My bipedal robot finally walking!

[vinmaster25]

cool dude i wanna do something like this but it took you pretty long it seems and im not smart instruction wise ._.

[Kyntheria]

That is neat.

[realloyd20]

CONGRATS MAN! CONGRATS!!

[grimtango5280]

The weight is entirely supported by the internal rod/ram. This means the robot hangs while taking a step. This also means you haven't done any center of mass/gravity calculation which is how a biped actually walk i.e. by shifting the center of gravity for balance when each foot and leg is moved.

I don't understand the need for the rod really? its clearly there to balance the bot. Why wouldnt you want to balance it in hardware and code? isnt that the whole point?

[dwidgit332]

@grimtango5280 The weight is not supported at all by the boom. It is only for lateral support. The purpose of this experiment was to demonstrate the algorithm for a planar robot, not a full 3d robot. The algorithm is called the foot placement "estimator" because some, but not all, of the dynamics are neglected from the calculation and this experiment validated that only the insignificant terms were dropped. A colleague has since proven the theory can also work in the full 3d case.

[ChudFapper]

@grimtango5280

How on earth would he be able to program it to balance, when the greatest scientific minds within the field can't do it with their massive budgets?

Every walking robot that's ever been made has basically been cheating, walking like a drunken retard or is pre-programmed to move in a specific pattern, rather than dynamically balancing itself.

[robertlam18]

after 4 years of research but still can't walk on its own ?

[williamthomasmi10]

Does it decide when/how to step on it's own, or does it have premeditated commands?

Very cool my friend!!

I once saw one on the television, and it had sort of kneecaps, so it could extend it's legs all the way and not have them hyperextend... Just an idea! Yours moves beautifully!

[dwidgit332]

I'm using a measure of balance that I called a "foot placement estimator. When this estimator moves in front of the standing leg then it means the robot has enough momentum to move onto the next step so the internal state machine changes from push->lift->swing->drop when contact occurs the state machine either goes into a standing state (by command), or onto the next four states of the other foot.

[williamthomasmi10]

@dwidgit332 That is so cool!

Congrats on being a genius!

[jimboscotty]

What is the next step with this project? (no pun intended)

[RaceAcer60]

Great LiL andriod.If i may ask,can it walk without the boom extention rod attached to it?.I wouldnt mind having one of those,thats just pure coolness. I like to build things myself,Good luck with it. :-)

[dwidgit332]

No, this one can't walk without the boom, but I'm working on something new that can. :-)

[AdamNeisler]

Maybe put on some rubber shoes so it has grip.... Also add a pelvis part. Their is more to walking then, having knee and ankle joints. Just my 2 cents

[dwidgit332]

The surface is neoprene to provide grip. With this configuration, you need wheels on the feet to move the legs radially because the radius from the center to the foot changes as the leg swings through. This architecture was selected because it is a particularly difficult configuration to control, and it was a good demonstration of the underlying control approach that I developed. If I could make it with with this configuration, then adding more degrees of freedom should make it more robust.

[Spectic]

Cool!

[randyhelzerman]

gratz dude

[1Vella]

ahahahah how cute

[Raiden4900]

OMG! Dwarf gekko! :O

[subtenko]

LOL ps3's metal gear solid 4 in the making XD

[1ThunderBird2]

WOW! Great Job! Good luck with further advancments!!!

[dimviesel]

the beginning stages of,.....METAL GEAR!

[sh3lbyguy]

Let's just hop they don't arm it with nukes

[newhaveninn]

are you going to make it like ASIMO?

(start with the legs and build up from then)

[dwidgit332]

Going to a full 3d version is an order of magnitude more difficult, but it's in the works.  :-)

[patridas]

Excellent work mate!

[werds00]

looks like the gekkos from metal ger solid 4

cool :)

[GabrielLegarda]

awesome you made it

[tjamassey]

Is it being held up by that attachment or is it balancing?

[dwidgit332]

The boom only supports it laterally. It is free to move up and down, around the circumference of the circle, and rotate about the axis of the boom. Friction is minimal. The purpose of this was to emulate a 2D, planar system.

[123BubberMiley]

wow that's awesome. why does it walk fast then slow? are you controlling it to do that or is it doing that by itself.

[dwidgit332]

It is a challenge to find the right balance between speed and torque (where one is the tradeoff of the other). I leaned towards torque which limits how fast it can go... but there are also two "modes" of walking. Around 10s it switches to a more dynamic gait which basically depend on how hard the push is. Still, with better and more efficient motors, it should be able to go much faster.

[kobasa22cm]

What is it?!

[dwidgit332]

If you want all the gory details on the implementation, my thesis is now online. Since we can't seem to post links in the comments, just google "uwaterloo electronic thesis". On that page follow the link for UWSpace Electronic Theses and Dissertations and search on Derek Wight. My thesis will be at the top of the list.

[phiberoptick]

great work man! when was this video from?

[dwidgit332]

8 months ago... the fall of 2007. I had it first working in the summer, I just didn't get around to capturing some video of it until then.

[crook99x]

of course its no asimo i cant imagine seeing that fall

[dwidgit332]

True, this is just design to emulate 2D rather than 3D, and it only cost $1000 as opposed to $1 million... but even Asimo has fallen. Asimo is based on a Zero Moment Point approach which has the limitation that once its feet are no longer on the ground, it has no idea how to restore its balance. The purpose of my control approach is to address that limitation. Even if it is flying through the air, it can calculated exactly where it needs to place its foot to restore its balance.

[danielr2d2]

thats the right way. Nature is not always flat ground. Even humans are falling often or nearly falling. The important point is, that the robot has to know how to save him of falling. So.... great work.

[hogsy12]

you sound wise :-D

[D4Shawn]

That's actually pretty awesome!

[FlailingDutchman]

Bravo, sir!

But alas, we are still quite a primitive species. :(

[richardperkinstv]

nice, you make it yourself?

[dwidgit332]

Mostly, the brackets and tubing of the legs are from Lynxmotion, the servos are standard hobby servos with custom electronics. The rest of the mechanical system, electronics, software and firmware I made myself. This was part of my PhD thesis.

[zapuruxo]

Very cool!

Is it supposed to sort of hesitate like that? If it's not, it still seems to work very well- awesome job!

[ilikeitsideways]

Very nice work!! 

Any plans to remove the boom?

[dwidgit332]

Thank you! There are no plans to remove the boom for this guy because a full 3D version will require a new/modified design anyway.

[clarxrawr]

That creeps me out.

[dwidgit332]

It applies very little force on the boom, but without it, it would just fall over sideways (there are no actuators for lateral control). The boom is connected to a central hub which has sensors on it though so the boom needs to be a large enough diameter to prevent vibrations which introduce artifacts into the sensors, but it's made of carbon fiber so it's relatively light.

[mmakrzem]

awesome work! How much of the side support is the robot using for balance?  Will it walk in a circle if it didn't have the rod connecting it to the centre?