I also thought about
I also thought about proportions. When do children start to walk? At 2 years, according to google, they should be able to walk well. At 2 years a typical child is about 80 to 90cm and about 10-15kg. And these creatures can walk straight. My estimations is also a structure of about 80cm height, but maybe about 8 kg (about 1kg per limb), so not that far from the “nature”.
But, overall, what I can see at Dr. Guero’s examples is that the size does not affect that much on the way how the structure behaves, because resistance and self-balance for me is most important. And the fact that I am choosing this structure is because I want to have better limbs than a “typical” connected servos can provide.
Size
That is too big for hobby servos and takes you into maybe one of the NEMA 23 steppers and a gear box of some sort and of course stepper controllers for each. That is quite a project.
If you cut that in half you can use some of the hobby servos in the 10 to 20 kg-cm range. If you work out the design and software there you can scale up. I had found a nice listing of hundreds of steppers with specs and good prices, but I can’t find it now.
What actuators are you considering?
Yes, I know. I am inclined
Yes, I know. I am inclined now to NEMA 11… And they fit with Actobotics well.
I am also still considering DC motors with encoders vs steppers (waiting for delivery to check)… but most likely steppers with gearboxes… I liked Phidgets motors because of possibility to install encoders on them. Already tested this NEMA11 motor, but it seem to be too weak, will not fit for everything, will need a gearbox. So far this one seem to be the best (considering there will be 2:1 bevel gear for the joint.
And now I am also calculating overall weight, and thinking about batteries. 12V 2800Ah battery will take up to 500g… :-/ and just for legs and pelvis with minimum 7 NEMA11 motors I will need 4,69A current, so at least 2 of these… so 1kilo more minimum… most likely 1,5kg… just for legs. and another 2kg for batterieis for hands and body (but that is second phase).
I have a question about servos, are there any which have continuous rotation and feedback? Coz those what I have seen, were just 180 degrees max. You see, I do not want to connect joints to motors shafts, as it will weaken construction. Instead I am planning joints on bearings and tubes, thus there will be a gear between the motor and a joint.
Servos
I like that stepper you are thinking about, I think it may be doable.
The servo database is here:
http://www.servodatabase.com/servos/all
I am not familair with any 360 servos that have feedback, cutting out the potentiometer is the standard way of making a continuous rotation servo out of a standard one. There is one servo that will do 3.5 rotations due to a geared down pot:
https://www.servocity.com/html/hs-785hb_3_5_rotations.html#.VgMUqd9Viko
Depending on how much movement you actually need to walk, that may be too slow for your project. I don’t know.
In the process of moving, some joints will need to rotate only a few degrees. I have these ranges sorted for a quadruped (which vary widely) but I’ve never looked at bipeds. I suspect that you may only need to move 10 degrees or so on the main joints and you can leverage for greater power if you have the speed.
The range of motion for walking is rather modest. It is all the other things, like sitting, that require the full range.
and another thing, just for
and another thing, just for you to understand how one should “program” a walking cycle: https://www.youtube.com/watch?v=tfio_gpJBPc
reality is that it would be almost impossible to “program” it (or rather too time consuming), paying attention on all the gaits characteristics, instead, it is much more faster to animate it.
Walking Kitten
http://design.tutsplus.com/tutorials/how-to-create-a-walking-kitten-animation-in-adobe-photoshop--cms-23046
The animation is fine if it matches the reality. There is no law that it must.
I see rather obvious flaws in the man, the feet are way too far apart, for one. Fix those flaws and you have something.
The guy who did the kitten pretty much nailed it. I’ve checked against reality and the proportions and movements are correct.
I see the reality of motion not in figuring out positions but in emulating the way forces work naturally. We rely on muscle memory, not inverse kinematic to walk. I don’t look at these as sine waves, that simplifies the reality. I look at them as Bezier curves where your movements are pulled by forces in certain directions. I think that is a better approximation. YMMV.
Just a small update here -
Just a small update here - the biggest challenge at the moment is to balance required torque of the motors with their size. So far what I can see that the biggest impediment in the creation of bipeds is lack of good actuators… steppers are not bad, but seem to be not that efficient, even geared ones… either tough or quick… but I need both. :-/ in progress…
Servos
I’ve noticed several 40kg-cm and larger servos and at not too high a price (just too high for me).
http://www.banggood.com/CYS-S8218-Digital-Metal-Gear-High-Torque-Servo-p-86652.html
That’s plenty big for me, you may need something bigger,
Good luck with your project.
thanks for the link
thanks for the link cyberjeff… please see some updates in the post above.
well, this is a typical
well, this is a typical animations tutorial… I am a bit lazy to search for it, but if you are interested in others, I am sure you can find some tutorials on youtube on character studio and/or CAT plugin for 3dsmax, where all the cycles are actually created using your mentioned Bezier curves.
At this moment I feel like it should be 2 layers - standard mechanical walk cycle and a self-balancing layer on top.
update…1. Overall we got
update…
1. Overall we got painfully undeveloped situation with motors. Steppers are rather weak, servos are rather stupid and expensive. If one would like to use a servo for a biped, it should be 360 degrees rotations. Ones of the best servos for a biped seem to be Dynamixel MX-series. Though they are a rather expensive.
2. Gears… good gear sets (especially right angle ones) for a motor are crazilly expensive (like 1k euro for a nema17 motor, while there are no for nema11)… 
Though they claim they have sort of precision, not sure how actobotics precision will be…
3. Framing. well… not a big selection either… Actobotics and maaaybe a little bit Tetrix (though expensive one). Btw, original dynamixel frames for building their constructions are crazily expensive (50 bucks for 2 cm of metal, it might be even cheaper to 3d print them… :-D)
4. Commercial little servo-bipeds seem to be too expensive… And I am not sure how programmable they are (if I can install my IMU on them and/or be able to re-program them (might take as much time as just to build mine own from scratch)
Next step is to finish frame for a leg, buy necessary components and start building inverted pendulum. Most likely actobotics + a bit of tetrix + couple of steppers with gears will be as expensive as 1-2 dynamixel servos…
Ah… and I have decided not to use rotary encoders but a 3600 degrees potentiometer, thus the system will always know which angle it it is at the moment.
to be continued…
this is what I am trying to
this is what I am trying to build with actobotics… I am not sure however if the limbs/joints will be strong enough and will not have any play… :-/
Though they claim they have sort of precision, not sure how actobotics precision will be…