ROBOTICS AND AUTONOMOUS SYSTEMS

DARPA Adds Difficulty, Expands Scope of Disaster-Response Robot Challenge

6/26/2014
By Yasmin Tadjdeh
When the Pentagon’s shadowy research laboratory last year asked for a humanoid robot that could perform dangerous tasks in disaster zones industry and academia delivered beyond expectations.
 
Android entrants in a trials phase of the Defense Advanced Research Projects Agency’s Robotics Challenge held in 2013 were so successful in completing various prescribed tasks that the agency plans to crank up the difficulty of the final test, program manager Gill Pratt told reporters June 26.
 
Teams that qualified for the finals will receive an additional six months to beef up their robots’ abilities to perform in emergency scenarios that are hazardous to human first responders. Teams who won funding during the trials will also receive a 50 percent increase.
 
"Things went better than we expected at the DRC trials … and because of that success and a number of other factors, we are actually changing the scope of the program a little bit," Pratt said. "We're actually going to try and raise the bar even more than what we had planned from the beginning."
 
Because the challenge will be more difficult, teams will need more time to prepare, he said.
 
During the trial phase, DARPA tasked 16 teams with building a humanoid robot that could perform a series of disaster response tasks. The trials were held in December at the Homestead-Miami Speedway in Florida. The top eight teams received up to $1 million each to continue their work.
 
Because the development period has been increased from 12 months to 18 months, each funded team will receive $1.5 million to finalize their robot designs, Pratt said.
 
The winning team will receive a $2 million funding award from DARPA, which has invested a total $95 million in the program to date, he said.
 
Eleven teams from the trials are prequalified to take part in the finals. In February, DARPA will hold a qualifying round for new entrants, including a number of prospective international participants. Pratt speculated that there might be 24 entrants in total.
 
Goals for the upcoming competition include robots that are more robust, have better stability and more advanced autonomy, Pratt said.
 
Teams will have to successfully complete eight rounds of exercises in one hour. They will be ordered from the easiest to most difficult, and requirements for one of the rounds will be a surprise, he said.
 
"We are trying to slowly move things so that we're closer to a more authentic test to what a real disaster is going to be like, " Pratt said.
 
Robots will have to complete the successive tasks without human intervention and with degraded communication. Additionally, they must not physical connections to power cords or wired communications, he said.
 
"We think it's going to require quite a bit of innovation on the teams in order to adapt to our adjustment of the goals here. We are sort of raising the bar. So in order to let them get there we are going to give them more time … to get that done," Pratt said.
 
Ultimately, DARPA hopes to spark the innovation of robots that can perform dangerous tasks in the aftermath of disasters. Once the competition is complete, he hopes commercial companies will take the resulting technologies, adapt them to a wider market, and make it more affordable.
 
"One of the big limiting factors in robotics for disaster response is actually something that is true for any technological equipment in the disaster field, which is that typical budgets to handle disasters are very low and fire departments, rescue squads … have to be able to afford the systems we come with," Pratt said.
 
The final challenge to be held in Pomona, California — initially scheduled to take place in December — have been moved to early June 2015 to accommodate the changes.

Topics: Robotics, Unmanned Ground Vehicles, Science and Engineering Technology, DARPA

Comments (0)

Retype the CAPTCHA code from the image
Change the CAPTCHA codeSpeak the CAPTCHA code
 
Please enter the text displayed in the image.