Weight repartition vs torque, balance

[Stephane]

Hi All,

I am trying to build my custom humanoid frame, using bioloid servos and I am currently working on the legs.
I can't make a decision on how should the weight should be repartited in the leg:

Initial solution was to keep the center of gravity of the leg as low as possible.
The points in favor are:

It will drag the overall robot center of gravity to a lower position and give it more stability.
It will reduce the load on the ankle servos.

More recent reflection rather rather set it in a higher position in the leg.
What motivates this new option is:

It is more like what we can see in humans or animals (especially birds).
It reduces the load on the hips servos.
The ankle's action is mainly about stabilizing the humanoid, working at small angles from the equilibrium position, so the torque expected from the servo is not very high.

Before I start drilling and cuting again, I'd like to benefit from your experiences on this.
From balance and torque perspectives, which solution would you favor?

Thank's for sharing.

Stéphane

[FabinovX]

Hi,
i'm a beginner too and i apology to flood your thread.
but according to your first name, you're certainly french ?
so i am.
as you can see, the 'espace francophone' looks like a no man's land.
It will be cool we try to develop this part of this really great forum in order to motivate french people to come here to share their own experience and knowledge.
Obviously, we will continue to share with English people too !!!
Merlin seems to be french too, and i asked him to do so.

Good luck with your legs and i apology one more time for the flood.

See you
Cheers
FabinovX

[mog123]

Unfortunately you are wrong.
The ankle servo need to have the highest torque in a humanoid, this is because when you stand on one foot the whole wight of the robot is shifted on just that 1 servo. With a center of gravity in a radius of 10cm you can only balance with 1/10th the momentum you have.

Let's say that the center of gravity is 10cm away from the ankle. Your robot weighs a total of 2kg. You need at least a 20kg/cm torque servo, to balance it freely. That's why building big humanoids is so difficult. The stability is really weak.

[billyzelsnack]

I dunno. I'd think that 10cm is quite far for a robot that weighs only 2kg.

[i-Bot]

I too am struggling to understand the answer from mog123.

When you say the cog is 10cm away , do you mean radially or is that the horizontal displacement of the cog from the ankle pivot ?

If it is horizontal displacement, you need a sin(theta) term, which reduces the torque required significantly in the angles possible for the ankle.

Or can you draw a diagram to help me understand your configuration ?

Your calculation feels wrong because the human ankle though a strong bearing does not have so much torque.

[luison]

You made a Biped?

Generally, the people thinks that the ankle is the most weak articulation in a robot... but when makes one, people realizes that the ankle and the hip (the 5th DOF) have to be the most stronger. If you don't believe me, check how Giger robot, changed her 64KKg/cm for other with 100Kg/cm because the hip haven't got enough power... not the knee like you suppose.

On my first robot, i used 6Kg for the ankle and the hip, and 8Kg for the rest... then i realized that i need more power in the ankle and hip, so i changed it for ones with 8Kg... and finally for others with 13Kg.

Now, on my second robot, I use 30Kg in all my articulations, but only for have the same servo.

[i-Bot]

I thought about this some more and I think the measure is the horizontal displacement of the COG from the servo pivot. The maximum would be when the COG goes over the edge of the foot, so stability is no longer maintained.
Does this mean the maximum torque of the ankle servo is the distance from the ankle pivot to the edge of the foot, times the mass ?
I realise this is a static case.
Is the ankle really the highest torque ? I guess the load as you raise your body on the balls of your feet is pretty awesome torque. and you can feel the muscles working in the leg. Some other parts of the body, though lower mass are further from the pivot. I'm interested to know, but too lazy to study the maths !

[i-Bot]

Thanks luison, What is the mass of your biped ?

What is the body doing when the hip torque is a maximum ?

[luison]

Thanks luison, What is the mass of your biped ?

What is the body doing when the hip torque is a maximum ?

My first bot, has a leg of 23cm and weights 2,6Kg.
My second robot, has a leg of 30cm (aprox) and weights 3,1Kgs.

