Home

Engineering Outlook

"Our robot is a tool to illustrate theoretical problems in intelligent control and explore them experimentally."
Mark Spong

Air Hockey Can Be Deep
by Jamie Hutchinson

Learn more about the air hockey project.

IBM’s Deep Blue beat Gary Kasparov at chess. How long before we see a robot air hockey champion? Not long, if the Air Hockey Project in the Coordinated Science Laboratory at the University of Illinois at Urbana-Champaign continues to progress. The project is directed by Mark Spong, who is an electrical and computer engineering departmental affiliate and a professor of general engineering. Funded since 1993 by the National Science Foundation, "Gretzky" represents more than an attempt to make the perfect air hockey player. "Our robot is a tool to illustrate theoretical problems in intelligent control and explore them experimentally," said Spong.

Faculty members and students in the Department of Electrical and Computer Engineering have played important roles in the project. Professor Seth Hutchinson proposed artificial intelligence techniques that add to the functionality of the machine. In his thesis, undergraduate student Brad Bishop derived mathematical methods from control theory that make the robot work. Graduate student Chad Partridge began work on the project in 1997. His improvements to the robot included stabilization, puck windowing, neural networks, and a linear slap.

The term "hand-eye coordination" aptly summarizes Gretzky’s system architecture, which consists of a three-jointed motor-driven arm, a camera that samples the puck path, and a variety of software and hardware (including a SunSparc workstation) that keeps everything working together. The result is no mean feat: a machine that achieves a goal (pardon the pun) by sensing its environment, then strategizing, and finally executing the strategy. In a speech in Champaign in January 1999, Vice President of the United States Al Gore cited the project as an example of important research done at the University of Illinois. And a video of the machine in action, which is available on the Web, recently won an award at an international robotics conference.

Watching Partridge play Gretzky is like watching a coach practice with a talented greenhorn—he is more interested in the pupil’s behavior than in scoring goals. With his mallet in one hand and "kill switch" in the other (in case the ill-disciplined student careens out of control), Partridge appraises Gretzky as they trade volleys across the center of the table.

"We had some instability recently because motors were supplied to us that didn’t perform as specified, but that’s been corrected," Partridge said. "It’s looking very stable."

Which means it is time for a test. "In the past, it’s been slow to react to these wall shots," Partridge says, aiming the puck to rebound into the goal, "but I revised the algorithm. See, it’s picking them up pretty well right now."

Such strides instill optimism in the teacher. When asked if a robot could ever beat a human air hockey champion, Partridge said, "It’s difficult to say because there’s a lot involved and unforeseen problems always come up. But yes, I think it will happen, possibly by the time I finish my PhD if I decide to stick around for that."

The Kasparov of air hockey is world champion Tim Weissman of Houston, who knows of the robot from exchanging videos with Spong. While Weissman is impressed by Gretzky, he is pessimistic about the prospect of a robot champion. "A robot could defeat a human champion," he said, "but it would be in the distant future in a galaxy far, far away."

Then again, Kasparov may have once felt the same way.

Chad Partridge is pursuing his PhD at Stanford University.

Jamie Hutchinson manages the Publications Office of the Department of Electrical and Computer Engineering.