Awesomo!

Dexter Walks. Paul Graham explains it's the first time for a walking robot to balance dynamically, like people do.

"target=new>He jumps!

Very impressive and very funny! Maybe I won't be laughing so hard in 30 years time :) Astonishing stuff really.

Quote:

Original post by Diodor Dexter Walks. Paul Graham explains it's the first time for a walking robot to balance dynamically, like people do.

Cool! I was studying biomechanics a while ago, and I was wondering when somebody would finaly achieve this. This ones looks like it has a bad case of Parkinson, but that should be solved fast enough from now on.

Quote:

Original post by oliii "target=new>He jumps! Very impressive and very funny! Maybe I won't be laughing so hard in 30 years time :) Astonishing stuff really.

Incredible how they can do that so soon after barely getting walking right. Can't wait to see it running.

One thing that would be cool, and possibly cheaper, would be to simulate a walking robot such as this one. I bet it could learn a whole lot faster too.

-One thing that would be cool, and possibly cheaper, would be to simulate a
-walking robot such as this one. I bet it could learn a whole lot faster too.

check out this:
artificial humanoid

Seeing that made me want to try to do something similar, how hard do you think that is?

Quote:

Original post by oliii One thing that would be cool, and possibly cheaper, would be to simulate a walking robot such as this one. I bet it could learn a whole lot faster too.

Such simulations exists in biomechanics and ergonomics. As long as you are taking a serious approach, and not do something dumb like trying to control everything with a neural net, its not too hard. The biggest difference is that, in a simulation, its easy to estimate your center of gravity and your motion. If you read that guy's webpage, the biggest problem they had was not the actual control, but rather that they didnt have gyros that were precise enough. They had to make their own, so that the robot could be aware of its own motion and situation.

Quote:

Original post by daniel_i_l -One thing that would be cool, and possibly cheaper, would be to simulate a -walking robot such as this one. I bet it could learn a whole lot faster too. check out this: artificial humanoid Seeing that made me want to try to do something similar, how hard do you think that is?

I was laughing my head off at that video! Whoa... That was funny! Whew!

Quote:

Original post by Steadtler Quote: Original post by oliii One thing that would be cool, and possibly cheaper, would be to simulate a walking robot such as this one. I bet it could learn a whole lot faster too. Such simulations exists in biomechanics and ergonomics. As long as you are taking a serious approach, and not do something dumb like trying to control everything with a neural net, its not too hard. The biggest difference is that, in a simulation, its easy to estimate your center of gravity and your motion. If you read that guy's webpage, the biggest problem they had was not the actual control, but rather that they didnt have gyros that were precise enough. They had to make their own, so that the robot could be aware of its own motion and situation.

strap a wiimote to every part of the robots body
EDIT:
now i want to build a wiimote driven robot, the accelerometer's could give feedback as to its motion, the buttons could be wired to touch sensors on the side and the force feedback wired to the motors, then it could be controlled remotely by a program on a bluetooth enabled laptop.
EDIT:
this would be cool for a battle-bot or something, not only would it greatly simplify construction but you could make the input semi automate complex commands like turn or dash-attack and use any off the shelf game controller to operate it.

[Edited by - Kaze on March 5, 2007 3:58:48 AM]

Watching that thing reminded me of Animatrix. DESTROY IT NOW OR IT WILL STOMP ON OUR RACE!

Oh yeah and that standing neural network thing had me cracking up, I want to see more.