>>9941 >Lyekka waifu? Okay. Reminder that we need a thread for all the freakish biology based plans, related >>7498, or the biological brains thread is being renamed and used for that. Don't get me wrong, I like that idea with the plant muscle. I guess it would be to weak though, not flexible and resilient enough, or would have other issues e.g. eating humans .

Hi guys I would try to explain a mechanism that I think would open the door to small form factor and cheap actuators. >1 Each degree of freedom of the robot is operated using a pulley system >2 The torque source of each system is powered by the same motor >3 There is a mechanism that sets the first pulley of each degree of freedom on Break or Coupled state The issue here is, how would you displace the the pulley, we would need a cheap small form factor of linear movement. I hope the image can describe my idea.

>>10001 Hello Anon, welcome. Call us 'Anon'. Contribute to our Embassy thread if you'd care to. I like the sound of your idea, but I'm not very qualified to comment on it. We have at least a couple of ME who are affiliated here. Be patient and maybe someone else can comment on your design idea. Again, welcome.

>>10001 That's a good question. I have similar thoughts on how to do things, reducing the number of motors by changing their place or by coupling pullleys. I don't really have a answer to your question, though. Where do you want to put that motor, or is it a more general question? I mean, winding the pulley up to some spool would be the obvious idea. If it would need to handle surprises and need to move the other way faster than it could, maybe adding a spring into the pulley (where it doesn't need to wind) might do the trick? Or the motor winding the pulley up could be connected to the rest of the body with a spring? We could also think of lock mechanisms for such springs, maybe by using some magnetic switches, solenoids or some another mechanism. I don't have any experience on how to use motors to wind up some pulley, so we might just use very small playtoy motors to try stuff out before getting bigger ones.

>>10121 The Mini-Cheetah servos on Aliexpress claim to have 6.9 Nm of continuous torque, whereas Dynamixel-P series has up to 44.7 Nm. If you were move the servos into the torso you still have to account for lifting the body's weight up when standing. If the upper body is 20kg you need at least 40 Nm in each hip to lift that alone. Distributing the work across different servos requires them to be at least 10 Nm. The MC servos would be sufficient though for a smaller or extremely lightweight robowaifu. The OpenTorque actuator has a tested torque of 25-28 Nm and costs around $150 to build but it's big and quite heavy at 1150g. Perhaps it could be remade for a different motor more suitable for a robowaifu? https://hackaday.io/project/159404-opentorque-actuator First-party test: https://www.youtube.com/watch?v=n_CiCqIRS2E Second-party test: https://www.youtube.com/watch?v=6lW2YGQQIQ4

>>10131 > If the upper body is 20kg you need at least 40 Nm in each hip to lift that alone. > Distributing the work across different servos requires them to be at least 10 Nm Could you explain how you arrived at these conclusions? And if you're going to break down the geometric physics math for us, if you would do you mind also adding this set Anon? This is all the same design experiment, simply testing the single variable of a 1kg mass: - How much accumulated torque (in Nm or w/e) would be requited in total to rotate an attached 1m long lever (say a 200g rod) 90 degrees, with a 1kg mass affixed to the distal end of the lever? - Conversely, how much force needed to repeat the exact same motion w/o the 1kg extra mass entirely? - Finally, with the 1kg mass affixed directly transversely with the rotation axis of the OpenTorque actuator. That is to say, the extra mass is affixed directly to the external side of the actuator shell, and rotates in lock step with it. TIA Anon, appreciated.

>>10135 Thanks, I didn't know about these huge differences.

anons , I am driving crazy trying to solve the problem of having many degrees of freedom using cheap high speed high torque small form factor actuators. And I have this idea, as you know solenoid air valves aren't small so what if we use this principle https://www.youtube.com/watch?v=MhVw-MHGv4s The clip pen mechanism could be use to reach low energy consumption for valves, we would only need an spike of energy to switch it. The air muscles could be done modifying this approach https://softroboticstoolkit.com/

Related to main topic: Hydraulic muscles by Automaton Robotics: >>10179 >>10150 >Clip power storage mechanism Interesting. Do you have an idea if we can get the plans from somewhere? Or if you could copy the design based on the video? Solenoids can be very small, but this might be good to store power from a motor to help it start moving some part faster. I mean by adding some startup power boost: >>9195. I'm concerned about the clicking, though.

