One of the disadvantages of robots is the fact that they are not good at learning by doing. Although you can fit a lot of information into the brain of a robot, you cannot expect it to teach itself a new motor task such as unscrewing a bottle of water or stacking blocks.
But this is now possible thanks to the researchers at the University of Berkeley which have developed a new algorithm that enables robots to learn and perform tasks using the trial and error method. Using trial and error the robot is able to learn how to perform tasks just like humans. This invention could lead to the creation of home service robots which could take care of tedious household chores which people would rather not do such as folding laundry, screwing in light bulbs and plunging toilets.
The name of the robot is BRETT, which stands for Berkeley Robot for the Elimination of Tedious Tasks. It is based on a promising form of artificial intelligence known as deep structured learning which creates neutral nests. In neutral nests the layers of manufactured neurons can process overlapping sensory data. This allows BRETT to recognize patterns and classify the received information.
Traditional robots are pre-programmed in order to deal with a series of scenarios, but this works well only in controlled situations such as medical facilities or laboratories. In order to become more integrated into people’s daily lives robots need to learn to adapt to the unknown. The algorithm used in BRETT requires a smaller amount of pre-programming and can function outside controlled environments. The human-like neutral circuitry on which the robot is based comes into action when the robot perceives or interacts with anything which surrounds it. The algorithm was developed by a team led by Professor Pieter Abbeel from the electrical engineering and computer sciences department. The Professor said:
“The key is that when a robot is faced with something new, we won’t have to reprogram it. 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.”
To prove that this method works the researchers required the robot to complete various tasks such as screwing a cap on a bottle of water, assembling a toy plane and putting a clothes hanger on a rack. And this was possible even though the robot had not received any particular details about the surroundings.
This algorithm is already used in iPhone’s Siri, the voice recognition software. However to apply it to motor tasks was a real challenge.
Image Source: Slash Gear