Human-robot Teams Compete June 5 at DARPA Finals
By Cheryl Pellerin
DoD News, Defense Media Activity
WASHINGTON, May 27, 2015 In eight days, 25 human-robot teams will compete on the rubble-strewn field of a mock disaster, the robots driving cars, using tools and communicating with their human partners over degraded communication links, just like in a real disaster.
Team KAIST, from Daejeon, South Korea, and its robot DRC-HUBO negotiate mock rubble at a test site March 6, 2015. DARPA photo
(Click photo for screen-resolution image);high-resolution image available.
The final round of the Defense Advanced Research Projects Agency’s Robotic Challenge, or DRC, will be held in Pomona, California, June 5-6. The challenge is a $3.5 million competition in which human-robot teams from 25 of the world’s top robotics organizations try to complete a simulated disaster-response course in the shortest time.
Each robot will have an hour -- under its own battery power and with no help in staying upright -- to drive 100 meters, get out of the car, open a door in a building, close a valve, use a tool to cut a hole in a wall, perform a surprise task, negotiate difficult terrain, exit the building and climb stairs to finish.
If a robot falls down, the DRC rules say, it must be able to get back up without help of any kind.
Hardest Test for Robots
To observers the robots may seem to move slowly, DARPA officials say, but the tasks they face represent some of the hardest tests of robot software and hardware ever attempted.
During a recent teleconference with reporters, DARPA Program Manager Dr. Gill Pratt said the DRC program began three years ago to improve robot disaster-response capabilities.
“The Fukushima disaster, caused by the earthquake and the tsunami and then the meltdown at the power plant, was a great inspiration for us,” he said, referring to efforts DARPA made after the 2011 Japan earthquake and tsunami to send robots whose development the agency had funded to the disaster zones.
“We don’t know what the next disaster is going to be but we know that we have to develop technology to help us address these kinds of disasters,” Pratt said.
Emergency Response Technology
Among the different disaster technologies, DARPA focuses on technology for responding during the emergency part of a disaster, during the first day or two, he said.
“This is not about, for instance, robotics for [restoring] the environment many weeks or years after a disaster, but rather the emergency response at the beginning,” Pratt added.
Robots have been around for decades, working in factories and cleaning floors, so, Pratt asked, why it is it necessary to develop new technologies for a disaster?
“The real answer is that when you have a disaster, one of the first things that occurs is a degradation of communications,” Pratt said.
During the DRC finals, observers at the public event will see 25 robots that are impressive mechanically, he said.
“Some of them look like an imitation of a person, some may look like some kind of four-legged creature –- there are all different shapes and sizes,” Pratt said, adding, “but that’s not the most important part of the technology we’re trying to improve.”
The critical goal is to improve how people and robots work together when they’re separated by a significant physical distance and the communication link between them is severely degraded, he said.
During the finals, Pratt said, “we will turn off communications for a significant fraction of a minute very often during the challenge.”
Beyond Physical Robots
“When you think about the DARPA Robotics Challenge, try to think beyond the physical robots that are there and focus on this very sporadic, very degraded communication between people and machines working together as partners,” Pratt said.
Because of degraded communications, the robots must have enough intelligence, for example, to open a door on their own rather than having the human partner tell the robot what to do every second.
But the human partners need tools as well, he added, “to give them situational awareness as to what is going on in the danger zone where the robot is operating.”
Half or more of the software’s computer science, or artificial intelligence, does not go into the robot but into the human interface -- the computers that human operators use to visualize what’s going on where the robot is, despite disrupted communications, Pratt said.
High Risks, High Rewards
All that computer software, he said, “is being used to help the effectiveness of both partners in this collaboration -– the human partner and the robot –- do something effective to mitigate a disaster during the first day or two.”
Pratt says observers will see a substantial fraction of the robots have difficulty as the 25 teams run through the course.
“We do that on purpose,” he said.
“DARPA takes high risks for high rewards,” he added, “and that means we also have a lot of challenges that we expect our performers to have.”
The challenge is quite hard, Pratt said, “ ... [but] we are expecting, or hoping at least, that some of the best teams will manage to do most if not all of the tasks.”
(Follow Cheryl Pellerin on Twitter: @PellerinDoDNews)