Blue is the name of a low-cost robot, specially designed for being “friendly” to its users,created by researchers at the University of California, Berkeley.
Blue was designed to be able to use the latest advances in artificial intelligence and deep reinforcement learning to be specified in more subtle human works, while remaining economical and safe enough so that every A.I.
researcher -and in the long run every home- can get one.
Blue is the brainchild of Peter Abil, professor of electrical engineering and computer science at UC Berkeley, the postdoctoral Steven McKinley, and graduate student David Gili. The team hopes Blue will accelerate the development of domestic robots.
“A.I. has done a lot for the existing robots, but we wanted to design a robot that’s just for artificial intelligence”
“Existing robots are very expensive, they are not safe when they work around people and are not safe for themselves; if they learn through trial and error, they easily cause damage to themselves. We wanted to create a new robot that is suited to the time of the A.I. instead of the time of high-precision factory automation.”
For years, Abil has been working on deep reinforcement learning algorithms that help robots learn through trial by error, or through people’s guidance as “marionettes”. He developed these algorithms by using robots from other companies that cost tens of thousands of dollars.
Plastic parts and high performance Blue motors cost less than $5,000 (construction-assembly). In each arm, the size of a hand of a bodybuilder, however is sensitive to external forces -as e.g. a hand that pushes- and have rounded corners and few points where fingers can be grasped. However, they can be rigid -like human hands that exercise power in something- or very flexible, like the members of a resting man. Blue can lift straight up to 2 pounds of weight, with arms fully stretched, but is designed with “Thermal constraints”, as McKinley explains. This means that, just like a man, he can force more than two pounds in a quick “explosion” until he reaches his thermal boundaries and needs to “rest”; a person can easily carry a heavy weight in a room, but not a long distance if he does not stop.
The team is currently building 10 arms, to be distributed to “early adopters”. At the same time, research is ongoing on areas such as Blue‘s resilience and its mass production capabilities, intended for use by the Berkeley Open Arms.