New robot learns from its mistakes - East Idaho News
Science & Technology

New robot learns from its mistakes

  Published at  | Updated at

BERKELEY, Calif. — A new robot can learn tasks through trial and error, just like humans.

The development is a major step in the field of artificial intelligence, according to researchers at the University of California Berkeley. The PR2 robot, nicknamed BRETT for Berkeley Robot for the Elimination of Tedious Tasks, is powered by algorithms which let the robot teach itself tasks like assembling toy planes and putting clothes hangers on a rack.

“The key is that when a robot is faced with something new, we won’t have to reprogram it,” professor Pieter Abbeel said in a statement. “The exact same software, which encodes how the robot can learn, was used to allow the robot to learn all the different tasks we gave it.”

Currently, most robots are programmed for specific applications in controlled environments, according to Berkeley Vision and Learning Center director Trevor Darrell. He said putting robots into real-life settings where surroundings change, like home and offices, is challenging.

The robot uses technology inspired by the human brain’s neural circuitry called “deep learning” to recognize patterns and learn from its mistakes.

“BRETT takes in the scene, including the position of its own arms and hands, as viewed by the camera,” a statement from Berkeley reads. “The algorithm provides real-time feedback via the score based upon the robot’s movements. Movements that bring the robot closer to completing the task will score higher than those that do not. The score feeds back through the neural net, so the robot can learn which movements are better for the task at hand.”

During tests, BRETT successfully stacked Lego blocks and screwed caps on water bottles, among other tasks. Researchers said it took the robot about 10 minutes to master a task when given relevant coordinates, or about three hours if it wasn’t given the object locations.

“With more data, you can start learning more complex things,” Abbeel said. “We still have a long way to go before our robots can learn to clean a house or sort laundry, but our initial results indicate that these kinds of deep learning techniques can have a transformative effect in terms of enabling robots to learn complex tasks entirely from scratch.”

He estimated that significant advances in robot learning could come through this line of work in the next five to 10 years. BRETT’s latest developments will be presented at the International Conference for Information Technology Research in Seattle Thursday.

beacon?cid=285837&pid=216

SUBMIT A CORRECTION