New generation of smart robots could change our lives
Tossing a ball through a hoop might not seem like a complicated task. But it is pretty impressive when you consider this robot learned how to do it all by itself.
"We provided only the goal, we did not provide the way to achieve or the means to achieve that goal," said Heni Ben Amor, an assistant professor of robotics at Arizona State University.
Ben Amor leads a team of researchers at Arizona State University where they're developing machines that can learn.
Instead of being programmed to carry out a task, the robots use machine learning-- a combination of algorithms and sensors that makes them teachable. The robots can adapt and react to a situation, like catching a ball by anticipating where it will be thrown. They don't always get it right.
"Ideally, the robot would not take a day or week for this but rather a couple of hours," Ben Amor said.
The technology is helping developers design smarter robots. Take, for instance, one robot that can learn the floor plan of a house or business and be taught to pick up misplaced items it finds. Researchers say a device like this could one day clean up your home.
Machine learning is also being adapted for medical use. One robotic prosthetic can analyze the nuances of how its user walks and adjust as needed.
"It's learning the relationships between the human and how it should react," researcher Jeff Clark said.
Researchers believe the teachable tech will ultimately help both man and machine learn to interact.