Building humanoid similar to Asimo and Qrio

[limor]

so i've started designing the aluminum sheet CNC which will hold the torso together. 2 delrin bearings hold the leg in place as it rotates (Roll DOF). I'm not convinced still about the current leg format and that big radius back-of-the-knee axis. The problem is that the AX12 stick out and i'm trying to avoid having to design another part because it takes hours churning these different parts out of the 3D printer)

I noticed that Asimo has a torso vertical rotation axis which helps it look human when it walks. Qrio has a horizontal DOF in the torso which i'm guessing was intended to help him bow (Japan..)

[limor]

This is better (smaller back-of-knee radius) but AX12 stick out.
I'll have to do some cosmetic adjustment to the polycarbonate design

[MOHIT JINDAL]

How will you move them ? From Pulley and Belts ?
And how much is the length ? 30 cm each like ASIMO ?
BTW very good Drawings.And you are working hard on it. Great

[limor]

I use the AX12 servos with a custom pulley system instead of a thick plastic/metal belt. The fishing rod can sustain huge pull without deforming and I designed a special pulley trick to create adjustable tension. (i'll post some pictures)

The height will be around 60cm (so far the length of the legs without the feet is about 28cm.
I'm working quite a bit on this project the past couple of weeks. The CAD + 3D printing allows for an iterative process of development. Shortly we will have to put together the leg with 3D printed parts and use standard Robotis CM5 or USB2Dynamixel to see how things move.

[limor]

simulated metal bending oddities resolved.
Bioloid feet will do for now because i want to focus on the aluminum system holding the arms in place.
The strange bell shaped thing on his torso will become the chest with arms coming out of the sides at 45degrees (Asimo style).



Turns out Qrio has 5DOF in the arm. This allows him to pitch a baseball like a pro. Guess I'll to give 5DOF per arm to my robot.
The problem is that couple of AX12 in the leg may need to be upgraded to AX18 due to the extra weight and heigh. fortunately AX18 with same form factor as AX12 give a couple of notches more torque.

Also, I'll probably throw into the design CM5 and CM530/CM510 controler form factors so that owners of old Bioloids can upgrade and so that Roboplus will work out of the box.

However i'll definitely throw into the design, a high Mhz embedded linux running in the robot to control the servos, camera, wifi and arduino to process some touch sensors.

[PaulL]

Hey Limor,

See below pulley mounted on Hitec HSR-5498SG hub, this is a pulley design I intend to use for hip / wrist rotation on an RN-1. The pulley groove isn't cut, but you get the idea. The nylon coated stranded wire (available at www.mcmaster.com nylon coated wire rope) has a 26 pound breaking strength, and is extremely thin (.019 inches, or around .5 mm), they probably have something that will work for your application. Alu in pic is .040 inch thick, about 1mm. Recess cut is .020 inch deep, about .5 mm.

The wire will mount first on the top screw in the picture, the wrap 90 degrees around the first pulley, then over to the other, around 90 degrees, into the recess like in this pic and looped around the single screw near the channel opening (to the right in the pic), around 90 degrees, and back to the first pulley. Tightening will be via the bottom screw in the pic. The channel opening (right in the pic) will be opposite for the two pulleys, allowing for about 180 degrees of rotation.

The finish is rough around the edges because this is 5052, I figured I'd throw a couple pulleys on the last sheet of 5052 I cut.

[MOHIT JINDAL]

Limor, now i am waiting for pulley thing. Hope to see your pics soon. Maybe a small video, how fish rod turn each joint.
I am wondering A pulley system is good to get more torque or A big Gear box is best ?

[limor]

Thank you for posting this.
How wide is the pulley? have you tested it in the robot ?

I found this manufacturer that makes jewelry wires. 49 strand stainless steel nylon coated. 12kg resistant. 0.61mm (.024") width. http://www.beadalon.com/49strandsatin.asp

Initially I thought of doing something similar to what you did.



But then I realized that this configuration may limit the possible resolution of the angles of rotation. If the wire is wound twice around the pulley (using something like a V profile pulley), then you can get a full 360 degrees of rotation resolution.

I prefer using plastic for the pulley for now and I'm not sure about using screws to hold the wire in place as these can break the plastic when the wire is under tension. So the the plan is to use some variant of a fisherman's knot to close the wire loop.



There should be enough friction in the contact areas with the pulley to prevent slippage I hope without the need for screws. I made a little tensioner that can be configured with a screw. the wire is looped twice around the pulley and then driven through a small tunnel through the pulley to the little tensioner. The tension applies a force on a little nut that is embedded in the plastic in addition to the threads of the screw.

[PaulL]

Thank you for posting this.
How wide is the pulley? have you tested it in the robot ?

No problem! It's .040 inch, about 1 mm wide. I have a turned adapter waiting to be finished on my lathe to cut the pulley groove, the Hitec hub is too thin to grasp in either my 3 jaw or 4 jaw chuck. Haven't tested it in the 'bot yet. My reason for wanting to use a pulley was to add a DOF without upsetting the geometry of the RN-1 too much, and have closer to 180 degree travel than with a pushrod with a thin profile.

