GRASP Lab's latest kick: robot soccer


Get ready for the first sports craze of the 21st century: robot soccer.

It's already sweeping robotics laboratories all over the world, and it's only a matter of time before it comes to an arena near you.

Okay, make that a matter of a good long time. But it is a legitimate phenomenon among robotics researchers for a good reason, explains Andy Hicks (G'95), postdoctoral researcher in Penn's General Robotics and Active Sensory Perception Laboratory.

Robots playing soccer, he said, represent the current hot subject in robotics: "cooperative robotics. What can you do with several robots that you can't do with one?"

Cooperative behavior is what enables humans to perform many complex tasks, but while people take it for granted, it presents a number of challenging problems for engineers and mathematicians involved in designing robots, which have thus far only been able to operate independently.

"There are tons of isolated projects that cooperative robotics could handle," Hicks said, such as two robots carrying a box together, a project that is the subject of a doctoral dissertation currently being researched in the GRASP lab, but the soccer exercise provides ways to examine the more complicated engineering, mathematical and computational problems involved in getting robots to work as a team towards a common goal. "Researchers settled on robot soccer as a common problem to investigate because it's complicated, but not too complicated," he said.

Robot soccer as a research problem is still relatively new -- "It's only recently that we've had the computing power to do these sorts of projects," Hicks said -- but it's spread fast: Hicks estimates that one-third to one-half of the world's robotics research labs have projects under way. GRASP Lab Director Ruzena Bajcsy and Assistant Professor of Mechanical Engineering and Applied Mechanics James P. Ostrowski are directing Penn's research team, which includes undergraduate and graduate students as well as postdoctoral researchers.

While most robot soccer teams thus far have been controlled by a central computer that issues commands based on visual information from a camera mounted over the playing field, the Penn researchers opted for robots that could "see" the field and objects by themselves.

pan.gif

Mirrored spheres mounted over camera-lens "eyes" allow Penn's robot soccer players to see everything that's around them.

The first set of robots, built from scratch by undergraduates using Lego bricks, were equipped with computer chips that sent data from the robots' camera "eyes" via radio to a central computer. The problem of peripheral vision was solved by pointing each robot's camera lens upward towards a mirrored half-sphere mounted on its top.

New custom-built robots from Newton Research Labs, the first of which arrived this month, will be able to direct themselves via on-board microprocessors. The new robots are faster and have a more sophisticated color-perception system.

The first two robots will be programmed this summer by undergraduates to handle a penalty-kick situation. If this test proves successful, a second pair will be ordered to complete the team.

But "league play" is not the primary goal for Penn's robot-soccer team. While a number of robot-soccer tournaments exist, and the robot-soccer "World Cup" in Japan has become a major event, especially for Asian researchers, the tournaments are mainly "something you set teams of undergraduates working on to produce well-engineered products," as Hicks put it.

"We're not looking at it as a technological issue, but a theoretical one," he said. Hicks, whose own academic background is in mathematics, explained that the main goal of Penn's research is to develop "a sound mathematical understanding" of the problem and to create models that could then be used to tackle other problems in cooperative robotics.

Originally published on April 30, 1998