[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/diy/ - Do It Yourself


View post   

File: 24 KB, 638x344, gits.jpg [View same] [iqdb] [saucenao] [google]
1560888 No.1560888 [Reply] [Original]

Are there any soft robotics artificial muscles that aren't slow as fuck to actuate, that are attainable to somebody outside a university research lab?

Or are home robotics still restricted to motors and servos?

>> No.1560900

>attainable to somebody outside a university research lab
why don't you just go to a university research lab that makes artificial muscles
and pay them to make you one
it really isn't that hard, they're pretty fucking thirsty for money

>> No.1560927

>>1560900
It hadn't very much occurred to me to ask. But now thinking about it, I would prefer something I could do myself so that I could do any and all design and prototyping on my own timetable, not being reliant on a lab to have the time and energy to divert from their primary research goals.

I imagine I would also be paying for their time and expertise, instead of just the materials. My own time is worthless, on the other time. I'd just be spending it masturbating anyway.

>> No.1560967

>>1560888
This approach looks relatively easy to implement, might need high voltages.

https://chunglabblog.wordpress.com/2018/02/08/hydraulically-amplified-self-healing-electrostatic-actuators-with-muscle-like-performance/

With the right electronics it would be very energy efficient as well (use a DC/DC converter to charge discharge the electrodes).

>> No.1561012

>>1560927
>I would prefer something I could do myself
Feel like you're going to need more equipment and research, not to mention materials, than would be worthwhile for you to put in to this..

>> No.1561054

>>1560888
In short all robotic muscles suck. Soft actuators are a meme. But if you want to play around with soft stuff there are pneumatic muscles. They are hard to control and valving is more expensive than you'd think.
>>1560967
was going to recommend this.
https://www.youtube.com/watch?v=M4qcvTeN8k0
https://www.youtube.com/watch?v=-TKjJBZEZe4
Construction is pretty simple, you either use a heat sealed plastic bag or bag of silicone filled with oil with flexible electrodes on it. For the flexible electrodes they use gel and vacuum coated on aluminum. You need high voltage DC to drive them though, about 10 KV seems to work. If you just want to play around with them without controlling them a cheap ionizer module is fine for this. Please be aware that if you do this, that basically every ionizer is not isolate so don't even think about driving it with anything other than a battery. I've fried a couple of AC-DC converters this way. You can buy a more expensive off the shelf DC-DC converter that's isolated and has proportional voltage control, but those are expensive. The best thing to do would be to make your own controllable high voltage power supply, also with capacitive sensing so you can have some idea of position. We're still figuring out how to actually control them. They are also messy. The dielectric fluid gets everywhere.
>>very energy efficient
theory says you can't recover all the charge on the electrodes. I forget the exact number, but I think you can only recover 60% of the charge even using the best circuit possible. This is the paper on the efficiency limits:
https://www.spiedigitallibrary.org/conference-proceedings-of-spie/8687/86870E/Understanding-efficiency-limits-of-dielectric-elastomer-driver-circuitry/10.1117/12.2010336.short?SSO=1

>> No.1561059

>>1560888
No. You’ve touched upon one of the bottleneck technologies for realistic androids

>> No.1561073
File: 133 KB, 960x652, A6CFE019-9D56-4E03-8D99-DAE739D97B85.jpg [View same] [iqdb] [saucenao] [google]
1561073

>>1561059
Actually a little over a year ago there was an advancement in artifical muscles. Normally they require high voltages, but this one required low voltage.

I would spoon feed but if OP cares that much then they will find it.

>> No.1561078

>>1561073
that approach is hard to scale

>> No.1561096

>>1561054
This is very interesting stuff

>They are also messy. The dielectric fluid gets everywhere
Why is this the case? Does the pressure involved within the bag break the seals? If so, could a more flexible or durable material be used instead, or would that increased durability be a detriment to the effective range of motion?

Additionally, it looks like the current soft actuators act more as a single muscle fiber rather than a skeletal muscle. Are there any current leading ideas on how to chain a bunch of these bad boys today to work in tandem for larger range of motion or total force generation?

