MIT building a robot without a mind of its own

"Hermes" the latest smart humanoid built at MIT picks up a bottle replicating the hand movement of operator Joao Ramos.
"Hermes" the latest smart humanoid built at MIT picks up a bottle replicating the hand movement of operator Joao Ramos.

In the 2013 summer blockbuster “Pacific Rim,”  Earth is under attack from mega monsters, “Kaiju,” who are beaming their way onto the planet through a portal that opens up on the bed of the Pacific Ocean.

The invaders rise up, cities fall, humans rally: Their weapons against the monsters are titanic fighting robots the size of jumbo jets operated by two humans suspended inside the machines.

The two humans are connected to each other, and to the robot, by a kind of mind-body meld. As they run and punch at the air in their harnesses, so do the robots they’re attached to.

So far, so science fiction. But a robotics team at MIT is taking baby steps toward a similar synchronicity between man and machine.

Hermes the robot's movements are controlled by graduate student Joao Ramos.

It’s called Hermes, and standing 4 feet 8 inches on perpetually bent knees, this robot is more Hobbit than humanoid hunk.

The team and robot traveled to California this year at the invitation of the US Defense Advanced Research Projects Agency, or DARPA, to demonstrate their invention at an all-star agency-run robotics contest in June.

For the two preceding years, teams had strived to hit DARPA’s goal of creating robots that could independently walk, drive, and carry around objects. It was DARPA’s simulation of mock-rescue scenarios where robots would play heroes: Competing robot teams had to open doors, walk over an uneven surface, use a power drill — all with minimum input from their minders.

Ramos watches the world through Hermes' "eyes."
Ramos watches the world through Hermes' "eyes."

MIT’s Hermes, in contrast, presents an alternative path to independence and autonomy. The puppeteer has a better chance of responding reflexively to sudden hazards like falling debris.

In the basement of an MIT lab off Massachusetts Avenue on a recent afternoon, PhD students Joao Ramos and Albert Wang demonstrated what Hermes can do.

"Hermes" the latest smart humanoid built at MIT picks up a bottle replicating the hand movement of operator Joao Ramos.
"Hermes" the latest smart humanoid built at MIT picks up a bottle replicating the hand movement of operator Joao Ramos.

Ramos was rigged up to what looked like elaborate climbing gear, with slim plastic pylons attached to ankles and wrists, and wires sprouting from a chest plate. The apparatus sensed Ramos’s movements — if he moved his shoulder upwards, or raised his arm to wave, and moved the limb of the robot correspondingly.

Virtual reality goggles channeling video from a camera perched on Hermes’s head allow Ramos to see the world as the robot does. Using a joystick, he was able to get Hermes to pick up a water bottle, and with three robotic fingers, grip the shaft of a fake ax.

A nest of wires sense Ramos' movements, and also provide realtime feedback to the operator, allowing him to adjust the robot's balance.
A nest of wires sense Ramos' movements, and also provide realtime feedback to the operator, allowing him to adjust the robot's balance.

Though they were reaching for independence, most robots on the track at the DARPA Robotics Challenge had trouble staying upright, as this compilation from the contest footage shows.

“A lot of people have compared these robots to toddlers,” Hermes maker Wang said. “If you’ve been on this earth for more than four years, you’re able to coordinate so much better than that.”

Walking on two feet while keeping balance and correcting for natural obstacles involves some complicated calculations. So at least in the short term, having a person in a harness think for the robot instead of building a machine that does so could have better results.

What’s unique about Hermes is that Ramos is able to sense the robot’s balance through the harness. If the robot is about to fall forward, he feels a pressure on his hip. As he rights himself, so does the bot.

Albert Wang watches Hermes' movements on a screen.
Albert Wang watches Hermes' movements on a screen.

Like the fictional heroes in “Pacific Rim,” the idea is that future versions of Hermes could also save the day by accessing areas dangerous for humans following natural or man-made disasters.

Except for one difference, noted Wang: “Why would you put the person inside the robot when it’s going somewhere dangerous?”

Of course, Hermes and humanity are several steps away from embracing that future. The team has only successfully demonstrate control of the robot’s arms and torso.

Like any robot in development, Hermes is not glitch-free. During the demo, the team lost control of Hermes’s right arm, which began flailing without warning. Whoa, said Wang. “That’s not supposed to happen.”

Nidhi Subbaraman writes about science and research. Email her at nidhi.subbaraman@globe.com.
Follow Nidhi on Twitter - Facebook