Slashdot Mirror
Kiro, the Foosball Robot
JasonFleischer writes "Although the official line from the RoboCup competition is that robots should be ready to challenge humans on the soccer
field by the year 2050, we don't have to wait that long to see man-machine
competition in the bar.
Researchers in Germany have developed a table football (foosball, table soccer, whatever) robot. The human challenger(s) take the red team, while the machine works the blue side, using an overhead video camera to see what's happening on the table. The conference paper shows that while the machine generally wins against the normal bar-amatuer it has no chance against a human grandmaster. But these kinds of things are always improving, after all look how big a deal the man-machine chess competition turned out to be. So perhaps the current table football world champions should be watching their backs."
but does the computer have full control over all the umm (forgive my foosball knowledge) "handles" at all time. I mean part of the game is having to let go and grab the next one. Kind of an unfair advantage if the computer does not have to do so.
I'll be impressed when they have a robot/AI that can play ping-pong. If you look at the plane the ball travels in, foosball is pretty two-dimensional...not entirely, I'll grant you, but I'm making a generalization. If you can create a robot that can deal with three dimensions, and can build strategies to play a good game of ping-pong, then I'll be impressed.
blog |
Chess is a thinking game. The idea of a machine being able to out think or out strategize a human fosters competition, so that's where the excitement in that lies. Foosball is simply hand eye coordination. I have no doubts that a machine will eventually be able to out manuever me in hitting a ball with a little plastic man on a rod. It his hardly the technical coup of a computer beating one of the best chess players in the world.
Rather than using a camera to read the table, if the table were built with sensors in each of the guys (to indicate state/current position/speed of current rotation) and a sensor in the ball (or positional sensors around the edges that could read the location of the ball), I believe that the software would win now, without any improvements. (Maybe we could put a GPS device in the ball!)
My guess is that the majority of the work that the computer has to do now is to figure out, from the video feed, where things stand before reacting. If the state were easier to read, the machine's reactions would easily outdo any human, champion or otherwise.
Why do I h8 apple?
I think it would be pretty much impossible to make a tournament-level competitive player this way.
50FPS is way too slow to react to even a simple three-man push shot. Humans generally can't see the ball when it's done well; they have to build up the skill to Just Know where the ball's going to go. It would be very easy to put some tiny variation into the shot that would completely throw off an AI.
Plus, the AI only models the field and the playing figures, so it can't respond to poker-style cues a human opponent might give off, like a slight pretensioning of the forearm, that would indicate when the shot would be attempted.
It would be amazing for practice drills though, both offensive and defensive.
Hey once one of the amusement table manufacturers sees this, I think they'd jump on the chance to develop a commercial version. Now that dotcoms aren't around anymore to put tables in their employee lounges, they probably need more revenue sources. It be coming sooner than you think. Give Kiro some smack talking ability and it'd be a hit. "Puny human, my Z-80 grandmother is faster than you!"
I've played this thing last February at the Hannover industry fair. I'm by no means a tournament-level player, but I'm not crap either, so I think I can judge this thing's playing power. It's not very good at planning shots (in fact, it's crap at that ;-)), but it's amazingly fast. Better reactions than I have ever seen in a human. And keep in mind, this was a year ago. They use motion prediction to increase the frame rate, so that isn't the limit.
I do think that they can make a tournament-level player out of this thing. Which isn't the motivation, of course, but give them two or three years.
Part of the reason IBM was able to create a program that beat a human was the fact that human chess masters prepare for a particular opponent (they study styles, weaknesses, etc), IBM never made any of this info on their program available, but did prep their machine to play a particular master.
If you ever get a chance to see a serious foosball tournament with high level players, go and prepare to be amazed. The difference between ordinary recreational foosball and hard-core tournament winners is like the difference between shooting a piece of wadded up paper into the garbage can from across the room and the NBA.
Total 13,233
Average Per Day 4
Average Visit Length 1:44
Last Hour 1,311
Today 8,038
This Week 31
coincidence anyone? i think not
Just by watching some of the video clips, it occurred to me that, while it's true the computer would be at a disadvantage on defense, it's strength is in it's ability to pick up the ball as it rebounds off a player or wall and quickly convert it into a shot. I imagine even a pro would have a difficult time with this. Imagine as the technology progresses, the computer's offense and defense would merge somewhat as it would be able to pick up the opponent's shots and passes and stuff them right back into your net before you have a chance to realize you even lost possession, sort of like slamming a fastball right out of the park.