See "Toddler learning to walk" at http://groups.csail.mit.edu/locomotion/videos.html (2004). At first, the robot walks passively down an inclined treadmill. But it quickly learns how to efficiently add just a little bit of power from its motors at the right time each stride, such that it soon walks on level ground and adapts to different surraces.
We could go downhill with (almost) no power too, if only we could coast-run fast enough (and could handle the shock in your leg when your stride becomes so long you fall several feet with each step)
True, but doesn't this mean that biped robots wouldn't be required to expend energy when walking downhill? Depending on the application, it can greatly increase the time that a robot can be in use.
BigDog was able to go both uphill and downhill, as well as even be kicked from the side while maintaining balance. Good point regarding being passive, I figured that somehow the robot would figure out the slope and provide extra power when the terrain was uneven or if it would be unable to maintain a given speed passively.
No, it didn't. Sorry if that's how it came across - I was just supposing that this concept could be use to extend the life of a robot such as BigDog or others by controlling how passive it was in response to downhill slopes.
This is pretty amazing! It'll definitely require power and motors to actually "walk", i.e. propel itself downhill, uphill or on a straight surface, but much less than what other robotic walkers do. And I don't see the need to make it two legged, either - that three-legged combination is fine by me, as long as it can clean my house, take out the garbage or carry boxes at a factory :-).
http://www-personal.umich.edu/~artkuo/Passive_Walk/passive_w...