With perception, robots could be programmed to retrieve a cane for a disabled person. Perception would allow them to navigate the shortest route and around obstacles such as a lamp or a chair without getting stuck. That’s the scientific intent behind Catch-bot, which works this way:
¢ A ground ball is tossed, and a color camera atop the robot’s head zeros in on the ball.
¢ The camera takes 30 pictures a second, sending the data to a computer for quick analysis. The robot is programmed to respond the same way a human would to a ball rolling on the ground, using mathematical rules that predict the ball’s path.
¢ Like a human, Catch-bot adjusts by scooting backward, forward, right or left and positions itself to catch the ball. Instead of a glove, the Catch-bot steadies the ball against a $1.50 foam bumper. The robot can catch a ball released 70 feet away. It makes the catch about 75 percent of the time.
The technology is still being perfected. The robot cannot catch fly balls, although that is a goal in the coming year.
Like humans, Catch-bot is susceptible to injury. The robot prefers a spongy Nerf-style ball, rather than a baseball, to protect the delicate camera equipment atop its head. Catch-bot was injured once while scooting under a chair to retrieve a ball. Surgery was necessary to repair the camera, and the robot was sidelined for a week.
In some ways, Catch-bot is better than humans. For instance, in one of the most infamous errors in baseball history, Bill Buckner of the Boston Red Sox missed a ground ball in the sixth game of the 1986 World Series. The error prompted a seventh game and a loss to the New York Mets.
“The robot would have caught that one,” Professor Mike McBeath said. For one thing, “it doesn’t let the ground ball go between its legs.”