Wired

Darpa’s Next Grand Challenge: Build Us Lifelike, Humanoid Robots

Imagine robots that can do everything you can do — and probably do it even better.

Brace yourself, because that era might be here sooner than you think: The Pentagon agency behind some of the most important robotics research will soon challenge experts worldwide to come up with humanoid robots that can navigate their environment and handle tools with near-Homo sapiens skill.

Within the next few weeks, Darpa, the Pentagon’s far-out research arm, is expected to launch its contest, which will likely ask roboteers to build a bipedal robot that can do things like drive cars, open doors, traverse rough terrain and show off its fine motor skills, perhaps by repairing busted pipes.

Word of Darpa’s plans was initially leaked to Hizook.com, a website that covers robotics. The site later confirmed details of the agency’s endeavor with Kent Massey, the director of advanced programs at HDT Robotics. Massey attended a recent speech by Darpa program director Dr. Gill Pratt, who outlined the new challenge. Danger Room confirmed Massey’s account with other attendees.

“The goal of this Grand Challenge is to create a humanoid robot that can operate in an environment built for people and use tools made for people,” Massey told Hizook.com in an e-mail. “The specific challenge is built around an industrial disaster response.”

This doozy of a Grand Challenge is the third such initiative that Darpa’s launched in recent years. The first three, focused on self-driving cars, were wildly successful in acceleration the development of autonomous vehicles. Not long ago, the idea of self-driving cars that could share the road with humans seemed like fantasy. Now, thanks in part to those two Darpa-funded challenges, the age of “robotic Prius-driving overlords” are nearly upon us.

In 2004, the agency’s first $1 million Grand Challenge asked researchers to develop autonomous vehicles capable of traversing circuitous off-road mountain routes; no vehicle made it further than eight miles. A year later, with the prize money upped to $2 million, five different robo-cars completed the 132-mile course, with an SUV built by Stanford scientists taking the top prize. In 2007, Darpa’s next challenge doled out $2 million to a collaborative entry from Carnegie Mellon and Tartan Racing, when their autonomous car managed not only drive itself, but obey traffic signs and avoid other driving ‘bots.

With such successes, it’s no wonder that Darpa’s now opted to turn its attention to humanoid robots. And the agency’s latest challenge will also give smaller, even amateur, robotics shops a remarkable opportunity to triumph over institutional favorites like Stanford, Carnegie Mellon, and Boston Dynamics, builder of the planet’s best-known two- and four-legged ‘bots. While Darpa will initially offer funding to 12 different competitors, it’ll also permit unpaid teams to enter the fray. And Massey notes that “if some of the unpaid teams perform better than the paid teams … they will begin receiving funding.” It’s worth nothing that one of the finalists in the 2005 challenge was a bunch of dudes from a Louisiana insurance company.

Given Darpa’s recent track record with robotics, there’s no question that any successful competitor will need to really knock the agency’s socks off. This year alone, we’ve seen Darpa back the development of the fastest-ever robot, a robo-beast that can take auditory commands and lug hundreds of pounds, all while trekking through rugged terrain, and even a robo-arm that can take phone calls and open doors.

So what kinds of tasks will roboteers have to tackle in this latest Grand Challenge? According to Massey, machine entrants might have to “get into the driver’s seat [of a vehicle] and drive it to a specified location.” The robot may then hop out of the car and walk toward a locked door, where it’d “unlock it, open the door, and go inside.”

That’s not all. The robot may be asked to “traverse a 100 meter, rubble strewn hallway,” Massey writes, before the ‘bot “locates a pipe that is leaking” and then fixes it.

Though Darpa hasn’t released full details on the challenge, there’s no question that this’ll be a tall order — even for the most advanced robotics experts. Then again, the agency’s earlier robot challenges seemed outrageously difficult, back in the day. Now, we’ve got states passing laws to allow autonomous cars on the road.