>> No.1561324

>>1560888
You can look into Nylon Fishing line muscles, with experimentation I was able to recreate them in my workshop. There are some instructable avaliable on the topic.

I've had success heating them by passing current through thin nickel wire wrapped around them.

>> No.1561331

>>1561096
>>why is this the case
it just is. Shit happens. Leaks can happen if you're not careful or they start arcing through. Sometimes you just spill oil while filling them.
>>chain a bunch
very carefully. You have to be careful when you make them otherwise you get arcing between electrodes and a nice burning smell

>> No.1561354

>>1560888
Why don't you show everyone you're the biggest brain in the room and invent some? Couldn't take too much effort, so long as you have PhD's in physics, chemistry, biology, and engineering.

>> No.1561371

>>1561331
>it just is. Shit happens. Leaks can happen if you're not careful or they start arcing through. Sometimes you just spill oil while filling them.
Why fill them with oil at all? It's necessary for the self healing stick, but ignoring that can't you just use demineralized water? Should significantly increase the force too for the same voltage due to dielectric constant.

>> No.1561400

>>1561073
This thing?
https://www.youtube.com/watch?v=EQLDyHGl68c
If "single wall carbon nanotubes" tells me anything, then you're not making anything big.
The video seems to prove it, too. If it was reasonable to make something large, they would have.

>> No.1561401

>>1560927
If it's not DARPA shit, then researchers are very much open to questions.
If you have a real and non-retarded question, then email them. Check your spam too.

>> No.1561490

>>1561054
>Construction is pretty simple, you either use a heat sealed plastic bag or bag of silicone filled with oil with flexible electrodes on it.

I'd try to use metallized Mylar/BOPET/BOPP, etch the parts of the aluminium you don't want off with ferric chloride. Attach a wire with some conductive glue, the oxide layer isn't going to mean dick to a kV.

>> No.1561492

>>1561401
Retarded Question: How do I know if my question is retarded?

>> No.1561539

>>1561371
force scales linearly with dielectric constant while quadratically with voltage. So you want to use as high a voltage as possible, so you need to use materials with as high a breakdown voltage as possible. So oil does the trick. You can try water though. I personally don't think water is a good idea though... The mess is the least of your worries, the difficult part is the drive circuitry. Driving multiple actuators will be fun. High voltage is hard to switch + converters can be sort of bulky cause they need transformers.
>>1561401
Even if it's DARPA shit they will still answer questions. DARPA funds top tier researchers, top tier researchers don't like being told they can't publish shit. So DARPA does everything they can to avoid having researcher work on stuff that would have to be restricted. The only thing researchers won't tell you about stuff that isn't published yet. Or stuff thats related to a startup
>>1561492
most of the time they'll answer anyway. You can also ask here.
>>1561400
CNTs aren't that exotic. The problem in this case is in order to make a big actuator you need a lot of tiny sheets glued together.

>> No.1561567

>>1561539
>converters can be sort of bulky cause they need transformers.
If you don't need a lot of power, the transformer for a say 12 kV power supply will be smaller than the rectifier.

>> No.1561619

>>1561567
transformers are bulkier than h-bridges. For muscles you need two muscles per 1 dof, which means two transformers. Even more muscles means more transformers, which can get fairly bulky. For robots, having a couple of watts or more would be ideal

>> No.1561950
File: 1.23 MB, 1279x853, Capture.png [View same] [iqdb] [saucenao] [google]
1561950

>>1561400
>>1561073
>spoon feed

Well I guess you people can't into google.
Pic is what it is.

>> No.1561952

>>1561950
https://www.youtube.com/watch?v=1J47difr3oo

>> No.1561999

>>1561950
that's shit. Any thermally based actuator is and will always be shit.

>> No.1562023

>>1561952
>x32 speed
Jesus christ

>> No.1562103

>>1562023
>>1561999
>that's shit.

If you paid even a little attention to the field you would see this as I do: a major advancement.

Comparing it and dismissing it to other thermal based artificial muscles is downright criminal.

First off no artificial muscle ever has gotten down to ridiculously low power levels this requires and that includes other thermal ones.

Second the weight to power ratio is just plain great.

