🔒 Locked

Biped Walking

Started by bzroom Jan 31, 2006 at 9:19 PM 15 replies 2.1k views

Original Post

Jan 31, 2006 at 9:19 PM

I've been reading up on biped walking ever since this QRIO video was posted. From what I can understand biped walking goes like this. With a simple pair of legs, its movements are limited by the amount of torque available to excert on the ground, not exceeding the actuator and foot size limits. With this information, calculate hip and swing foot trajectories that never cause the stance to over torque considering all momentum the moves and gravity are affecting. Any idea if I am close? Right now I'm just trying to get a simulator running, that will allow me to tweak the equations and tell me if it will fly or not. It's also going to drive the servos in the real world. I'm eyeing this (1x) and this (4x). Hopefully soon I will more completely understand the physics envolved and make the trajectory calculation include the momentum limits instead of guess and check. Which I believe technically is the the difference between static and dynamic walking. I'm posting to poll any information Gamedev may have about biped walking. It's a challenge but I'm sure someone has some suggestions. The simulator currently looks like this. The equations and mass's I've used have no realistic value, once it's tracking time, I'll get it all to scale and it will output all the physical data. Image hosting by Photobucket

You can download it here, but it requires the .net framework which can be found here.

choffstein

Jan 31, 2006 at 9:50 PM

I have almost 0 idea what you are talking about, but it seems freaking cool.

Shamino

Jan 31, 2006 at 9:53 PM

Don't you have to take in consideration the swing and momentum the arms would give you?

Luckless

Jan 31, 2006 at 9:58 PM

Are you tracking this in 2D, or 3D for the forces?

Also try sucking the hips up higher, your walker is going to be wasing a lot of power of a minor thing like having the top float a little more than bounce.

bzroom

Jan 31, 2006 at 10:17 PM

Quote:
Original post by Shamino
Don't you have to take in consideration the swing and momentum the arms would give you?

Yes, if you had arms. Right now I'm only considering the leg's and a fixed torso.

Quote:
Original post by Talroth
Are you tracking this in 2D, or 3D for the forces?

Also try sucking the hips up higher, your walker is going to be wasing a lot of power of a minor thing like having the top float a little more than bounce.

This is only in 2d for now. The trajectory for the hip right now is just a straight line. I do need to put an arc on it so as you said hips are not floating and "ride" over the stance leg. But I don't want the hips to ping pong because that would really jerk the system, I need to get my math straight first.

I need to create some polynomial equations that give me some really nice loop-able curves. That will allow me to modify the whole step with parameters such as velocity, step length, step height (for stairs).

Shamino

Jan 31, 2006 at 10:20 PM

I hope you realize that this is like trying to invent the theory of relativity.

As in, no one has yet found an algorithm for dynamic walking on a biped, or a quadroped for that answer, not even a fish.

AnonymousPosterChild

Jan 31, 2006 at 10:28 PM

Quote:
Original post by Shamino
I hope you realize that this is like trying to invent the theory of relativity.

As in, no one has yet found an algorithm for dynamic walking on a biped, or a quadroped for that answer, not even a fish.

Wrong. Many studies on fish motion and algorythms have been done, we just haven't been able to create a working model outside of a computer.

Honayboyz:

If you use a system where the legs connect ABOVE the bulk of the frame, you could probably solve a lot of balance issues.

bzroom

Jan 31, 2006 at 10:41 PM

Quote:
Original post by Shamino
I hope you realize that this is like trying to invent the theory of relativity.

As in, no one has yet found an algorithm for dynamic walking on a biped, or a quadroped for that answer, not even a fish.

...I had to look up theory of relativity.

I beg to differ. Both Honda and Sony's new robot's appear to be a fully dynamic setup. They can cope with varying terrain, unexpected outside forces such as pushes, and when they are about to fall they try to recover or prepare to crash. And Sony's Aibo (quadraped) at least knows when it's upside down.

Shamino

Jan 31, 2006 at 11:39 PM

Quote:
Original post by honayboyz
Quote:
Original post by Shamino
I hope you realize that this is like trying to invent the theory of relativity.

As in, no one has yet found an algorithm for dynamic walking on a biped, or a quadroped for that answer, not even a fish.

...I had to look up theory of relativity.

I beg to differ. Both Honda and Sony's new robot's appear to be a fully dynamic setup. They can cope with varying terrain, unexpected outside forces such as pushes, and when they are about to fall they try to recover or prepare to crash. And Sony's Aibo (quadraped) at least knows when it's upside down.