>>10150 >cheap >high speed >high torque >small form factor You're kind of violating some engineering principles there. a) No one 'outsmarts' the laws of physics, and b) economics is a thing. Pick any two Anon.

>>9754 Thanks, Anon!

>>8988 So one of my main project goals is to make a horse leg for a pony waifu. There may be as many as 7 joints that can move in each leg, and I think that the mini cheetah motors are not suited for the torque required by looking at the later anon's posts. I could see making a bowden system or a chain or belt-driven system that could work reasonably well with dynamixel servos at the high end. What do you think is the best way to predict how much torque is required to move a thing with 4 legs?

>>10470 >What do you think is the best way to predict how much torque is required to move a thing with 4 legs? Unfortunately, I'm not a mechanical engineer (though we have a couple here who stop by occasionally)), so I can't give you the formulas, etc. for it. They are out there. Calculating lever-moments of torque, etc. One thing you can take for granted is that while the kinematics with be more involved with a quadruped, the actual design-work itself will be simpler. This is one reason that Boston Dynamics Spot is already on the market today. Personally, I simply rely on basic mechanical experience from working on cars & construction, etc., to give me a good instinct as to what will work and what will not. 'Keep things Lightweight & Strong' is almost always a good design decision Anon. As for my own approach to such things, the answer is quite a simple one: >test everything I'm much closer to a Michael Faraday than a James Clerk Maxwell. :^)

>>10475 Thanks anon. I'm graduating as a ME in a few months, and I figure I can read a few research papers on something like this done already, or skip to modeling a leg in a gait simulator to come up with the maximum torque/power. Also for the anons who are making hands/toes, it could be lighter/smaller to make a single motor with a clutch/transmission system to distribute power to whichever fingers instead of cramming 5 motors or more for each finger. The downside would be higher complexity.

>>10488 >Also for the anons who are making hands/toes, it could be lighter/smaller to make a single motor with a clutch/transmission system to distribute power to whichever fingers instead of cramming 5 motors or more for each finger. The downside would be higher complexity. Yes, all engineering endeavor is a constant balance of tradeoffs. Such is physics. I've thought numerous times about a central motive drive shaft in core locations, with radial clutches all lined up like ducks in a row along it. As each clutch engages, it would translate it's new force out to a linear actuator associated with it. My notion vaguely hearkens back to the days of yore when steam-powered factories were all the rage at the start of the industrial revolution. Stationary steam-engines typically drove several long driveshafts down the lengths of the factory, and radial clutch/pulley arrangements transmitted some of the power off through belts into the machines on the factory floor. Ofc, our needs are quite different, but the fundamental physical mechanisms (and even some design-work) can be quite similar. I expect you could even capitalize on such an arrangement for a real steam-punkish design motif for robowaifus!

I posted this once in the robowaifu thread on mpol.net: https://hackaday.io/project/175888-motorized-walking-atat-using-servos-and-arduino - maybe this helps a litte. It's about building a ATAT from Star Wars.

>>10492 Neat Anon, thanks! I'd say it's a good example of design approaches that could be useful to everyone when creating robowaifus. >other related links he gave https://www.thingiverse.com/thing:1083338 https://www.thingiverse.com/thing:4641541 https://www.thingiverse.com/thing:4651937

>>10509 Okay, but it doesn't show what this is about. Maybe this should be part of these crosslinks? If the posting isn't moved to the right thread. >Hazel actuators

>>10492 was meant for >>10470 I'm trying to find out how to calculate stuff like this myself right now. I got some bldc motors for experiments and try to find out how well I could use them (after buying them, lol). Look at this: https://maker.pro/custom/tutorial/motor-sizing-math This software here might also be good, but it requires Windows and registration: https://www.orientalmotor.com/technology/motor-sizing-calculations.html - here the video: https://youtu.be/9MG8-3QuSyQ I also calculated something in Python, but have to post that from my other computer. I divided the weight of an arm and the length through ten, so I've got ten parts, each closer to the motor, then used the formula and added the values up.

