Massachusetts Institute of Technology graduate student Jared Glover is developing a new way for robots to learn with a ping-pong-playing robotic arm that improves how it hits the ball by receiving constructive criticism and advice.
Glover provides the robot with feedback after every swing. Over time, the robot learns to swing its arm in a certain way depending on the position and speed of the ball, and based on the advice it has been given.
Glover says coaching the robot could help it improve its ping-pong game faster, and he notes the robot's ability to learn by taking advice could have far-reaching implications for robots that have to define objects. "The idea is a pretty revolutionary one," Glover says. "I'm sure it's something other people have thought of in passing, but as far as I know, there haven't been any other papers that have actively explored learning and advice giving for any difficult robotic tasks."
The robot uses a high-speed camera to track the ball, which it identifies by its motion and color.
From Boston Business Journal
View Full Article
Abstracts Copyright © 2014 Information Inc., Bethesda, Maryland, USA
No entries found