Building humanoid similar to Asimo and Qrio

[PaulL]

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.

I didn't realize you meant monofilament / fishing line... That stuff tends to have a lot of stretch, and it's very slippery. You definitely want the fine steel cable!

Btw, great drawings - I wish I had better CAD skills, but I'm making do with 2D stuff. What software are you using? I've only found one package I could test drive long enough to feel like I could be proficient with it (Rhino3D).

[limor]

indeed we'll order some alternative wire. i'm using Inventor. seems like we managed to get enough friction between the wire and the pulley for prototyping but I'll order some of that wire discussed previously and also look for other inelastic wire materials such as kevlar.
Quite difficult to find online industrial kevlar wire with strength and width parameters. http://www.thethreadexchange.com/miva/m ... t-selected

[limor]

After building the leg I found several "bugs" in the design like screw holes that weren't matching and spacing between parts that weren't right. but the major issue I see is the height of the robot. In CAD 25cm looks kind of ok for a leg but in reality i don't know of 60+ cm robots based on AX12 or AX18 so its best if i try to shrink down the structure. This can only be done by placing the servos horizontally along the leg line rather than vertially. it should be ok given the Universal joint system with the little hole for the wires. (ie: the line between the center of the pulley can be at an angle vs. the line joint-to-joint






A trick to create friction between the pulley and the plastic wire is to extrude a little pillow in the pulley and link it back with a screw thereby squashing the wire between the two pieces.




Here you can see that the fishing rod's elasticity is a big issue even once the friction between pulley and wire is resolved

<object type="application/x-shockwave-flash" width="640" height="427" 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=53f918db3a&photo_id=7714166382"></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=53f918db3a&photo_id=7714166382" height="427" width="640"></embed></object>


Here the bevel gears needed some post-3D-printing smoothing so that the Bioloid plastic washer and nut can fit in. We will eventually have to make these gears in derlin or nylon using the CNC.. sorry for the lack of auto-focus in my Nexus phone's latest Cyanogen OS.

<object type="application/x-shockwave-flash" width="640" height="427" 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=9489e850e5&photo_id=7714157236"></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=9489e850e5&photo_id=7714157236" height="427" width="640"></embed></object>

[limor]

Here's the new servo format of the leg to shorten its length

[limor]

I went to the local sports superstore and bought a couple of "braid" fishing lines which seem to not have any elasticity. One is made with Dynema and the other a mixture of Gore(tm) and Dynema.
Both are supposed to have "very low" elasticity but I saw that there's recommendation to use triple fisherman's knot if material is used in climbing ropes as they may break. Don't know how applicable this is to my pulley system. Another concern is that they are very thin - less than 0.2mm. Can this tear through the ABS pulley groove?.

[PaulL]

That video really shows how much flex you can get from some of that stuff! Even at much less levels of stretch, you could easily end up with "bounce" in an axis - that's been my concern with my pulley design. I know a fixed belt would do for what I need if the cable doesn't work out - could you use a fixed or toothed belt in your design?

Shorter arm / leg segments is better, less leverage working against the servos and pulleys, better effective torque. I have never heard anyone complain that their 'bot had "too much torque"...

Tear - maybe, but more like CUT! Depends on how much tension is needed, and how much load is applied. I wouldn't put my finger in such conditions!

Excellent work, btw!!!

[Tyberius]

Can I ask why you wouldn't just start with MX-28s as your base servo?

Seems that developing this level of hardware for AX-12s is sorta counter-intuitive as they have a fairly low performance ceiling and making it walk will be a struggle with the lower torque. You'd also have PID at your disposal for fine tuning the response on your pulleys.

[garycmartin]

Can I ask why you wouldn't just start with MX-28s as your base servo?

Seems that developing this level of hardware for AX-12s is sorta counter-intuitive as they have a fairly low performance ceiling and making it walk will be a struggle with the lower torque. You'd also have PID at your disposal for fine tuning the response on your pulleys.

Limor did mention some leg servos will likely need to be AX-18s, but I'm guessing many more folks have AX-12s and may be interested if any of these parts ever become available as a kit (would be great to see that happen).