Now if all you are worried about is speed, then I guess we are great because this solves the power problem and weight to power ratio!

>> No.1562105
File: 10 KB, 112x78, Capture.png [View same] [iqdb] [saucenao] [google]
1562105

This can be run on AA batteries!!!

>> No.1562109

>>1562103
it's too slow to build a fuckbot with tho, even if the fuckbot could hold me up

>> No.1562113

>>1560888
Another step towards the Kira korpi robot

>> No.1562132

>>1562113
>not Brian Boitano
Fag.

>> No.1562232

>>1562105
>>1562103
Thanks for the chuckle, I laughed pretty hard
>>ridiculously low power levels
It's 0.2% efficient. 8 watts is not a ridiculously low power level. HASELs have been demonstrated to have a 21% efficiency.
>>power to weight ratio
is god awful. They demonstrated a power density of 4 watts/kg. HASELs have demonstrated an average power of 358 W/kg
>>speed
Anon, I don't think you understand what efficiency is. Thermal actuators will always suck. For one carnot says that heat engine efficiency is pretty goddamn limited and two thermal mass means the response rate will be pretty slow.

>> No.1562238

>>1560888
OPs question is related to making an artificial muscular orifice for benetrations

>> No.1562249

>>1561952
>2025
>buy soft actuator from aliexpress
>3d print hand exoskelleton
>use next-gen arduino as controller
>build a functional hand for less than 1% what biomedical companies typically charge for such a device

>> No.1562290
File: 38 KB, 700x438, 25.jpg [View same] [iqdb] [saucenao] [google]
1562290

>>1562249
>Technological Singularity hasn't happened yet
>world might collapse before then
>no post-scaracity civilization
>no open-source big tiddy cyborg gf plans
>no nanobot assembly clouds

why even live

>> No.1562305
File: 30 KB, 644x554, thefutureisnow.jpg [View same] [iqdb] [saucenao] [google]
1562305

>>1562249
Already happening mate, people 3d print prosthetics that use wrist movement to close the hand or mechanical prosthetics that use an open source arm band that reads nerve signals through the skin.

https://www.youtube.com/watch?v=Cl8ijPGEKO8

https://www.youtube.com/watch?v=QOhHRzTTpeM

https://www.youtube.com/watch?v=KSP4o_WCqVs

https://youtu.be/vxcUknfPzl4?t=147

>> No.1562312

>>1562232
>power density of 4 watts/kg
> HASELs 358 W/kg

Are you supporting my argument or yours? By your numbers mine is 89.5 times more power efficient in lifting weight.

>> No.1562331

>>1562305
But all those use either servos or leadscrew linear actuators, they're pretty shit. No artificial muscle for arduino yet.

I have a 3D printer but it would be pretty bad trying to design a hand without one in first place. Maybe I'd need a pedal to use fusion 360.

>> No.1562352

>>1562312
That's the power to weight ratio not efficiency. 1 kg of VPS actuator can output 4 watts mechanical power, 1 kg of HASEL actuator can output 358 watts mechanical power. To supply this much mechanical power we must provide 358W/0.2=1704 watts for the HASEL actuator and 4W/0.002=2000 watts for the VPS actuator.
>>1562249
Of course if you want the same amount of control with a real hand you have to do surgery to wire up electrodes to your nerves for control and sensory feedback. Since nerves are tiny you need a home semiconductor fab to make the electrodes. Then there's the fun part where you need to have to connect the hand to the stump. Well I mean you can do all that at home, but...
>>1562305
Electrodes on the surface of the skin are very limited.
>>1562331
Honestly, there's no reason why robot hands have to use artificial muscles. Piezoelectric actuators have the potential to be quite force/torque and power dense, they're just really difficult to design.

>> No.1562361

>>1562352
You have confused me something awful. Why would you express mechanical energy in watts in this situation?

So now you are saying to lift one kilogram one needs 1700 watts and the other 2000 watts? In input power? Right?

>> No.1562380

>>1562361
What else are you going to measure power in? Horses?

>> No.1562384

>>1562380
Obviously newtons if you are lifting weight and watts for electrical energy.