I found this manufacturer that makes jewelry wires. 49 strand stainless steel nylon coated. 12kg resistant. 0.61mm (.024") width. http://www.beadalon.com/49strandsatin.asp

That should work! 26.45 pounds, very similar to what I found. I was amazed to find that this kind of stuff is readily available.

Initially I thought of doing something similar to what you did.

...

But then I realized that this configuration may limit the possible resolution of the angles of rotation. If the wire is wound twice around the pulley (using something like a V profile pulley), then you can get a full 360 degrees of rotation resolution.

Yes, I've seen that done in old plotter - printers (multiple loops on the pulley). I have a hub from one right here. They just cut a flat bottom on the outer diameter, with two "rails" on the edges to keep the cable on it (no true "V"). If I remember right, they just used a spring to tension the cable, but then space wasn't an issue for them (and the load was very light compared to bots!).

I am only looking for 180, and to add loops will mean I need a thicker pulley - it's a slightly different application.

I prefer using plastic for the pulley for now and I'm not sure about using screws to hold the wire in place as these can break the plastic when the wire is under tension. So the the plan is to use some variant of a fisherman's knot to close the wire loop.

I like the closed loop idea to get around having to worry about the tension of terminating the cable with a screw. I would think you'd be fine with your approach - loop the wire through the pulley, knot it, slip it on your pulleys, tension it. You'd need just enough adjustment for the difference to slip the loop onto the pulley.

There should be enough friction in the contact areas with the pulley to prevent slippage I hope without the need for screws. I made a little tensioner that can be configured with a screw. the wire is looped twice around the pulley and then driven through a small tunnel through the pulley to the little tensioner. The tension applies a force on a little nut that is embedded in the plastic in addition to the threads of the screw.

In this plotter wheel, they just do hard right angles, it locks the cable down pretty well (entry point of cable from outside to inner area of the pulley).

See below, this is from a very old plotter. This was a driver wheel, attached directly to a stepper motor. You can see how they terminated the ends - they used two screws on top of the pulley (threaded holes, wire is tucked into one in the pic). This had multiple loops for travel on the plotter, but you can get 360 with one extra loop. Note the "groove" shape - just a flat channel.



I do have concern for this approach in our applications - I haven't done any "stress stretch" testing yet. I need to hang a weight of several lb / kg and see how it bounces, how much force will make the cable stretch. I hope it isn't too springy.... Next step for me would be a flat belt approach.

[limor]

Thanks for the picture!
No idea is ever original i guess.

[limor]

I added the upper arm rotation as used in Qrio and Asimo.
The lower arm and hand are still under construction.
I'm thinking of making a small hand and putting a couple of mini servos + arduino to give the hand a couple of moving fingers (similar style as the inmoov robot hand but with fewer DOF)

[limor]

I'm positioned the arm on the body. i think the proportions are ok. It will be interesting to see how the servos internally will be aligned at an angle and the whole chest needs to rotate aswell. Already 21 servos with the rotating chest and without hands and head. Should still be within range of acceptable torque for the knees and feet with AX18
climing stairs using one foot at a time, means one knee has to be able to lift the whole torso at perpendicular upper leg. ie: 18kg*cm / 15cm = 1.2kg thats the maximum weight of the torso. and this is theoretical. probably the real maximum weight of the torso should be under 1kg if the upper leg is 15cm in length.

[limor]

Today we got some 3D printing done. It takes some knowhow to get decent details with Makerbot and the likes. eventually we managed to get the universal joint all together







Turns out the friction between the plastic and the fishing rod is too low and there's too much friction in the universal joint hinges (Bioloid plastic parts) because they area squashed because the 3D printed holes are deformed etc. so some drilling needs to be done to improve the holes but with regards to fishing rod maybe a screw will hold the wire in place.

generally dealing with these tiny bits is tricky but after half an hour of playing around i think i can assemble the joint in 10 minutes.

<object type="application/x-shockwave-flash" width="400" height="267" data="http://www.flickr.com/apps/video/stewart.swf?v=109786" classid="clsid:D27CDB6E-AE6D-11cf-96B8-444553540000"> <param name="flashvars" value="intl_lang=en-us&photo_secret=bd49edac2c&photo_id=7700426730"></param> <param name="movie" value="http://www.flickr.com/apps/video/stewart.swf?v=109786"></param> <param name="bgcolor" value="#000000"></param> <param name="allowFullScreen" value="true"></param><embed type="application/x-shockwave-flash" src="http://www.flickr.com/apps/video/stewart.swf?v=109786" bgcolor="#000000" allowfullscreen="true" flashvars="intl_lang=en-us&photo_secret=bd49edac2c&photo_id=7700426730" height="267" width="400"></embed></object>

[limor]

Upper body rotates so it needs a wide circumference low friction contact with the torso otherwise there will be side to side wobble.. (I'm thinking maybe some part circle shapes in Delrin or Nylon will do the trick)

[PaulL]

Thanks for the picture!
No idea is ever original i guess.

No problem, just thought you might want to see the "groove" in the pulley.

... I think technology is in large part itterating through context at this point. The originality comes from the combination of concept and context in which it is used.