BERKELEY, Calif. — In an engineering laboratory here, a robot has learned to screw the cap on a bottle, even figuring out the need to apply a subtle backward twist to find the thread before turning it the right way.
This and other activities — including putting a clothes hanger on a rod, inserting a block into a tight space and placing a hammer at the right angle to remove a nail from a block of wood — may seem like pedestrian actions.
But they represent significant advances in robotic learning, by a group of researchers at the University of California, Berkeley, who have trained a two-armed machine to match human dexterity and speed in performing these tasks.
The significance of the work is in the use of a so-called machine-learning approach that links several powerful software techniques that make it possible for the robot to learn new tasks rapidly with a relatively small amount of training.
The new approach includes a powerful artificial intelligence technique known as “deep learning,” which has previously been used to achieve major advances in both computer vision and speech recognition. Now the researchers have found that it can also be used to improve the actions of robots working in the physical world on tasks that require both machine vision and touch.
The group, led by the roboticist Pieter Abbeel and the computer vision specialist Trevor Darrell, with the graduate researchers Sergey Levine and Chelsea Finn, said they were surprised by how well the approach worked compared to previous efforts.
By combining several types of pattern recognition software algorithms known as neural networks, the researchers have been able to train a robot to perfect an action such as correctly inserting a Lego block into another block, with a relatively small number of attempts.
“I would argue this is what has given artificial intelligence the whole new momentum it has right now,” Dr. Abbeel said. “All of a sudden there are all of these results that are better than expected.”
Roboticists said that the value of the Berkeley technology would be in quickly training robots for new tasks and ultimately to develop machines that learn independently.
“It used to take hours on up to months of careful programming to give a robot the hand-eye coordination necessary to do a task,” said Gary Bradski, a roboticist and computer vision specialist who founded OpenCV, a freely available software library for machine vision. “This new work enables robots to just learn the task by doing it.”
Previously, the Berkeley lab had received international attention for training a robot to fold laundry. Although it was viewed almost one million times on YouTube, the laundry-folding demonstration noted that the video had been sped up more than 50 times. The new videos show the robots performing tasks at human speeds.
Despite their progress, the researchers acknowledge that they are still far away — perhaps more than a decade — from their goal of building a truly autonomous robot, such as a home worker or elder care machine that could perform complex tasks without human supervision.
The researchers said that while their new approach represents an important leap, it is also fragile. For example, the bottle cap-threading technique will work reliably when the bottle is moved from one location to another or if the bottle is of a different color. But if the bottle is tilted at an angle before it is picked up, the robot will completely fail.
“There is nothing better to ask a roboticist, ‘If you change the conditions, will it still work?’” Dr. Abbeel said.
To explain the new approach, the researchers draw the analogy of how baseball players track and then catch balls. Humans do not do mathematical calculations to discern the trajectory of the ball. Rather, they fix the ball in their field of vision and adjust their running speed until they arrive at the spot where the ball lands.
This, in effect, short-circuits a complicated set of relations between perception and motion control, substituting a simple technique that works in a wide variety of situations without having to worry about details like wind resistance or the ball’s velocity.
Until now, robots have generally learned with a variety of techniques that are laboriously programmed for each specific case. The Berkeley researchers, who will present their results in a paper at an industry conference on robotics and automation next week in Seattle, instead connected the neural networks, which learn from both visual and sensory information, directly to the controller software that oversees the robot’s motions. As a result, they achieved a significant advance in speed and accuracy of learning.
“We are trying to come up with a general learning framework that allows the robot to learn new things on its own,” Dr. Abbeel said.
The advance underscores the rapid impact that the deep-learning approach has had on the field of artificial intelligence. Pioneered several decades ago by a small group of cognitive scientists, the techniques were blended in 2012 with the “big data” power offered by cloud computing systems. Researchers were then able to capture billions of images or samples of human language. Their software was able to show rapid progress in accuracy in recognizing objects and in understanding human speech.
Now computer scientists are pushing the techniques in new directions, including self-driving cars and a host of other applications. In December 2013, Deepmind, a British start-up, first demonstrated deep-learning techniques that could be used to play video games with more skill than most human players. The company, which Google acquired for an undisclosed sum in 2014, published a paper describing its advance in the journal Nature in February.
- http://bit.ly/1FqnYc8
Niciun comentariu:
Trimiteți un comentariu