>> No.1562387

>>1562361
>>1562380
>>1562384

God damnit you faggots, newtons for force, joules for energy and watts for power.

Also fuck off with your stupid heated soft polymer shit that takes 35 hours to actuate dumbass. >>1562103

>> No.1562388

>>1562384
Newtons is a measure of force, not energy.

>> No.1562393

>>1560888
just buy neodynium wire and silicone sleaves for said wire, they WILL wear out, but can be driven directly with DC currents.

Otherwise go the spider route and get a strong dc powered hydraulic pump with LOTS OF SWITCHES, lots of hydraulic tubing and anchor points with auto-off sensors for ruptures/leaks. That would be your cheapest possibility. There is a reason most animatronics use hydraulics to move their robots and such.

>> No.1562397

>>1562388
We are talking about the force to lift 1kg.

>> No.1562398

>>1562352
>Honestly, there's no reason why robot hands have to use artificial muscles
Reminds me of when 640k was more memory than anyone would ever need in a computer

>> No.1562403
File: 151 KB, 764x1280, F1.large.jpg [View same] [iqdb] [saucenao] [google]
1562403

Check out how many thousands of watts HASEL needs. I don't believe for a second it is any where close to the energy efficiency of the Columbia VPS.

>> No.1562413

>>1562361
>>1562384
Oh dear you really are quite hopeless. First let's go over what work is. Work in physics is defined as force times the distance over which it acts and has units of energy. So if you lift something that weighs 1 newton 1 meter up, then you will have performed 1 joule of work on that object. Power is the work/time. So if we lift 1 newton, 1 meter in one second we will outputting 1 watt of power. So our actuators apply a force over a distance in an amount of time, so that's power. Now then before we proceed further let us discuss the difference between weight and mass. So you weigh less on the moon than you do on earth because the gravity is different, but your mass does not change.
>>1562397
No we are not.
So in my post I mentioned power density, which is equivalent to power to weight ratio, but has much nicer units. In practice, engines, actuators, and other things that output power can't apply infinite power. We can try putting in more power to an electric motor, but at some point it's going to break. We can make a bigger electric motor, engine, or actuator, to get more power, but now it will have more mass and thus weight. In general, the power we get will be proportional to the mass of our engine or actuator. The number we multiply with the mass of the actuator or engine is called the power density. In general it is desirable for this to be as high as possible. So a regular car and a sports car have about the same mass, but the the sports car can apply more power. Because the sports car can apply more power it can accelerate faster than the regular car.
https://en.wikipedia.org/wiki/Power-to-weight_ratio

>> No.1562424

>>1562403
That is voltage not watts. Electrical power is voltage multiplied by current (amps). While the HASEL actuators use high voltage, the amperage can be quite small. After all to get the same power we can use low voltage and high amperage or high voltage and low amperage. 1 watt =2 volts* 0.5 amps= 20000 volts * 50 microamps.
>>1562398
No I'm simply stating that we don't need artificial muscle to make a hand. We don't need some soft contracting thing to make a hand, we could use something else entirely. We could use a sliders that that crawl up and down a rods in the arm to pull cables connected to the hand. We could put tiny rotating motors in each of the finger joints. The last is a bit absurd, as the motors would need to be ridiculously strong, this is not impossible though. We don't have to use soft contracting material to make a hand. In fact, soft materials tend to be less durable than hard materials. Right now the actuators in robot arms are good for 10 years of continuous operation, while many of these soft actuators are only capable of lasting months.

>> No.1562458

>>1562424
Muscles don't need a gear box and aren't as prone to break when there's suddenly a lot of external force applied.

I don't think the pneumatic/hydraulic muscles using a flexible tube and a woven shell have been mentioned yet. The the voltage controlled pressure regulators look a bit bulky and expensive though.

https://www.youtube.com/watch?v=AkaVsoEd6Vo

>> No.1562460

>>1562424
Something just seems shady. I'll have to look into it more. But if they are so good and throw in self healing, then why aren't they everywhere?

>> No.1562471

>>1560888
Here ja go faggot.

https://www.youtube.com/watch?v=P2qhQftLy3g

>> No.1562507

>>1562460
they just came out last year. