Electromagnet Actuator

Btw, please forgive that it looks ridiculous. I think it would work with EM Shielding and a different shape. Obviously, it would need something between the coils so they don't touch, as well.

>>1077377
What part of the sex robot is this supposed to be for?

>>1077387
There would be maybe 3 or so at each joint. Small ones for digits on the hands, and large ones for limbs. The spine will be straight for the first design.

It might not be able to supply enough force for a large body, but it would allow a small teddy-bear sized prototype to be cheap.

>>1077392
Wait, you're doing this for each joint? Yeah that's not gonna work senpai, use some kind of linear actuator. Or motorize the hinges. Or just get a gf.

>>1077394
Well, motorized hinges move slow.

We're working on an AI to go along with it so that it can walk, or at least do simple body actions.
The problem with manufactured actuators is that they are expensive, heavy, and slow. We need lightweight and responsive in order to avoid the jerky movements commonly seen in industrial robots.
It's gotta look good, bro.

>>1077398
Mechanically, it simply won't work. The way that this pivots is incorrect.

File: Electromagnet Actuator2.png (4KB, 294x212px) Image search: [Google]

4KB, 294x212px

>>1077402
Ok, I made a different design.
I drew it quickly so that I could get it up before the thread died.

The attached pic is sort of what it is. It's really just 2 separate "S" shaped bars with one coil for each side. The "S" bars move along rails so they can move up and down (not rotate) when that side is activated.

On the top/bottom is a bar that would effectively rotate a limb attached to that bar. I haven't thought of a way to make the bar, but I don't think it's an issue.

I'm mostly concerned with whether opposing electromagnets can apply a reasonable amount of force, or if they're just very inefficient.

I really want to make my own actuator so it doesn't cost much.

>>1077411
I'm trying to draw something that explains how this won't work
Stand by.
Bump.

>>1077411
>>1077417
Nvm too lazy. It won't work. Get a gf.

>>1077419
Ok. Well, I think it will work enough for something small.
I guess I'll think about the design for a day or so and come back with something better.
Thanks for the help.

>>1077377
To get a better understanding of what you're trying to achieve with your actuator, can you please give a list of user needs? Maybe there is some other methodology you haven't considered that can meet your need.

>>1077422
>Thanks for the help.
What help? he didn't help you at all.

>>1077424
Primarily cost.
The prototype robot will be something small, like a teddy bear. Maybe 1.5 feet (45cm) tall.
I guess large joints will need to support 30lb of pressure without changing position.

It really just needs to supply enough force to allow it to walk, but I don't know how to calculate that, and it depends on how heavy it is. I think it could be as light as 50lb with the correct materials.

>>1077411
>I'm mostly concerned with whether opposing electromagnets can apply a reasonable amount of force, or if they're just very inefficient.

Electromagnets can be extremely strong, but the inherent problem with limb joints is that they're at a mechanical disadvantage. They need to be able to produce 5-10x the expected force at the other end of the joint in order to work satisfactorily. This is why so many electronic actuators have built-in gearing systems; it's usually impractical to have a mechanism that can produce the necessary forces on its own, for various technical reasons.


That particular design would also grant only an extremely limited range of motion, and, contrary to your intent, the "s" bars do rotate slightly. Worse, stacking them causes the power demand for the joint to go up linearly, while not actually increasing the strength of the actuator.

Truth be told, I have no good recommendation for you. You're looking for the holy grail of cheap, simple, and powerful. If there were such a thing, it would already be used all over the place, because every other entity with even a passing interest in mechatronics is trying to figure one out. That's not to say it can't be done, but it IS to say that, if you're the one to do it, forget trying to make a sexbot out of it and just sell it off to someone else for a hundred million dollars.

File: pl903367-304h_hexsteel_304h_hex_mesh_304h_hexmesh_stainless_steel_hex_mesh_flexible_hex_metal.jpg (100KB, 600x520px) Image search: [Google]

100KB, 600x520px

>>1077431
Ok. That makes sense.

So, along the lines of a holy grail actuator for joints, the issue as I understand is that a lot of energy is wasted on bending the material of the limbs where they connect to the joint, so that the metal is bent instead of rotated.

I think the reason that an animal muscle is efficient is because the muscle pulls together while attached to the entire limb, and not just where the joint is. You could partially emulate that effect by having a gear system that pulls strings connected in multiple places down the limb so that the force is spread out more. That would require gears though and it would be complicated to build.

The idea I'm thinking of now is something that looks like a lot of cells. Basically, a 3D mesh with same-facing electromagnets between the joints. When the mesh is activated, it squishes together. To make it push apart I guess you could flip the current on every other magnet, but the downside is it wouldn't be able to apply force from both directions at once so as to hold something still.

The pic is basically what the mesh would look like, but made of cloth. Each joint would have a horizontally aligned electromagnet/coil.

Do you see any problems with this?

Electromagnets eat power, first of all. Second, how are you going to hold joint positions? Some kind of feedback? Coils have inductance so switching directions causes a voltage spike.

File: Electromagnet Actuator3.png (4KB, 133x278px) Image search: [Google]

4KB, 133x278px

>>1077450
I drew a picture of it.
The entire mesh is attached to the limb at certain points, ideally as a muscle would be to the bone.
When it pulls together, if the whole thing is glued to the limb it wouldn't do anything, but it would need to be connected at certain places. I don't know how yet. I'm thinking of thick rubber strings that connect the mesh to the limb at ~6 points, and a hard connection at the joint.

This is a weird design so I'm having trouble making it work, but I think something unusual is probably where a solution to this problem would lie.

>>1077461
I didn't know that. You could have a separate set of electromagnets set to the opposite direction for every other coil to reverse direction. Still, it would not hold a position, only move toward and past it.

I didn't know that electromagnets eat power. That might kill the whole design.

I'll think about a different design. It might take me a few days if I can't use electromagnets.

>>1077380
You'd get what, 45 degrees of movement from that design?
What joints are you planning on using it for?
Also fuckall holding power, and absolutely binary, either one side or the other, no inbetween.

>>1077398
Then work on wire muscles or something.
This is gonna be jerky as fuck.
>>1077450
>>1077462
I'm running this through my head and it's fucking brilliant.
It'd eat up power, but make the cells small enough and you've really got something.
A problem with what you've drawn is when it compresses the two rows would crash into eachother, so there needs to be a gap.

>>1077495
That's my brothers work, I told him he should reply to you. Do you have any ideas? This is going to be for the benefit of all, we're not selling it. Could you draw something up in Gimp real quick? I'd like to see what's on your mind.

>>1077490
Yeah, that was the first design. You would have to use multiple actuators attached to each other to get a full ~170-180 degrees rotation.
All joints should use the same actuator. The main idea is to reduce the cost of the thing. We're hoping to make it really cheap to build, with free open-source software.

>>1077495
Thanks! I thought it was cool myself, and you're right about the cell size. I don't know how small they should be, but it seems that the smaller and closer together each electromagnet is, the smaller EM field they make which I think is where most of the energy gets wasted.
I thought a little bit about the gap, but I actually haven't played with electromagnets before, so I don't know how they would act.

We'll be building a version of it probably a week or two from now when we have some extra money.
If we have something that looks neat we'll post a video.

>>1077450
The problem you're having with things being held still is why there are opposing muscles on the human body. Think bicep/tricep and what happens when they're flexing with the same forces. Humans aren't actually steady, we oscillate slightly when "holding steady", but if you can reduce the oscillations it'll be good enough. I like your idea of the electromagnetic mesh because you can simply turn on more or less for control over the force applied, though I don't know how much force you're going to get vs the weight of the materials.

>>1077506
About it oscillating, in software, I think you could make a simple math that tests for the recent angle change of the limb, and tries to apply the correct force to the pull and push muscle meshes. I think it could be solved through software or hardware, but in animals, it's probably mostly software.

That's a really good point about the two different muscle types. I think there are nearly infinite ways of doing this, but if this version is good enough I don't see a reason not to use it.

I think cloth is very lightweight. One problem I see is that the mesh will attempt to bulge away from the limb when it flexes, but I don't think it will be hard to solve. I have a solution for it now, but it's still vague.

>>1077514
You can deal with oscillations with analog damping circuits (which is how it's done with things like synchros). Software damping might be too slow, or overreactive to reduce the oscillations enough, but analog damping isn't all that accurate. Maybe a hybrid method would be the best solution.

>>1077501
>I actually haven't played with electromagnets before

You should probably rectify that before coming up with designs that don't work. Not that your designs aren't creative, but a mental model of whatever you're working with will help you eliminate dead ends and wasted time.

>>1077377
Hahahahaha no m8.
>>1077392
Wait what. You are building fingers?
>>1077398
Oh shit and now you want it to walk?
>>1077411
>> reasonable amount of force
so here's the deal, electromagnets only produce reasonable amounts of force when you apply unreasonable currents.

Unreasonable currents tend to turns electromagnets into slag. Now go do the math.


Basically you haven't a fucking clue what you are doing and are in way over your head. There is a reason we don't do this with robots.

>>1077561
Thanks for the recommendation. We'll probably end up using damping circuits. I'll have to do some research on them.

>>1077580
Well, the designs do work, I just thought they would be inefficient. Now, how well they work is up in the air, at least for me. The point is to make a prototype on the cheap. I am not a wealthy person, but hopefully we can design a diy actuator that works well enough to not be replaced by an expensive manufactured version.

>>1077604
I don't expect each one to supply much force. The image doesn't show how small they are, or how squished together their rest position is. There are going to be hundreds of these for each limb, and they'll be very tiny. It'll look like a bed sheet you wrap around the limb (sort of).

I figure a design like this is not used because it's inefficient, and there isn't a reason for a well-funded team to craft their own inefficient versions of something that could do the job with a little money.

Given, solenoids are pretty similar to this, and while this won't even be as efficient as that (maybe a third?), it should still work.
We'll post a video when we've made the first version.

>>1077656

i don't understand why you're using this overly complex method that might work rather than linear actuators (or even geared servos) that will.

>>1077664
Because they're slow.
I would use Pneumatic actuators, but what I saw was in the $150 range for just one. I'm sure you could find them cheaper, but it needs to move quickly to actually walk, even if the actuators are not strong enough to propel a heavy body.

We're trying to avoid the problem of a slow moving robot. If there was a linear actuator that was fast and inexpensive, we would probably go with that.

Another benefit of going this way is the robot could be made for $100 with the parts we're thinking of using.

>>1077675
>>1077656

Anon, are you building these things as tentacles? Because that seems like the only whay this would sort of work. At the very least this will be a cheap way to fail.

>> design a diy actuator that works well enough to not be replaced by an expensive manufactured version
Unlikely. Anon servos are fucking cheap even high quality servos are considered cheap. Brushless motors are cheap.

Anon roboticists aren't well funded, if these actuators worked at all they would use them. Even if they were inefficient.

An issue you will run into is torque density, electromagnetic actuators do not produce much torque relative to their weight which is why we have to gear them down a lot. If gets bigger it may not be able to lift its own weight.

Second of it is small why not just use regular servos.

>>1077675
So how are you going to do walking?

There's no good servo solution at the moment. To get high power to weight ratios you need an electric motor with 1000:1 gear reduction. Nasa recently developed a gear configuration called a gear bearing that reaches those in a simple device. Another recent option is magnetic reduction gearing, no one sells them so you have to make your own.

>>1077377
Dude, use hydraulics. Seriously.

>>1078074
I don't know what you mean by torque density. There isn't any torque, it's linear/vertical movement. It'll look like a muscle. It'll probably be a sheet of the muscle mesh that's wrapped into a tube, and attached to the limbs as a muscle would be to bone. That's what I'm thinking, anyway. I don't have the money to buy materials, yet. Won't till the end of this week, but it's possible I'll have to wait an additional 2 weeks.

I do have a different, similar design I'll try if this doesn't work, (though it should). It's basically a string with wire fed in the center, and coils going down the string, where each coil is about 2 coil's height apart. So when you activate it, it pulls together/shortens. Whatever design I use, it'll ultimately look like a muscle.

Honestly, I don't know if this will work, my estimate is if each coil can lift the weight of ~10 quarters, it'll be enough.

We have some software that we're testing for walking. It's basically an adjustment system that predicts its future motion, and modifies its expected path so as to keep the hips from going below a certain height. Other body movements will use the same system. I guess it's similar to reverse kinematics.


>>1078110
As far as I understood, electric motors/servos are too slow. I don't even know why they are, honestly. I'm not that informed. So I just abandoned whatever methods people were using because nothing looked effective. Not something within my price range, anyway.
I'll look into magnetic reduction gearing. Still hoping I won't have to use it.
If none of my simple designs work well enough we'll look into how magnetic fields flow, and hopefully make something that feeds efficiently off of it. If so, I'm imagining some kind of torus-shaped actuator, but we'll see if that's necessary. Maybe a torus with rings fitted around it that spin when electricity is passed through it? But I guess the issue is why servos are slow. I'll research it when I get back from work.

>>1078128
Whoops, I guess Torque Density is just Power/Weight ratio.

I guess you were just saying that it at least needs to be efficient enough to lift it's weight. Sorry, I overlooked that.
But yeah, I think it will be. So long as it doesn't catch fire I should just be able to increase the amperage and use a slightly thicker EM shield.

Assessing your knowledge and know how, you'll never get anything working. Can't you just fuck each other in the ass?

p.s. Use condoms

>>1078132
TORQUE DENSITY IS NOT POWER TO WEIGHT RATIO!

It is torque to weight ratio. Brushless motors have HUGE power to weight ratios, but very low torque to weight ratios as this power is in the form of low torque at high speed. Which is why you use a high reduction gearbox to get high torque. Servos aren't fast because the gearbox they use isn't efficient.

Ridiculous brushless motors can generate decent amounts of torque without a gearbox http://www.ghostrobotics.io/minitaur/

>>1078182
Holy shit, that's awesome.
Why are powered doors usually so slow, then?
Thanks, that should be all we need. They're cheap, too.

File: image002-1.gif (10KB, 279x278px) Image search: [Google]

10KB, 279x278px

>>1077462
You should definately build this, even if it is inefficient, because it is really cool.

This is inefficient because the distance between magnets is large. Magnetic force decreases quite a bit with distance so you want the distance between magnets or airgap to be as small as possible.

Might be worth using auxetics rather than hexagons for the muscles pic related
>>1078204
Electric doors are slow because they don't need to be fast.

Be careful with using such motors. Part of the reason it's able to do all this is because of some special driving circuitry.

Oh and the special motors it uses cost $279.90 each
http://www.rctigermotor.com/html/2013/Efficiency-Type_0928/94.html
The robot in the video costs $10,000 dollars

>>1078217
Thanks!
I did think that the distance between magnets was the main efficiency draw, but even still I'm not sure how efficient it could be if fully optimized. The design is simple and inexpensive, so even if it doesn't work at first we'll keep trying with it or something similar.

For alternative actuator types it would help if there were other ways to convert electricity to kinetic energy aside from electromagnets. Of course there is, but does current science know of any?

Thanks for the info on auxetics and the motor cost. I tried to find the motors they used, but I didn't have any luck. I looked up brushless motors and those were cheap, but that was as far as I got.

We'll have some money on Friday, so depending on whether we order online or via a local store we'll upload a video between Saturday and next Wednesday.


I do like the muscle system because it's cool, but hopefully it will work and we won't need to focus on iterating it like we will the accompanying software.

As far as the rest of the robot goes, once we have a working actuator design and a light body frame to attach it to, we'll begin testing our movement software on the robot. On the side we'll update a "Sexbot" wikia and develop prototype skin and fat materials. For fat we're thinking of using Gelatin mixed with Oil at different ratios for firmness, and maybe some sort of preservative to keep the gelatin from rotting if that's an issue. Gelatin is about $10/lb, so for a small bot we may need 5lb or so.

I'm imagining that all muscle operations can be controlled through a single Raspberry Pi, but we'll need around 200 input/outputs for a full model, so I'm not sure if one Raspberry Pi can handle all of that.

At any rate, there don't seem to be any major obstacles so far. It will take a few months to make something impressive, but our progress until then should still be interesting.
We'll post updates on the project through a wikia in the future.

>>1078952
That is a rather agressive time line for development. Even well funded teams have trouble going from CAD model to robot in 6 months. Getting said robot to actually do things takes longer.

You should get it working in simulation before you start building anything.

>> gelatin
Most certainly rots. It has been considered for making soft robots that biodegrade.

>>1078952
>For fat we're thinking of using Gelatin mixed with Oil at different ratios for firmness, and maybe some sort of preservative to keep the gelatin from rotting if that's an issue. Gelatin is about $10/lb, so for a small bot we may need 5lb or so.
Like the previous post said, gelatin is a bad idea. Go with solid silicone, silicone gel & saline, which are what's used in breast implants.

Try muscle wire. It contacts with electrical currents.

>>1077398
>motorized hinges move slow
wot?
https://www.youtube.com/watch?v=m_6cmu_2p1M

>>1079521

I just about guarantee that thing's direct-drive or nearly direct-drive (with just a single belt reduction). "Motorized hinges" explicitly implies...you know...electronically actuated hinges for opening doors or what have you. They're nowhere near that fast because they don't have to be.

>>1080033
Moar like a heavy torque stepper motor.

File: 1365369277976.gif (1MB, 320x240px) Image search: [Google]

1MB, 320x240px

>>1077377
>Hey /diy/,
>I and my brother are designing a sex robot