dynamically self-balanced biped robot difficulties?

[roboSonny]

why is so hard to build a robot that is dynamically self-balanced and can mimic human walk?what are the problems?

[Metal Slug 2]

First of all they are expensive.  The quality of the parts needed to create a biped which even remotely mimics human walking have to be very precise, and thusly very expensive.

Next comes computational power. The mathematics involved in determining where a limb must be placed, how fast to move that limb, how much force is needed to support to weight of the biped (during walking/running/support on one leg/moving side to side etc.), the bipeds position in accordance to its surroundings, balancing etc etc. require a lot of processing.  creating a system that has both a small form factor and the required computational power is quite difficult, just take a look at ASIMO's "backpack".  dont forget that the biped needs its own power source...a biped tethered to a wall outlet at all times would be of not much use.

[roboSonny]

that applies to all scales?

[Metal Slug 2]

dynamically self-balanced and can mimic human walk

in relation to the above sentence, yes.

hobbyist-size bipeds wouldn't have such requirements.

[azy]

lag and inaccuracies means that you wont ever get the system to balance in any stable way. (unless you have an uber setup)

the heavier it is the less it will blow about in the breeze as it where, (upset by tiny changes)

you will need some smart fudgy code, need to think in terms of buffering, overshoot, undershoot , damping, ahead of time instead of just in time

big pushes followed by little pushes ? , charge forward and the reverse brake ?

complicated stuff is balancing

--------------------------

theres 2 kinds of walking, static always balanced short shuffle step kind of thing & dynamic

Humans don't walk by pushing there legs backward. ( you could of fooled me to my grave if i hadn't watched a video about it)

A you lean forward (now your centre of gravity is no longer passing through your feet)
B you start to fall
C only then can you put a leg out in front of you
D you partly! catch the fall with your leg, but are still leaning forward (allows for continuous smooth motion because you are always falling)
E some of the falling momentum is deflected into forward motion
F forward momentum plus the continuous leaning over (which is always there until u decide to stop walking) keeps you going

something like that anyway.

Still confused, remember how hard it was to walk forward when you where neck deep in a swimming pool , why ? it wasn't the slippery pool floor, it was the fact that you couldn't LEAN forward to make the first step. The buoyancy kept you upright. If you had a dense 10kg weight in each hand you could lean forward and walk.

-----------------------------------------------------------------------------------------------------

Its not actually that hard, you are an expert at all kinds of dynamic movement, but you do need "to become the robot leg"...get inside its mind  ,become acutely aware of what its doing ,its strengths and limitations that and a lot of practice.

The key to smooth motion is understanding the centre of gravity both of the whole and the moving parts. Remember hitting a cricket ball and it felt just right , that's the sweet spot that you are after. (when you get it wrong you know about it, and it also wastes precious energy

Big Dog, prolly the best example of biomechanics out there. 1:20 dancing on ice wins the cigar for me , like to see Asimo pull that out of the hat ! the jump at the very end is superb
Big Dog

still along way from perfect but clearly demonstrates the kind of dynamic movement we are after

-----------------------------------------------------------------------------

Better than Human

Chuck enough computational power and engineering at it and you can quickly surpass anything any human could ever do. Smaller robots are much harder to balance, like flying a tiny RC plane. need much faster reaction times (all objects fall at the same speed) and environmental disturbances are greater.

Murata Boy - Sexy Robots Videos

A 6ft 6 android could run at 100mph in theory, for a very short while. (preheated lithium batteries and precooled brushless motors as they do with RC racing and electric record breaking. Just give me 3 million bucks and 18months and ill build you one....any takers ? ..... High performace electric now almost equals internal combustion power to weight ratio, but run time is 100 times less or worse.

Advanced Robotic research is hugely underfunded, trillions have been spent elsewhere rolling out the global internet,mobile phone netowork, satellite and cable TV, Ipods and other gadgets with short Innovation to marketplace cycles. The robots you see today are decades behind what we could build if we really tried. Many top hobbyists blow away ye old men in white coats. I can buy an off the shelf toy robot that puts most PHD students efforts to shame.

The only thing holding back casual robot hobbyists is the problems interfacing hardware with software. A 5 buck web cam and a 5 buck servos and cheap as chip sensors isn't what holds us back. Yesterdays PC can be fished out of your local skip so we have uber computational power.  Gluing the Hardware to the Software takes us a month of sundays and is our bigest barrier and killjoy. The day that rickety bridge is replaced with an ironcast one robotics will advance a decade overnight. And then we wont be able to move for waltzing bipeds !