Challenging robot makers to build a ping-pong champion
San Francisco — Robot ping-pong, anyone? That is the invitation that John Billingsley has been spreading around the world recently.
A professor of electrical engineering at Portsmouth Polytechnic in England, Dr. Billingsley has gained some notoriety in the past by sponsoring MicroMouse competitions - contests where robot mice ''race'' to the center of a labyrinth. Now, however, ''the robots have simply gotten too good,'' he explains.
The first robot mice took 15 minutes or more to puzzle their way through the maze, but these days the winners take 30 seconds or so, the dapper electrical engineer reports. So he decided robotmakers needed a bigger challenge, and this new form of ping-pong was born.
What the good professor has in mind is something much more modest than arming ''Star Wars''-like robots with table-tennis paddles and turning them loose to serve and volley on an official-size ping-pong table. Such a feat is well beyond the state of the art. Instead, the game will be tailored to the more limited abilities of current machines.
The one-armed mechanical gamesters will compete on a table 6 feet (2 meters) long by 20 inches (50 centimeters) wide. They will be armed with paddles that cannot exceed a circle 5 inches in diameter.
From a robotics point of view, one of the most difficult problems the game poses is that of vision: acquiring, tracking, and analyzing the optical image of the ping-pong ball. The mechanics of sticking a paddle in the ball's way is relatively straightforward.
To reduce these visual problems from prohibitive to just difficult, high lobs are outlawed: The ball must pass no more than 20 inches above the 10-inch net so the competitor does not loose sight of it. The tiny sphere will be put into play each rally from a specific spot at the center of the table so the contestants can start with their ''eyes'' firmly on the ball. And the net will be transparent so the robots never lose sight of their bouncing target.
The first such match will take place next year in London at the annual Computer Fair exhibition, and Billingsley has already received 30 serious inquiries.
''I imagine that, at first, the best defender will win,'' the engineer says. Simply hitting the ball consistently back to the center of the table will be a big enough challenge. Next, he foresees robots being refined so that they start aiming for the corners, trying to win points by positioning their shots. Finally , the mechanical contenders may begin trying to put spin on the ball. By then it may be time to move on to a more difficult game.
The serious purpose behind competitions such as this is to stimulate people to improve the capabilities of low-cost robots. For instance, it should encourage the development of inexpensive vision sensors and software for tracking moving objects.
(Those interested in taking up Dr. Billingsley's challenge can write to him at the Department of Electrical Engineering, Portsmouth Polytechnic, Anglesea Road, Portsmouth, England, PO1 3DJ.)
Who knows? This might be the first step toward the Robot Olympics. Perhaps even professional robot sports: After all, although robot athletes will undoubtedly be expensive to build, they could be programmed never to turn free agent.