FWIW: I have a 57cm tall 22DOF build that can hold its weight (2210g) and walk (if not yet with confident strides), 20 AX-12 and 2 AX-18 in the hips. The down side to AX-18 is the extra power draw & heat, notably more than AX-12s.

If you can afford them (!!!!) MX-28 would seem to give a similar power profile as the AX-12, but with almost double the torque, and better accuracy. Even at the rather frightening cost (for a hobbyist), I am tempted to get a bulk pack of 6 MX-28s for hip/knee/ankle use, but that'll also be quite an effort making robust hybrid brackets needed for holding the AX-12 / MX-28 pairs (I'm a SW guy more than a HW).

[Tyberius]

Can I ask why you wouldn't just start with MX-28s as your base servo?

Seems that developing this level of hardware for AX-12s is sorta counter-intuitive as they have a fairly low performance ceiling and making it walk will be a struggle with the lower torque. You'd also have PID at your disposal for fine tuning the response on your pulleys.

Limor did mention some leg servos will likely need to be AX-18s, but I'm guessing many more folks have AX-12s and may be interested if any of these parts ever become available as a kit (would be great to see that happen).

FWIW: I have a 57cm tall 22DOF build that can hold its weight (2210g) and walk (if not yet with confident strides), 20 AX-12 and 2 AX-18 in the hips. The down side to AX-18 is the extra power draw & heat, notably more than AX-12s.

If you can afford them (!!!!) MX-28 would seem to give a similar power profile as the AX-12, but with almost double the torque, and better accuracy. Even at the rather frightening cost (for a hobbyist), I am tempted to get a bulk pack of 6 MX-28s for hip/knee/ankle use, but that'll also be quite an effort making robust hybrid brackets needed for holding the AX-12 / MX-28 pairs (I'm a SW guy more than a HW).

Due to the higher heat generation and current draw, I don't think the AX-18 should ever be considered a torque upgrade to the AX-12. I essentially consider them the same, just faster.

If there are pulleys being implemented, why not trade some of the 300 degrees of rotation for extra torque by implementing some gearing?

[limor]

The reason i'm focusing on AX12 is affordability. We intend to make this into a kit so it will cost a bit more than Bioloid.

The main issue with torque is in the knees getting up, walking up the stairs. Walking knee torque requirements will depend on how the walk management will be done. I may put two servos on the knee if the AX18 doesn't do the job.

We have a couple of GPs that use AX18 and we haven't experienced the AX18 heating up as you describe.

The reason I wanted these universal joints is for aesthetics, so that they can be better covered and the leg will look more streamlined like Asimo. The way Robotis implement joints is to use the servos as structural components, not just actuators. this has a downside in not being very pretty and can't be well hidden under a shell.
So to relay the servo output to the joint, i first thought of using a belt. But the available belts for the precision and torque needed, were too bulky. Using gears for precise rotation with heavy torque would be an option, but a more complicated one to fabricate. For the upper arm rotation (Asimo style, instead of wrist), I used a big gear which can be 3D printed. But the torque on this gear is not great so plastic should suffice.

Hence the idea of using wires. In theory there should be 0 backlash using those kevlar fishing wires. I've been away this week so I couldn't test this further. But the design is continuing and next week we'll hopefully see some of these wires in action. I've rotated the servos so they are horizontal in the legs. The height of the robot now looks to be around 60cm



The pulley on the AX12 was modified to serve also to fasten the wire and to tighten the wire. The oval thing with the hole has a screw through it and it squashes the wire which also wraps around the screw to improve the friction area). The horizontal screw pushes against a small piece of plastic where the wire will pass so it stretches the wire tight. There are 4 tunnels in this pulley to allow the wire to pass through.

[siempre.aprendiendo]

It's starting to look great. When do you think you will have a first protoype?

[limor]

It's starting to look great. When do you think you will have a first protoype?

I think a prototype without the polycarb shell should be within a couple of weeks. The polycarb prototyping requires CNC'ing blocks of wood or other dense soft material that will be used as mold for vacuum forming.

