Skip to Content

Google’s Robot Recruits Dominate DARPA’s Rescue Challenge

Two companies acquired by Google demonstrate remarkable feats of agility and dexterity (albeit slowly) at a competition held in Florida.
December 21, 2013

Crowds gathered at a NASCAR racetrack in Miami this weekend to witness a more sedate sport than usual, as some of the world’s most advanced legged robots inched their way through a range of emergency tasks, including clambering over rubble, clearing debris, and operating a fire hose. And two of the robot-makers acquired recently by Google, Boston Dynamics and Schaft, dominated the contest, giving some sense of why the company was so keen to snap them up. 

In all, sixteen teams took part in the challenge (see photo gallery: “Robots to the Rescue, Slowly”). The robots were operated remotely but still required sophisticated automation to cope with the complexities and uncertainties faced when dealing with the real world. Teams scored points by completely tasks inspired by a real emergency faced at Fukushima-Daiichi nuclear disaster in 2011: as hydrogen leaked from the stricken plant, human rescue workers risked their lives trying to reach and operate a valve that might’ve stemmed the leak. The robots faced eight tasks: walking over uneven ground; moving chunks of debris from a walkway; turning a valve; drilling a hole in a wall; climbing a ladder; maneuvering through several doors; manipulating a hose; and driving a golf cart along a snaking course.

Such jobs are, of course, simple for humans. But creating machines capable of stepping into role of a rescue worker is no easy feat. Walking across uneven, unfamiliar terrain and reliably manipulating everyday objects remain incredibly difficult engineering challenges (which explains why the robots worked at speeds that often felt exasperatingly slow). But if such skills can be mastered, they could also be useful for much more than just rescue missions.

Schaft, a spin-out of the University of Tokyo that acquired by Google in recent months, won the contest with 27 out of a possible 32 points. The teams that placed second and fourth both used a humanoid robot called Atlas, developed by another Google recruit, Boston Dynamics (see “Google’s Latest Robot Acquisition Is the Smartest Yet”).  The second and fourth placed teams, from IHMC and MIT respectively, had just a few months to learn how to program and operate these Atlas machines. A team from CMU placed third with robot called CHIMP.

Atlas, made by Boston Dynamics, strolls over some rubble during the DARPA Robotics Challenge.

While much has been made of the military funding Boston Dynamics has received in the past, to understand what Google has planned for its robot acquisitions it may be more instructive to look at Schaft. This machine is the culmination of many years of research in Japan, inspired in large part by concerns over the country’s rapidly aging population.

In fact, Gill Pratt, the DARPA program manager in charge of the contest, believes that home help is the big business opportunity humanoid robots. 

“Most people don’t realize that the military market is quite small compared to the commercial market. And the disaster marketplace is even smaller than that,” he said from the sidelines. “My feeling is that where these robots are really going to find their sweet spot is care for folks at the home—whether that’s for an aging population or other uses in the home.”

Pratt added that the challenges faced by the robots involved in the DARPA event are quite similar to those that would be faced in hospitals and nursing homes. “The rough terrain requirements of going up and down slopes will not be as great, but the robots will certainly have to go up and down stairs; people will leave clutter all over the floor. Because we arrange our houses to suit human beings, it’s very important that the robots have the same competencies of locomotion and manipulation as human beings do.”

The robot developed by Japanese company Schaft removes debris.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.