That is really cool but it is super cutting edge technology...

And it is probably at the point in stage of research where the equivalent to einsteins book of relativity is being written still..

So essentially, You're trying to discover relativity while einstein is writing his book..

But, by all means, it is definately a FUN thing to do, if that is what you like. Very cool experiment still :)

Shamino

Feb 1, 2006 at 12:07 AM

I must admit, I read that whole site about sony's QRIO...

AMAZING.

PERIOD.

Think Chobits (an anime)

And already, Sony's marketing ploy has caught me. QRIO is so damn cute!

"QRIO what is your favorite thing?"
*he puts his hands on his hips, looks into the sky, thinks...*
"I like dreams!"

The ASIMO from Honda looks like a space man, less a robot..

I Like Bread

Feb 1, 2006 at 9:00 AM

Quote:
Original post by honayboyz

A person's legs will NEVER buckle that much, unless he has crippling shit spasms or something. Realistically, the human gait is minimalistic, the legs doing little more than to counteract gravity during the forward momentum of the torso.

ViLiO

Feb 1, 2006 at 9:37 AM

Quote:
Original post by Shamino
I must admit, I read that whole site about sony's QRIO...

AMAZING.

PERIOD.

Think Chobits (an anime)

And already, Sony's marketing ploy has caught me. QRIO is so damn cute!

"QRIO what is your favorite thing?"
*he puts his hands on his hips, looks into the sky, thinks...*
"I like dreams!"

The ASIMO from Honda looks like a space man, less a robot..

Hate to break it to you, but sony have cancelled it [sad]

Iron Chef Carnage

Feb 1, 2006 at 1:19 PM

That's not the sort of thing that'll ever be profitable, but it's an important step toward my robot legion of horror.

I'd also like to point out that QRIO has "emotions" and sometimes doesn't follow instructions when it's in the wrong mood. Let me give you a little scenario:

Me: QRIO, go get me a beer.
QRIO: Negative.
Me: (agitated) Listen up, you bipedal iPod (biPod, if you will), I told you to get be a beer, and you'd better hop to it!
QRIO: (looking at me with eyes that are now suddenly glowing red) Nega-
Shotgun: BOOM!
Me: You like that, you little twerp?
QRIO: Default heuristics engaged. Must Crush Hu-Mans! Must Crush Hu-
Shotgun: BOOM! Click-clack, BOOM! Click-clack, BOOM!

I hate robots.

Steadtler

Feb 1, 2006 at 1:22 PM

Seems like that thing is ignoring balance, which is the biggest problem in biped movement.

Iron Chef Carnage

Feb 1, 2006 at 1:36 PM

Isn't the Segway guy working on a stair-climbing all-terrain leg system for parapalegics? It probably has the necessary software for balancing.

I'm amazed by the ability of anything, including humans, to balance a bipedal system. It'll be an awfully long time before any robots hit the Olympics. Even weight-lifting, which seems like it would be robot-dominated, requires so much fine motor control that there's no way current tech could do the job.

You could have the balance system on a purely reactionary connection. Just use the functional equivalent of an upside-down joystick with a weight on it to get an idea of how fast and in what direction the bot is tipping, and then swing a leg out there to stop it. If you get it just right, the thing would stagger around drunkenly, then run into a wall and fall down. I'd pay a nickel to see that.

bzroom

Feb 1, 2006 at 3:18 PM

Quote:
Original post by Iron Chef Carnage
You could have the balance system on a purely reactionary connection. Just use the functional equivalent of an upside-down joystick with a weight on it to get an idea of how fast and in what direction the bot is tipping, and then swing a leg out there to stop it. If you get it just right, the thing would stagger around drunkenly, then run into a wall and fall down. I'd pay a nickel to see that.

Like this? or tripple the fun with something like this. I can't wait to actualy have a pair of these legs standing there, and then watch them trip up over and over.

Yes right now the program does ignore "balancing" but thats because balancing involves physics which are not yet supported.

Poor balancing robot getting thoroughly abused by his maker.

[Edited by - honayboyz on February 1, 2006 5:18:44 PM]

Iron Chef Carnage

Feb 1, 2006 at 11:57 PM

That was pretty neat, but the wheels undoubtedly make it simpler.

I've seen vids of bipedal bots on a rigid tether, running and jumping around a central point, like the training rigs for RC helicopters. QRIO obviously had disproportionately large feet, and move very deliberately.

I wonder how far away we really are from matching human-style ambulation.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.