The maximum torque is when the robot is on one leg, for example walking... it has to support not only the normal weight, that in the first bot is: 2,6Kg x 5cm aprox (for centre the weight on the foot) plus the lateral acceleration. 2,6X5=11,2... extremely near the top torque of the servos... and generally, the torque is calculated on stall move...

[SK]

If you don't believe me, check how Giger robot, changed her 64KKg/cm for other with 100Kg/cm because the hip haven't got enough power... not the knee like you suppose.

Well have a look at the robots of my team ( http://www.dribblers.de/, http://www.tzi.de/humanoid/bin/view/Website/Teams2010 or link in my sig ). We´re using RX-28s for all joints but knees and the waist (for which RX-64s are used). Especially the knee RX-64 are a result of lessons learned when our robot was completely DX-117 servo equipped in its first iteration.
I think the amount of required servo torque is largely a function of kinematics/geometry as well as the walking gait used. On our HR30 robots, using RX-64 in the knee and RX-28 in the ankles and hip is a good solution.

[billyzelsnack]

Well have a look at the robots of my team ( http://www.dribblers.de/, http://www.tzi.de/humanoid/bin/view/Website/Teams2010 or link in my sig ). We´re using RX-28s for all joints but knees and the waist (for which RX-64s are used). Especially the knee RX-64 are a result of lessons learned when our robot was completely DX-117 servo equipped in its first iteration.
I think the amount of required servo torque is largely a function of kinematics/geometry as well as the walking gait used. On our HR30 robots, using RX-64 in the knee and RX-28 in the ankles and hip is a good solution.

Thanks for the info. I figured the non RX-64's were just in upper body.

[billyzelsnack]

My first bot, has a leg of 23cm and weights 2,6Kg.
My second robot, has a leg of 30cm (aprox) and weights 3,1Kgs.

The maximum torque is when the robot is on one leg, for example walking... it has to support not only the normal weight, that in the first bot is: 2,6Kg x 5cm aprox (for centre the weight on the foot) plus the lateral acceleration. 2,6X5=11,2... extremely near the top torque of the servos... and generally, the torque is calculated on stall move...

Do you have pics or vids of your robots online anywhere? I'd love to see them.

[luison]

Do you have pics or vids of your robots online anywhere? I'd love to see them.

Yes, i have vids of my first robot, but it never works as well as i spected, because the weak servos.

http://www.youtube.com/watch?v=Y8sRPNMDV5k

From the second, only pics, the following month we make a competition in Madrid, so i will upload videos if you want.

http://img3.imageshack.us/i/dsc00653b.jpg/

http://img188.imageshack.us/i/dsc00652b.jpg/

The second, is really really impressive, with two axles servos modified to fit in the SES, open body, CAM, giros, accelerometers... and a really really cool library make for the AXON. I'm writing all the libraries for the AXON because the originals are really bad (i think).

[luison]

If you don't believe me, check how Giger robot, changed her 64KKg/cm for other with 100Kg/cm because the hip haven't got enough power... not the knee like you suppose.

Well have a look at the robots of my team ( http://www.dribblers.de/, http://www.tzi.de/humanoid/bin/view/Website/Teams2010 or link in my sig ). We´re using RX-28s for all joints but knees and the waist (for which RX-64s are used). Especially the knee RX-64 are a result of lessons learned when our robot was completely DX-117 servo equipped in its first iteration.
I think the amount of required servo torque is largely a function of kinematics/geometry as well as the walking gait used. On our HR30 robots, using RX-64 in the knee and RX-28 in the ankles and hip is a good solution.

Seems like differents algorithms produces differents results.
I know perfectly "your" robots

[mog123]

I didn't think people wouldn't understand my post, so here's a picture with an explanation:



blue = center of gravity
green = 'lever'

Momentum = Force x lever (radius)

That is wy the ankle needs to be the strongest part in a humanoid.