One thing that's blocking progress now is I need to order some sheets of Derlin or Nylon in order to CNC the bearings. All the other parts are 3D printable although there's still work in adding support material here and there because the automatic support material functionality in ReplicatorG is not as good as one can get by doing it manually.

[mog123]

It's starting to look great. When do you think you will have a first protoype?

I think a prototype without the polycarb shell should be within a couple of weeks. The polycarb prototyping requires CNC'ing blocks of wood or other dense soft material that will be used as mold for vacuum forming.

One thing that's blocking progress now is I need to order some sheets of Derlin or Nylon in order to CNC the bearings. All the other parts are 3D printable although there's still work in adding support material here and there because the automatic support material functionality in ReplicatorG is not as good as one can get by doing it manually.

You can always go with the cheaper option - get some styrofoam and sculpt it according to 3d drawings. Why always include heavy and expensive machinery?

Can I ask why you wouldn't just start with MX-28s as your base servo?

Seems that developing this level of hardware for AX-12s is sorta counter-intuitive as they have a fairly low performance ceiling and making it walk will be a struggle with the lower torque. You'd also have PID at your disposal for fine tuning the response on your pulleys.

Not everyone works for a robotics company that (what a coincidence) sells servo's that you're advertising. I know they are great and all, but people can manage with much much cheaper.

[Tyberius]

The reason i'm focusing on AX12 is affordability. We intend to make this into a kit so it will cost a bit more than Bioloid.

The main issue with torque is in the knees getting up, walking up the stairs. Walking knee torque requirements will depend on how the walk management will be done. I may put two servos on the knee if the AX18 doesn't do the job.

We have a couple of GPs that use AX18 and we haven't experienced the AX18 heating up as you describe.

Affordability is a good reason.

The GP doesn't exhibit this behavior because it's super light weight. Trying to build a heavier robot where it puts the AX-18A under a tad more load, and you'll see the heat issue come into play. My experience on this stems from my own experience building a smallish custom AX-18 biped, and half a dozen builders in Mech Warfare who chose AX-18s over AX-12s. The common factor with all AX-18 based mechs: overheating. I think they're great for places where you need the speed, but I can't recommend them over an AX-12 for reasons of torque alone.

Hence the idea of using wires. In theory there should be 0 backlash using those kevlar fishing wires. I've been away this week so I couldn't test this further. But the design is continuing and next week we'll hopefully see some of these wires in action. I've rotated the servos so they are horizontal in the legs. The height of the robot now looks to be around 60cm

I haven't seen an AX based biped in the 50-60cm range that could actually walk yet, but I'd love for someone to prove me wrong! That was another reason I mentioned the MX-28; the bot looks pretty big.

Not everyone works for a robotics company that (what a coincidence) sells servo's that you're advertising. I know they are great and all, but people can manage with much much cheaper.

Uhh. "Advertising" is quite a stretch. I don't see a link to my company nor a link to the MX-28 anywhere in my post. Also, Robosavvy sells them here, so if anything I'm promoting them on their behalf.

I was suggesting it because I've built half a dozen various sized custom bipeds using the Dynamixels and happen to have a good idea of what the AX-12s are capable of. I'm more than allowed to participate in an engineering conversation as a hobbyist, regardless of my employer. My sole intent was to contribute to the conversation and had zero to do with "advertisement", but thanks for jumping to that assumption.

[mog123]

Not everyone works for a robotics company that (what a coincidence) sells servo's that you're advertising. I know they are great and all, but people can manage with much much cheaper.

Uhh. "Advertising" is quite a stretch. I don't see a link to my company nor a link to the MX-28 anywhere in my post. Also, Robosavvy sells them here, so if anything I'm promoting them on their behalf.

I was suggesting it because I've built half a dozen various sized custom bipeds using the Dynamixels and happen to have a good idea of what the AX-12s are capable of. I'm more than allowed to participate in an engineering conversation as a hobbyist, regardless of my employer. My sole intent was to contribute to the conversation and had zero to do with "advertisement", but thanks for jumping to that assumption.

I sincerely apologize for that post then. It just ticks me off when people ask "why not go with the best and most expensive option?" because that's when they make the assumption that other people have a lot of free cash to spend.