EARL, the Enhanced Automated Robotic Launcher
It's a fun contest but I think before you can really say machines > humans, the machines will have to operate at human scale and not as these huge powered robotics system.
It surprised me that it lost because of oil removal on the track. Apparently humans adjust for that or just don't repeat it that accurately.
I would say it should still be relatively easily to let it make small changes that all throw strikes.
Either model the track and oil removal, or just let it run for a week or so with some optimization algorithm. I believe it shouldn't be that hard to let this machine beat a human bowler.
At least this showed me that bowling isn't only about just repeating the same motion, or this robot just sucks ;).
> It's a fun contest but I think before you can really say machines > humans, the machines will have to operate at human scale and not as these huge powered robotics system.
Human scale in size and power? Or a true android? It will take some time before an android beasts a human in bowling. That would be truly impressive.
Wondering how hard it would be to make a snowboarding robot? I should ask my brother in law, his is doing a phd in robotics. He works on child sized android robots, trying to let them walk (with new methods). He says one of the main problems of real world applications of robots is in the durability of mechanical parts. The child-human sized robots just fail too often for practical use outside the million dollar industry. So it's not only modelling and software development, but it also needs huge mechanical improvements.
Yes, android(like) robots are taking waaaaay too long in my opinion. I thought the Japanese would have it figured out by now. I also think, historically no one thought it would be this hard.
It surprised me that it lost because of oil removal on the track. Apparently humans adjust for that or just don't repeat it that accurately.
ReplyDeleteI would say it should still be relatively easily to let it make small changes that all throw strikes.
Either model the track and oil removal, or just let it run for a week or so with some optimization algorithm. I believe it shouldn't be that hard to let this machine beat a human bowler.
At least this showed me that bowling isn't only about just repeating the same motion, or this robot just sucks ;).
> It's a fun contest but I think before you can really say machines > humans, the machines will have to operate at human scale and not as these huge powered robotics system.
ReplyDeleteHuman scale in size and power? Or a true android? It will take some time before an android beasts a human in bowling. That would be truly impressive.
Wondering how hard it would be to make a snowboarding robot? I should ask my brother in law, his is doing a phd in robotics. He works on child sized android robots, trying to let them walk (with new methods). He says one of the main problems of real world applications of robots is in the durability of mechanical parts. The child-human sized robots just fail too often for practical use outside the million dollar industry. So it's not only modelling and software development, but it also needs huge mechanical improvements.
Yes, android(like) robots are taking waaaaay too long in my opinion. I thought the Japanese would have it figured out by now.
ReplyDeleteI also think, historically no one thought it would be this hard.