Anti-Social College Kids Build Autonomous Foosball Table

Click to viewFour engineering students from Georgia Tech have built an Autonomous Foosball Table, quite possibly the key to my never-ending quest to replace the humans in my life with robots. The funny thing is, this thing isn't just a gimmick; it may soon become a decent foosball opponent. The table system cost about $500 to build, and combines a webcam, an 800MHz Pentium PC and servo-controlled paddles to move, twist, and kick. Here's how it works:


The computer runs a Java Media Framework app that tracks the ball as well as the human opponent via webcam. The PC then issues commands to a PIC microcontroller across a serial line, based on the information it's receiving. From there, the microcontroller tells the servos to either move or kick the ball. In addition, the table can predict the trajectory of the ball, lifting its midfielders in time to set up a clear shot.

Despite how awesome it looks to the layman, the project wasn't perfect according to the team. For one, they'd like to improve the moving and kicking speeds, a task which would simply require larger (and more expensive) gears. Also, they would ideally have a webcam capturing video at 60fps instead of the current 30fps. These improvements, combined with a better AI engine, could pump the table up to beating a serious player, or at least most drunk and/or stoned college kids.

I would love to try this table out, and maybe I will. But until those improvements are made, I guess I'll have to settle for actual social interaction. [Autonomous Foosball Table via Hack a Day]



@Captain Angry: Fail? Is that what your school said to you to allow you to play fooseball everyday? Regardless of how successful you could potentially be in your life, if you're playing fooseball that often, you may need to re-evaluate your life, or at least find something other that fooseball to play to break up the monotony.

As for the robot fooseball player, I must say that it's not a bad idea if only the speed, AI, and reaction time of the movements can be improved upon to challenge at least the casual player, if not a professional player such as Captain Angry here. This would be cool if applied to other two player table games, both for practice and for entertainment. I could see this being fun for a billards table, and could potentially be easier and more successful to apply since speed and response time is not an issue and the computer would have time to analyze the shots. Of course, the AI would have to be programmed to not be so perfect that the player would never have a chance. Hurray for forward, if not, different thinking.