DARPA Pushes Machine Learning with Legged LittleDog Robot

With phase two testing wrapped up, six teams of roboticists are focused on improving LittleDog's speed and agility














Share on Tumblr



ON YOUR MARK: For an autonomous system like LittleDog, all of the difficulties with perception, cognition and action are combined with the engineering challenges posed by the mechanical system. Image: Courtesy of Boston Dynamics/DARPA/Carnegie Mellon University

Editor's note: Legged robots have the ability to follow troops on long journeys across extremely difficult terrain. In our series on legged robotics, Scientific American Online explores the challenges such technology poses as well as two DARPA projects—BigDog and LittleDog—that have shown great promise.

If BigDog is the Defense Advanced Research Projects Agency's (DARPA) dopey but lovable Great Dane, LittleDog is its extremely intelligent—if high-strung—Jack Russell terrier.

Shortly after DARPA commissioned Boston Dynamics to build its BigDog autonomous legged robot, the agency decided it should broaden its research to include a likewise legged device that was aware of its environment and deliberately placed its feet to avoid falling. LittleDog's software spells out the robot's route and its cameras and sensors help it "see" obstacles so it can avoid missteps.

While BigDog's quick thinking and nimbleness has its limits—particularly if it happens to step off of a high ledge or cliff, LittleDog's specialty is being able to sense its surroundings and avoid such dangers all together. It methodically moves over obstacles much larger than its leg length and body size—it measures 11.8 by 7.1 inches (30 by 18 centimeters), stands 5.5 inches (14 centimeters) tall and weighs 4.9 pounds (2.2 kilograms). "We wanted LittleDog to deal with the locomotion problem," says Larry Jackel, a DARPA program manager responsible for robotic vehicles who spent four years at the agency until June 2007 and now works as an independent consultant.

DARPA is looking for its mini-legged robot to cross progressively difficult terrain at increased speeds. "BigDog and LittleDog are related in that they are both focused on solving the problems that will enable legged robots to accompany war fighters as they cross complex terrain," says Tom Wagner, program manager in DARPA's Information Processing Techniques Office. (For more on BigDog, read "Leggy 'BigDog' Robot Set to Step Up for the Military.")

Phase two of LittleDog's development recently wrapped up, and phase three is set to begin this summer. In the first phase, which began in late 2005, DARPA asked six teams of roboticists—from Carnegie Mellon University, the Florida University System's Institute for Human and Machine Cognition, the Massachusetts Institute of Technology, Stanford University, the University of Southern California and the University of Pennsylvania—to improve on the same basic quadruped robot platform, which DARPA paid Boston Dynamics more than $1.6 million to design, build and support. To successfully complete this phase, each team's LittleDog needed to move at the rate of at least a half an inch (1.3 centimeters) per second over terrain that included obstacles 1.9 inches (4.8 centimeters) in height.

To succeed in phase two, which ends today, the teams need to tune their LittleDogs to scurry 1.7 inches (4.3 centimeters) per second across obstacles 3.1 inches (7.9 centimeters) tall. Now the teams have their eyes on funding for the next phase, whose requirements are 2.8 inches (7.1 centimeters) per second across obstacles 4.3 inches (10.9 centimeters) tall. To do this, DARPA scientists created a specific terrain for LittleDog, which is equipped with sensors in each leg. "We knew where the robot was with respect to its environment," Jackel says.

One of the LittleDog competition's biggest challenges has been improving on the original software so that the robot can read any map and then navigate the map's terrain, says Carnegie Mellon Robotics Institute research scientist Drew Bagnell. The teams are asked to ship nothing more than a hard drive containing their software to DARPA, which then loads the program into their own version of LittleDog. "There's a blind component to the test," he adds. "We get tested on terrain that we've never seen nor will ever see."

DARPA's strategy is a sound one, says James Kuffner, an associate professor at the Robotics Institute. It is a good testing strategy because it forces the roboticists to write software that works for a variety of terrains, he adds. "We don't want to hard code something into the robot that works for only a few examples."

One thing DARPA will not do is commission a remote-controlled legged robot, which the agency believes would be impossible for one person to manage. Jackel likens it to driving a car that has four steering wheels. Instead, DARPA has called on LittleDog's research teams to develop algorithms that manage each leg. "People have been talking about legged vehicles since the 1960s but they didn't have good algorithms to make the work," he says. "If you know that every step is going to be a repeat of the previous step, it's not that hard. But when you get to unstructured environments, that's where things fall apart."


5 Comments

Add Comment
View
  1. 1. AET RaDAL 04:29 PM 4/16/08

    Cute! I wonder how long before they get a cat version like the one in Red Planet -http://www.videodetective.com/movies/RED_PLANET/trailer/M00169794.htm .

    For a somewhat whimsical view of what I'd be doing if I was running DARPA, check out my blog - http://science-community.sciam.com/blog-entry/Aet-Radals-Blog/Future-Weapons-Warfare-Scenarios/5800000569

    Reply | Report Abuse | Link to this
  2. 2. geoguy 04:30 PM 4/16/08

    eventually all these robots will be armed and kill us.
    That's no joke or exaggeration.
    We better be damned sure who has control of the automated army becuase without soldiers you don't need general compliance.

    vivzizi.com

    Reply | Report Abuse | Link to this
  3. 3. AET RaDAL 11:05 PM 4/16/08

    Hey Sci Am! You've got to fix this posting software. I didn't do all of those reposts of my comment. Something is really screwed-up. The entire post didn't even appear either...

    Reply | Report Abuse | Link to this
  4. 4. The_Stranger 02:27 AM 4/17/08

    The idea thyat a legged robat can't be remote controlled is a fallacy. All that needs to be done is for the operator to choose the direction and the robot to figiure out how to move in that direction

    Reply | Report Abuse | Link to this
  5. 5. SteveBallmer 01:22 PM 5/5/08

    That thing is ugly!

    Reply | Report Abuse | Link to this
Leave this field empty

Add a Comment

You must sign in or register as a ScientificAmerican.com member to submit a comment.
Click one of the buttons below to register using an existing Social Account.

More from Scientific American

See what we're tweeting about

Scientific American Editors

Tweets could not be retrieved at this time

Free Newsletters


Get the best from Scientific American in your inbox

Solve Innovation Challenges

Powered By: Innocentive

  SA Digital
  SA Digital

Science Jobs of the Week

Email this Article

DARPA Pushes Machine Learning with Legged LittleDog Robot

X
Scientific American MIND iPad

Tap into your MIND

Get Both Print & Tablet Editions for one low price!

Subscribe Now >>

X

Please Log In

Forgot: Password

X

Account Linking

Welcome, . Do you have an existing ScientificAmerican.com account?

Yes, please link my existing account with for quick, secure access.



Forgot Password?

No, I would like to create a new account with my profile information.

Create Account
X

Report Abuse

Are you sure?

X

Institutional Access

It has been identified that the institution you are trying to access this article from has institutional site license access to Scientific American on nature.com. To access this article in its entirety through site license access, click below.

Site license access
X

Error

X

Share this Article

X