>>10539 I used this for torque: https://www.electrical4u.net/calculator/electric-motor-torque-calculation-formula-torque-calculator-online/ >For Calculating, Torque for DC motor T = V x I / (2 x pi x N(rpm) / 60) N(rpm) is the speed of the motor V => Input DC Voltage I => Input DC Current From Kv to rpm its Kv multiplied by voltage: https://drones.stackexchange.com/questions/49/what-does-kv-mean-on-motors Gear reduction calculation then seems to be simple, just multiplying by the reduction. I'm sure, like the other calculations here, this is not very precise. Good news for me, though. With a reduction of four or five in the ellbow I should get more than what I need, if I didn't make a mistake.

>>10540 I also really like Quora for stuff like this, if only their app would support dual screen... *whatever* https://www.quora.com/How-many-Newtons-does-it-take-to-lift-100g-of-mass?share=1 >A Newton is defined as 1kgm/s^2. Lifting is force against gravity; so using the formula F=ma use g in place of a. Also mentally convert 100g to .1kg. >F=mg = .1kg x 9.8m/s^2 = .98kgm/s^2 = .98N >Answer .98N

>>10541 My Python code for calculating it: newton = 9.8 arm_in_kg = 0.5 section_weight = arm_in_kg / 10 sections = [i / 100 for i in range(3,33,3)] # ten sections with different distance to the motor newton_meter_required = sum([section * section_weight * newton for section in sections]) newton_cm_required = newton_meter_required * 100 newton_cm_required >80.85 motor_in_newton_cm = 3.8 gear_reduction = 30 motor_in_newton_cm * gear_reduction >114.0

>>10490 I've completely abandoned hope in artificial muscle fibers that are viable for now. The best things available aren't purely electric powered or are a heat system. Those that are are made from polymers or flexible materials which can't take the stress while sized correctly to a normal muscle. Therefore the best way forward for a beginner like me is to make bots with hobby servos and then upgrade to low-mechanical impedance BLDC motors when I know what I'm doing, or skip this step and finish it. >TLDR: using whatever hobby servos I can afford until perfect electric muscle fibers (not you nitinol) are invented Anyone know of a gait simulator program and ways to make 3d models? I'll check the board rn.

>>10601 >Anyone know of a gait simulator program An anon mentioned cyberbotics.com/ (>>10531) >and ways to make 3d models? Blender (>>8107) . I'd suggest you have a look at Grant Abbitt's YT channel as a good place to find information. Be sure not to miss his 'YT channels you may be missing' video in his Blender Quick Tips' playlist. Also, Dev Enabled has good stuff for you probably, since this is kind of like the same needs of a traditional vidya in many ways.

>>10601 Nice looking paper Anon, thanks. Interested ME spin on the medical leg prosthetics domain.

>>10574 No comments. So I can assume, it's correct and a decent approach?

>>10608 I"m not qualified to analyze it personally myself Anon, so I kept my mouth shut. Your code seems to remind me a little of a process I've heard about called 'Finite Element Analysis'. I don't see anything wrong with your code, but to my (again, admittedly untrained) eye, I'm missing any calculation related to lever-action multipliers related to each element's distance from the base axis of rotation. Am I missing something here?

These are copied over from where I commented in the wrong place Gromment(Grommets are useful to stick things together) Found this paper. "Air/water interfacial assembled rubbery semiconducting nanofilm for fully rubbery integrated electronics" Quote,"...rubber-like stretchable semiconductor...Here, we report the scalable manufacturing of high-performance stretchable semiconducting nanofilms and the development of fully rubbery transistors, integrated electronics, and functional devices...their electrical performances were retained even when stretched by 50%. An elastic smart skin for multiplexed spatiotemporal mapping of physical pressing and a medical robotic hand equipped with rubbery multifunctional electronic skin was developed to show the applications of fully rubbery-integrated functional devices..." https://advances.sciencemag.org/content/6/38/eabb3656 I think something like this will eventually be seen as the most viable path. All these mechanical and electromechanical devices take too much machining and work to make all these little disparate parts fit together. This process I expect you could use films and silk screen the circuits like t-shirts. Means designs would be easier to replicate. Send pdf or CAD drawings, and print them out to screen on the material. There's also a wealth of machinery, design and materials already used for screen printing that could be adapted. Another Hydraulically amplified self-healing electrostatic actuators with muscle-like performance https://science.sciencemag.org/content/359/6371/61.full Notable capabilities "...tested the cycle life of a donut HASEL actuator used in Fig. 1E for more than 1 million cycles ..." "...The use of liquid dielectrics enables HASEL actuators to self-heal from dielectric breakdown..." Fluid could be heated to keep skin warm. "...large actuation response up to a frequency of 20 Hz..." Fast "...specific power during contraction of the two-unit actuator was 614 W/kg.." For humans,"...According to our data table the body uses 685 W to climb stairs..." so the power of these far out weighs human power. Waifu could take care of you when you got old and carry you around.

