The Defense Advanced Research Projects Agency – the government agency that funds future technology – has awarded $3.4 million to two Cambridge groups to build speedy miniature drones that function as scouts for first responders.
Some day, the idea goes, tiny flying robots will zip into burning, crumbling buildings and scope out the space before firefighters and first-responders enter.
It’s all part of DARPA’s Fast Lightweight Autonomy program that the agency announced in December last year.
In a request for proposals DARPA compared the agility required of their machines to nature’s aerial champs: “Birds and flying insects maneuver easily at high speeds near obstacles. The FLA program asks the question ‘How can autonomous flying robotic systems achieve similar high-speed performance?'”
In structures or areas with a failing infrastructure, the telecommunications link between an pilot and their remote-operated vehicle is likely to be severed. So the drones, said DARPA in its challenge, would have to fly themselves.
Stepping up to the plate is a team from the Draper, a nonprofit research facility in Cambridge, and a group at MIT, who will collaborate on building the software systems that will allow the drones to be swift, but also autonomous.
The leader of the MIT team is professor Nick Roy, back on campus after a hiatus helping Google get its delivery-by-drone endeavor, Project Wing, off the ground.
“They have to be lightweight as possible and as quick as possible that’s really the crux of it,” said Julius Rose, the Draper Labs program manager who will lead that group’s involvement in the study.
Off-the-shelf hobbyist drones that can be guided by remote control can already hit the target speeds of 45 miles per hour envisioned for this study. But lose the human controller, and even the most autonomous of the lot can’t travel nearly as fast.
The ability to navigate precisely while flying at high speeds requires sophisticated and intense computational power, Rose explained. “They generally have to move quite slow in order to make the appropriate calculations,” said Rose said. Software to support the kind of zippy flight that DARPA has in mind doesn’t exist – yet.
The project kicked off last month and is funded for up to three years.