- Professor Rod Grupen, director of the Laboratory for Perceptual Robotics, where the uBot was created.
Researchers at two campus
robotics labs have emerged at the cutting
edge of the field on the backs of three projects: UMan, uBot, and Dexter.
The UMan, or UMass Mobile Manipulator, couples a robotic arm with a
digital camera that allows it to learn about objects it encounters
in the real world. The fifth iteration of the uBot has inspired the
creation of startup DigitRobotics
LLC. And Dexter, which comprises
two arms and a two-camera head atop a stationary torso, can pick up
objects with one or both hands.
The UMan was developed by computer science doctoral candidate Dov
Katz and professor Oliver
Brock in the Robotics and Biology Lab. The robot
uses a digital camera that allows a robotic arm to “see” objects and
the environment. An algorithm turns that data into trial-and-error-based
learning. The arm pushes on objects and observes how they change, similar
to a child exploring the physical dimensions of the world. UMan stores
this knowledge for later use in performing specific tasks, such as
opening scissors. “Once robots learn to combine movement, perception,
and the manipulation of objects, they will be able to perform meaningful
work in environments that are unstructured and constantly changing,”
says Katz.
Not far away, in the Laboratory for Perceptual Robotics directed by professor Rod Grupen, Computer Science, the uBot-5 recently put all that information together to autonomously pick up a cylinder. The uBot, originally a government-funded project for small robots meant to be deployed in swarms to search emergency sites, has evolved into a balancing, two-wheeled mobile manipulator with a “face” made up of a Skype hook-up and a flat-screen head. Across the room sits Dexter, which uses its arms and camera eyes to learn about interacting with the physical world with single-hand and full-body motion.
PhD students Patrick
Deegan and Bryan
Thibodeau have spun their lab
experience and uBot technology into a business. They sell robotics
platforms to researchers at colleges. Digit already has customers;
three units are in production for MIT’s
Media Lab, which uses the uBot
as the chassis for Nexi, its emotion-expressing, dexterous robot. Digit
is also working under a NASA
Small Business Technology Transfer grant
to develop software to control whole-body mobile manipulator robots,
such as the uBot-5. (YouTube
video)
That’s a healthy start for a new business, when you consider the most
commercially successful robot is the Roomba, which sells for $130,
cleans floors while you sleep, and was developed by MIT spinout iRobot.
(Grupen uses a Roomba to keep his lab tidy.)
As the technology becomes more cost effective, the market should expand.
Experts predict that in a decade, the sales of personal robots could
surpass sales of industrial robots, now about $3 billion and $4.6 billion
a year, respectively.
Investors and researchers are eyeing the home health care market, especially
as baby-boomers get on in age. “Health care seems to be the place where
the evolving price point and the value of service are crossing first,”
says Grupen. Digit’s uBot has already demonstrated in the lab the ability
to check on a fallen person, call emergency services, and conduct a
remote check on vital signs.
Grupen says he’s motivated to discover what robots have in common with
human beings and other animals. He says researchers are now beginning
to appreciate that the description of natural systems is incomplete
without an understanding of the system’s computational nature—how it
perceives information, models the world, and makes decisions. “Robots
bring the natural sciences together with computer science and the social
and behavioral sciences to learn how humans really work.”
Grupen says playfully that while robots may indeed take over the world, only some of them will be evil: “It’s unlikely that they can all be recruited to the same common purpose—just like human beings.”
