If you can’t beat them, make a robot that can. A team of BYU students and faculty took this philosophy to heart over the last couple of months, as they built an artificial intelligence machine that can not only play foosball, but it can win against humans.
The robot began beating about 80 percent of the people it played in April.
“It wasn’t perfect then and we’ve been working on it since,” said D.J. Lee, the faculty advisor in charge of the machine’s development. Lee estimates the machine could outplay an even greater percentage of people now.
The idea for the A.I. machine came from a YouTube video, where another school made a similar table.
“We looked at their design and thought, ‘well, we could do better,’” Lee said. That included a more simplified design with only two motors, one to kick the ball and the other to move the players.
Only six students were part of the design team.
“It was cool to see that we as students were attacking a legitimate problem because our professor a lot of times had no better solution than we had,” said Nathan Warner, one of the students that worked on the project.
Everyone who labored on the project enjoyed it. According to Warner, the most fun part was ”seeing it work, putting in the algorithms, plugging it in and playing, seeing it go, seeing it not let you kick the ball right past it whenever you wanted to.”
However, the machine wasn’t all fun and games for the makers.
“It’s got a lot of frustrations as well. You build something and you think ‘oh this is going to work’ and then you put it on and it crashes and breaks” Lee said.
One of the major issues that had to be overcome was lighting. The camera and software responded differently depending on if the foosball table was in a darkened room, played during the morning or night, or even if it was near a window.
“We just had to adjust wherever we were. We had to figure out how to adjust best to our environment,” said Taylor Simmons, another student who worked on the project.
The team solved the problem by increasing the number of LED lights. This helped the robot to keep better track of the ball and enhance its accuracy.
However, there is still hope for the humans.
“A human is still better, in that a human can adapt,” Warner said. “We were able to beat the machine. The longer we worked on it, the harder it become to do that though. By the end of the semester it was a challenge that we really had to fight for it, to be able to beat it.”
The team hopes to continue to work on the machine. According to Lee, “There are still some weaknesses. It can be improved.”
For the time being, however, other projects may begin to take precedence. Lee mentioned making a robotic arm that could play Ping-Pong, while many on the student on team have graduated and moved away for graduate school and work.
Despite the uniqueness of the project, students said this project helped them to prepare for the future.
“It wasn’t necessarily that we were doing something groundbreaking. What we implemented was a lot of stuff that was available,” Simmons said.
The experience gained from the project should help the students better utilize current technology in solving today’s problems.
But for now, Simmons hopes he can just continue to work with “small little devices that have a lot of intelligence.”