Want to see how some robots of the future may find their way around? Hold a finger in front of you and stare at it.
Now shake your head, still focusing on that finger.Objects behind your finger appear to move in the same direction as your head, while those in front of it move in the opposite direction.
Using that principle, researchers have demonstrated a fast method for depth perception in visually guided robots, which should help them navigate around objects more quickly, a scientist says.
A robot at the University of Rochester in New York state trains a pair of cameras on an object to gauge the distance to it, said Dana Ballard, a computer science professor at the university.
The distance measurement is updated 30 times a second, Ballard said, so "it potentially allows you to move at high speed." Current robots guided by visual feedback can move only slowly because of the time it takes them to judge the location of objects in their path, he said Tuesday.
Chuck Thorpe, research scientist at the Robotics Institute of Carnegie-Mellon University in Pittsburgh, said the Rochester robot is the first to combine several previously known concepts into such a high-speed system.
The cameras are mounted on a motor-driven platform that allows them to move up and down or side to side. The platform is mounted on the end of a large industrial robot arm, allowing the platform to move over a range about 12 feet in diameter at speeds up to about one yard per second.
Currently, a human directs the cameras to a reference point, but eventually that will occur automatically, Ballard said in a telephone interview.