster-related tasks, like accessing controlled areas and climbing rubble, as opposed to facing those tasks separately and having thirty minutes for each of them. And bulky power cords won’t be allowed. Each robot will have to be battery-powered and energy self-sufficient. If a robot falls, it will need to be able to get up. Most importantly, the robots will perform in what Pratt describes as a communication-degraded environment. This will require each team to improve the autonomy of their robot. In a sense, the human researchers won’t have nearly as much opportunity to steer, or tele-operate, their machines the way drone pilots steer drones. The robot will have to do much more on its own.
The exact obstacles have yet to be determined but will be heavily influenced by the events at Fukushima.
“We learned a whole lot” on that trip, Pratt said. “Certainly, a lot of what you’re seeing is inspired by what happened, including the bad communication. That was inspired by the fact that the reactor buildings have thick concrete with re-enforcement rods in them that tends to shield wireless and the big platting, made the [communications] very, very bad there.”
The trick now is to keep from making the simulation for the challenge resemble Fukushima too closely. “We don’t want to overfit,” Pratt said. He described how he and other researchers had been talking with first responders and other emergency personal to find broad themes related to disasters to help them better design the challenge.
Communication was a debilitating factor not only at Fukushima but also in the aftermath of 9/11, where New York City police and fire had different radio systems. The recent ferry capsize in South Korea is a case of robots faced with the task of navigating difficult territory.
History is, of course, littered with case studies of disasters. But none represent quite so Carthagenian a challenge as does Fukushima. The events of March 11, 2011, provide a textbook worst-case scenario, a concatenation of emergencies. Much of the area, indeed much of the north-eastern portion of the country, was only partially accessible after the massive wave. But accessing the power plant was an urgent necessity, a job that became more deadly by the moment.
Researchers now need to find “the goldilocks zone in terms of how hard [the challenge] should be,” said Pratt.
The entire cost to the U.S. government (excluding the Japan, South Korean, and EU sponsorship) of the Robotics Challenge is $95 million, with qualifying teams each getting about $1.5 million to develop their robots over the entire course of the challenge. The winning team will get another $2 million dollars in June.
It’s research that will complement what’s going in the private sector. The leading team, SCHAFT, withdrew from the event, according to DARPA. It was acquired by Google last year. It’s all evidence of how quickly the field is advancing, according to Pratt. “Because we [at DARPA] take high risks, things often don’t turn out as well as we hoped. This happens to be one of the times when things went better,” he said. “The mobility was better than we thought it would be [during the trials]. The grasping and manipulation of the arms was better than we thought it would be.”
If DARPA and the roboticist teams succeed in pushing the field forward, improving the ability of robots to do more on their own, the events at Fukushima Daiichi will be at least partially the reason.