>>10620 Interesting, Grommet-Anon. I think this technology will turn out to be important to us here in the future. When we are trying to optimize our designs to increase modularity, reduce weight, and reduce manufacturing costs, then directly integrating electrical and electronics directly into flexible materials will certainly be a very valuable thing to be able to do. Do you know if there are planned approaches with this stuff that allows for simple and easy maintenance yet? Thanks!

>>10627 I'm not exactly sure what you mean by simple maintenance but I have commented on manufacture and modularity. I'll review and link some of it here as the comments were really in the wrong place. I first found this board from there and didn't realize the structure at the time. It's a bit of a dump but maybe having it all close together will help. The general idea is as above Dielectric Elastomers. They use electric fields to actuate. The good thing is they are made of silicon and other elastomers which of course you can get at any hardware store to experiment with. Here's some links on this stuff. Here's a link to a good book on these called "Dielectric Elastomers as Electromechanical Transducers: Fundamentals, Materials, Devices, Models and Applications of an Emerging Electroactive Polymer Technology" >>8502 Dielectric Elastomers can also be used as sensors A good page with links to 100 papers on Dielectric elastomers >>8505 A link to a SUPER video. Note that D.E. can be used for computer logic. This means that with a few lines we could control muscles and possibly feed back sensor(touch) data. Using serial lines vastly reducing cost. Just like SATA and USB drive interfaces are cheaper than the older parallel interfaces. Large numbers of wires are a pain in the ass, costly and troublesome. With the right logic you could move some logic into the actuators themselves. I think actual human muscles have some logic in them??? An example would be if the waifu hits something it would reduce it's force in the muscle. >>8519 >>8522 I describe above how you could make these. The idea is to roll out thin sheets of this stuff like a round homemade taco shell or tortilla. You lay out a layer. Silk screen on the control lines, logic and add parallel lines of strong fibers and power/control lines sticking far outside the ends of the layer, fishing line should do nicely, maybe or likely have several layers on top of each other for different functions, add one or more power/control lines to the logic then you roll the thing up from one end. What you have is something that looks very much like a muscle with a thick middle and skinny ends with the fishing line stuck out the ends that are tendons. Visualize each little muscle much smaller and shorter than the whole muscle but with long tendon and power/control lines. Say you are building a muscle, the one on your upper arm that you flex to show how strong you are. You would build this out of a bundle of maybe 30 of these small muscles. You tie them together in series for length and add them in parallel (each offset from each other to stack smoothly). All the fishing line tendons would wrap around anchor points in the bone Once you have a muscle bundled together in a group you could in turn pack it in some sort of cloth covering to keep them all together. Obviously you could remove these and replace individual muscles or the whole group. As for sizes you could maybe have 3 or 4 sizes. The smallest would be very small such that it would still take 20 or 30 of these to make a finger muscle. This gives you redundancy and having small muscles in a bundle let's them move around realistically like real muscle. Another book. "Muscles, Reflexes, and Locomotion" by Thomas A. McMahon >>8975 Some ideas. I also touch on skin. >>8311 I think the best way to make skin is to use that fabric called microfiber with a little bit of elastic fiber added in to make it stretch. It's super soft to the touch. It's used for towels and you could throw the whole waifu in the tub, scrub it off then drip dry. If it's smart enough it could do it itself. The skin could be one big piece from toes to head. The vagina, butt could be pockets that go into a cavity. To connect them you move aside the abdominal muscles pull into the body tight and attach to a holder. Possibly be held in place with Velcro??? Reach into the top of the head for the mouth and ears. Skull cap could be removable. Wig, skull cap to cover??? If you connect the vagina and butt with the mouth the robot could drink water and wash itself out. Don't forget the elastomer muscles in the sensual areas. They could vibrate and/or contract to do wonderful things. These parts could be removable. I look at all these mechanical things, electric motors, gear boxes, noisy hydraulics, pnumatics and I can't imagine you could ever get the cost down on these things. They're great but they produce a lot of noise, they take a vast amount of machining resources and if they are 3D printed most plastic is fairly weak. If you use electric servos or motors you're going to have to have some sort of transmission and all of them will make a huge racket. You will have a grinding, thumping washing machine for a mate. Hydraulics and pneumatics are very inefficient and make a lot of noise. I think the Dielectric Elastomers are the most fruitful path but...it's going to take a lot of experiments and work to get these to work. Motors, pnumatics, etc., well you can cobble something together right now that will work but I expect it will never reach a satisfactory conclusion and will always be expensive, noisy and inefficient. The D.E. made be a more difficult path but I could see it being extremely lifelike. One last thing about skin. I'm opposed to silicon. It's messy, tears easily, collects dirt and everyone will know it's a robot. Why not cover it in micro-fiber that's easy to keep clean and super smooth to the touch.

