Students from Carnegie Mellon University say they will have one huge advantage this month at the international soccer competition known as RoboCup 2010: robots that understand the laws of physics.
The goal of students participating in RoboCup is to one day develop fully autonomous humanoid robots capable of beating the world’s top soccer team. This year’s contests will take place June 19-25 in Singapore.
Stefan Zickler, a computer-science student at Carnegie Mellon, has contributed the work of his doctoral thesis to the university’s robot soccer team: a physics-based algorithm that allows the robots to anticipate the movement of the ball, and plan a reaction to it.
“In the past, they’ve written a lot of code to deal with what could potentially happen,” says Mr. Zickler. “When something unforeseen happened, the robot didn’t know what to do.”
According to Mr. Zickler, the algorithm will help the robots out-dribble their opponents on the field by predicting what may happen in certain situations.
“When the robot has to manipulate the ball, the algorithm will creatively, on the spot, come up with new solutions,” he says.
Byron Spice, a spokesman for the Carnegie Mellon School of Computer Science, says the university has competed in RoboCup since the competition began, in 1997.
CMDragons, the Carnegie Mellon team, has won twice before, in 2006 and 2007. The team came in second in 2008 and placed in the quarterfinals in 2009.
In previous years, Carnegie Mellon’s robots have been programmed to play reactively based on the coding provided. However, this year, the robot players will be able to plan ahead to a certain extent based on knowledge of the ball’s behavior.
“Essentially, we teach the robots the laws of physics,” says Mr. Zickler. “We’re giving the robots more choice. They’re more adaptive.”
Mr. Zickler also tested the algorithm in robot mini-golf, robot billiards, and animations. He says the algorithm is general enough that it could have more practical applications as well, such as operating a dishwasher.
He says he hopes to improve the algorithm such that the robots could develop long-term strategies — like negotiating lucrative contracts with sponsors? Maybe in version 2.0. —Kelly Truong
(Image of Carnegie Mellon team: Stefan Zickler)