Slashdot Mirror
Robot soccer - AIBO Blown Away
Chilli writes "The results of RoboCup 2000 (the 4th RoboCup world championships) are out. There was a big suprise in the four legged robot league, which consistently uses AIBO hardware. UNSW United blew the competition away. In the final, they won 10-0 against last year's champion. The success was largely due to a new walking technique developed by the UNSW team."
Yes, these football robots are all good and fine, but what I want to see are some ice hockey robots able to duke it out like the goons they should be. Besides, a robot able to skate would be quite a challenge to engineer, methinks...
I'm from Canada and just HAPPENED to be backpacking in Melbourne, Australia when I heard Robocup 2000 was on. Being a dedicated geek I decided to show up on the last day and see it for myself. I must say I was a little disapponted. It was cool an all, but I was there on the LAST day, which should have all the best stuff and the snazziest teams, but there wasn't a whole helluva lot to see and do. A comment on the robosoccer players: It was blatantly obvious (and probably is to a lot of you already, but I'll say it anyway) that the larger/more complex the robotics involved the shittier the soccer you watched. The best were the pure AI players displayed on a big-assed screen for all to see. No robotics at all. The winning team was FC Portugal and they did a fine job of getting their players to always go after the ball/opposing players in two, which in my opinion cinched their victory. The worst were the humanoid biped walkers. It was sort of neat to watch (the one I actually saw walk), but the thing was painfully slow, and watching the little 2' robotic imp turn was excruciating. It just sort of rocked back and forth for half a minute until it had changed direction. In it's favour it did manage to kick a ball into a goal, but they aren't anywhere NEAR being able to play soccer. They were more of a techo demo. Aside from that they had the small wheeled robots, the larger wheeled robots (sorta Dalek looking), and the Aibos. I didn't realize you could program EVERYTHING on there, including their now famous kneeling crab-walk gait that really did help. Even without their new walk their strategy was better. I couldn't tell you what it was, but it was aetherally THERE and helping them out. It was hilarious to watch the other (standard gait) continuously get bowled over by the crab-gait team. At one point I think ALL of the opposing team's Aibos were on their backs flailing their legs in the air trying to right themselves! WEll that's it that's all. $20 entry fee. Ciao bella. -=Cam ===== cam_marsollier@hotmail.com
You aren't going to get anywhere with "high level strategy". That's why no one has anything to show for 50 years of AI research.
Who cares if you find the optimal strategy for playing soccer within certain constraints? Programming a computer to play a perfect game of tic-tac-toe doesn't get you any closer to building something that's actually intelligent.
Robots aren't near human level computing power, so anything that looks like "strategy" is "pre-computed formula". On the other hand, Rodney Brooks' robots are in the same league as bugs, and they do bug like things in ways that may be close to how bugs do them.
So let the state of the art advance based on hardware for a while. Eventually, though, everyone will be running very similar hardware, and the better software will win out. Everyone is probably already running a good overall strategy -- it's the local control abilities that no one knows how to do well.
Are they robotic too?
"Free your mind and your ass will follow"
It would be cool if you gave them guns like that other security robot a while back. Then they could act like Rap stars.
Sig it.
or better yet can they pick up pop stars??
mv
How do you pronounce that in binary?
Or did you perhaps "access and browse" their site without checking their terms of use?
Oh no! Now Sony will firewall *slashdot* in my computer!
All opinions are my own - until criticized
The current problem with robocup is that you can win quite easily by devoloping some new "gimmick" like a new walking technique. Iran for instance in the middle sized league has a robot that can turn it's wheels so that it drives in a cirkel around the ball until it's pointing in the right direction and drives of with the ball( turning its wheels into the traight position again ). :) ) with almost no budget to speak of. They beat the combined effort of something like 5 italian universities because the italians where using mostly standerd robots. So until a "standard" robot has emerged that has all the good gimmicks we will not sea a real competion between the (robot)intelligience of the differant teams.
Most universities however are more interested in letting real robot's do higher level strategy and therefor use standard robots. You can write much more interesting reasearch papers about strategy and communications between autonomous robots, than you can about a set of turning wheels. Therefor the iranian team was last years world champion and this years european champion (it was an open championship