I've just come across this youtube channel, they seem to be making progress with a biometric hand using air muscles. They've even started work on some which are actuated by heating an engineered fluid with an electrical element to create pressure internally. https://www.youtube.com/channel/UCd0xLOw6No5IAsq3Y2-b0eA

>>10639 Thanks for the excellent post. Lots of consolidated information and ideas here. Much appreciated, Anon.

>>10641 Yeah, that's Automaton. I put links into the thread for humanoid robotics videos, a while ago: >>5136 >>10179 With a crosslink here >>10180 It's actually hydraulic, with water, not pneumatics (air muscles). I only realized that myself recently, before I thought these were pneumatics.

>>10640 He's fun to watch as well. I feel like we're looking in on a Slavic mad scientist at work.

>>10617 >missing any calculation related to lever-action multipliers related to each element's distance from the base axis of rotation There's none. There are two calculations, one for how much torque is necessary for the whole arm and the other for how much reduction attached to te motor would be necessary to have even more than that.

>>10639 Thanks. Do you plan to replicate some of the DIY muscles on YouTube? Maybe making your own videos to show it is possible for some enthusiasts? https://youtu.be/PDqmGHHKkWw[Embed] https://youtu.be/eUdcBjo19oU[Embed] https://youtu.be/hsd7_vQqt5w[Embed] https://youtu.be/uw8FLgiXsmk[Embed] https://youtu.be/PgNKeqOCOKE[Embed] I hope to get one of the research set of Artimus Robotics one day, when I have time for it: https://youtu.be/i1QmwOxGGFA[Embed] (old video) https://youtu.be/usvoiGBAflY[Embed] (newer) Problem is, they need very high voltage, which might be dangerous. They can be very strong, similar to a human muscle. Composite DEA made out of strain-stiffening elastomers and carbon nanotubes. https://youtu.be/LWDRlfruSBM[Embed] https://www.pnas.org/content/116/7/2476 (pdf related)

I maybe should have ordered an ESC with my motors or some mosfets and other parts. But I didn't bother to look, since I thought I already have so much stuff and it will be there. Also, I probably should have watched this video https://youtu.be/erppWLMzw8I before buying some ESC (electronic speed controller) for my bldc motors. The cheap ones cost 5€, but they seem to be simple to make, with an Arduino and some mosfets.

>>10730 I know Tamiya make some of the best R/C models and have been in the biz since 1946. Maybe get one of their ESCs?

Never heard of them and I'm not in US. I'm not sure if I need something special, or if it would even be better to be able to make ESC especially for my use cases. Since it sems to be doable, I'll probably at least gonna try. I ordered two cheap ones and some mosfets as well, so I can build my own.

