Scientists have created a new algorithm that allows robots to learn motor tasks through trial and error "like a human."
The researchers demonstrated their breakthrough method by having a robot perform tasks such as putting clothes hanger on a rack and even assembling a toy plane without being preprogrammed to do so, the University of California, Berkeley reported.
"What we're reporting on here is a new approach to empowering a robot to learn," said Professor Pieter Abbeel in UC Berkeley's Department of Electrical Engineering and Computer Sciences. "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."
Instead of requiring the tedious but conventional preprogramming needed to allow a robot to navigate a 3-D environment, this new method employs a concept referred to as "deep learning," which is inspired by the neural circuitry of the human brain.
"For all our versatility, humans are not born with a repertoire of behaviors that can be deployed like a Swiss army knife, and we do not need to be programmed," said postdoctoral researcher Sergey Levine. "Instead, we learn new skills over the course of our life from experience and from other humans. This learning process is so deeply rooted in our nervous system, that we cannot even communicate to another person precisely how the resulting skill should be executed. We can at best hope to offer pointers and guidance as they learn it on their own."
Deep learning programs create what are known as "neural nests," in which layers of manufactured neurons process overlapping sensory data. The process allows the robot to recognize patterns, and categorize the information it is receiving. This algorithm is currently used in voice recognition software, such as the iPhone's Siri, but applying it to motor tasks has been a much larger challenge.
"Moving about in an unstructured 3D environment is a whole different ballgame," said Ph.D. student Chelsea Finn. "There are no labeled directions, no examples of how to solve the problem in advance. There are no examples of the correct solution like one would have in speech and vision recognition programs."
To move forward in this field, the research team worked with Willow Garage Personal Robot 2 (PR2), nicknamed BRETT (Berkeley Robot for the Elimination of Tedious Tasks.). BRETT was presented with a number of motor tasks, such as stacking Lego blocks. The new algorithm provided a reward function that scored the robot's activity based on how well it was doing on the task. A camera on the robot allows it to view the scene, including its own hands, in 3-D. The algorithm provides real-time feedback and delivers the score based on robot's movement. As BRETT makes more and more correct movements, the score increases. The algorithm could allow robots to learn a new task on its own in as little as 10 minutes.
"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. In the next five to 10 years, we may see significant advances in robot learning capabilities through this line of work."
The findings will be presented at the International Conference on Robotics and Automation (ICRA).
WATCH:
