In Robot Soccer, US Team RoMeLa Dominates Robocup 2011 60
Narmacil writes "U.S. Team RoMeLa has swept Robocup 2011, winning first place in both the kid size and teen size divisions. (Video) CHARLI, America's first full size autonomous humanoid, managed to also make the high score record for the teen size division. DARWIN, the kid-sized robot, beat the Darmstadt Dribblers, the previous world champs, and continued on to the finals to win out." There must be joy in Blacksburg today.
Re:Is it really that hard? (Score:5, Insightful)
Our sensors _are_ pretty high precision, and very well integrated with our brain. They're also integrated well with each other. The three _big_ ones that are used in football would be sight, equilibrioception (balance), and proprioception (limb awareness), IMO (with touch and hearing playing lesser parts). Those 3 alone are very hard to replicate well in robotics, and combine into a complete package.
Re:Is it really that hard? (Score:5, Insightful)
Did anyone else watch the video [youtube.com]? Is it really that hard to make robots that can walk decently?
It seems like from a mechanical standpoint, it wouldn't be too hard to make something that mimics the muscular structure of the body......it's not like human muscles are particularly strong, or human nerves fast transmitting, or human sensors high precision. I'm having trouble understanding what the challenge is in building a robot that works reasonably well (or at least not dreadfully slow).
The devil is in the details. There's lots of muscles involved. Newbie mistake #1 is thinking you can build a walking mechanism using a really simplistic count of leg joints. Like trying to get away with about 4 servos per leg. You really do need a nearly fully articulated body including arms and spine, even if its cost is incredible compared to just two articulated legs. While walking to refill my water glass I felt my sore elbow moving as I dynamically balanced walking down the hall, and theoretically I wasn't using my elbow. Newbie mistake #2 is trying to use a single canned routine. Walk around some time and try to straighten or slump your back, it has an effect on gait. Even if you're not paying attention, your gait is pretty dynamic, and newbs can't program dynamic gait very well. Newbie mistake #3 is applying human male standards of beauty to something inhuman. To me, a reasonably well shaped 20s female human just coincidentally happens to have a "decent walk" that I would admire as she walks by. Other people, live and let live, their idea of living is Really liking the 4-legged gait of a sheep, or the 2 legged gait of a 'roo, etc. I'm sure to a robot, a similar robot would seem smooth and graceful, and if we don't like it, the robot's opinion would probably be to shrug shoulders and grunt "eh". Newbie mistake #4 is not emulating or handling the shock adsorption of soft tissue, both semi-statically and also dynamically. Not just female swaying as per above #3 or whatever, but the cushion of the spine and feet (and shoes!)