>>10714 Do you plan to replicate some of the DIY muscles on YouTube? It won't be anytime soon I don't think but eventually. Still reading about them. I saw this video not sure if it's been linked yet. Harder to make probably but high power. https://www.youtube.com/watch?v=xMGXqT0LWUI One thing that I like about these is that book and video I linked where it says you can make logic with this stuff embedded into the actuator. That is fantastic. If you try to run wires and controls to every single muscle in a waifu it's going to be a mess. Lots of potential to have bad connections, trouble. If you could get one good connection, easier, and then send signals to each muscle it would dramatically cut down on cost and manufacturing complexity. Of course nothing is for free as the complexity of designing it goes up. That research set of Artimus Robotics I bet is expensive. I bet you would be better off buying a bunch of different materials and trying different stuff instead. Of course if you are well off it would be faster to buy the kit.

Related: Buying or building your own electronic speed controller for a BLDC motor >>10729

>>10738 That's neat stuff Anon, thanks.

>>10602 >>10606 >>10738 Thanks bros. >>10620 I agree that soft hydraulic/pneumatic muscles are likely the best solution in the future. However, the electric motor platform is far more developed at the present, and therefore is the best solution to make one now. The fabrication is indeed cost-intensive for current robotics, but very well defined to the point where just about anything you need can be made with great precision and repeatability. Remember that if you don't choose servos you are most likely choosing high voltage, loud pneumatic compressors (or figure out a phase change system idk), loud and very messy/heavy hydraulics, or heaven forbid one of those auful heat-activated wire actuators. Maybe the best thing to try is to use lower voltage in some kind of parallel/series array with narrow or short DEs so you don't have lethal voltage millimeters away from you while you actuate the muscles. I've thought about an idea where there is a low-displacement coiled actuator that functions like a spring, with two sides of opposite voltage for actuation. >>10639 >You will have a grinding, thumping washing machine for a mate What if I told you I have seen a lot of people that seem to prefer that kind of bot? ^:) Since you want to be satisfied with a realistic bot, why not cover the cost of your dream by building the dreams of others? It would be extremely experiential. To build a big bot. For you. Personally, I now have some little blue sg90s to learn basic electric motors, build a little bot, and to go down my own roadmap of ideas.

>>10842 Also speaking of >washing machines I found that there is a samsung washing machine stator that looks like it maybe could be hacked into an actuator, DC31-00074C. Maybe other washing machine motors and such appliance bldc motors are viable for a high-current hard robot leg similar to the mini-cheetah.

>>10854 >Maybe other washing machine motors and such appliance bldc motors are viable for a high-current hard robot leg similar to the mini-cheetah. I think you're right. My opinion is that as more and more men around the world -- impoverished men -- seek to devise robowaifus of their own, that we'll see a veritable explosion over the coming years of creative innovation at reusing 'junk' components for robowaifu creations. And fortunatly with video sharing both commonplace and easy (even if all you have is a goyphone) then we'll see more examples of it too. For example, there was a set of video linked here where a brown man (possibly Indian) created a small moe-bot class robot with mostly foamcore boards and hotglued parts. It's here somewhere, but my apologies I can't locate it.

>>10854 James Bruton uses windshild wiper motors. >>10815

#Noise related Noise level reference: https://youtu.be/C62-Yl_V4NI Digital vs analog servos: https://youtu.be/2TVmWbIM3IM #Other servo topics Low quality servos might have shifting arms, not going back to the correct position: https://youtu.be/b4jAGz3YBSY Arduino and Servos (Drone Bot Workshop): https://youtu.be/kUHmYKWwuWs Removing jitter with 6000uf, 16V capacitor: https://youtu.be/S7EVRDPQmfE Wiring to reduce electro magnetic emissions: https://youtu.be/vMdJdTcW5WQ

Oh, wow. Spherical gear for a ball and socket joint. This is something I thought about myself, but just didn't know how to do it and thought it might be impossible anyways. https://t.co/rxWgbMnKKB via @61laboratory

>>10951 Wow, that's pretty interesting. If we could manufacture/purchase lightweight ball and socket actuators for our robowaifu's shoulders & hips then that would be a dramatic simplification of many kinematic problems in robowaifus. Thanks Anon.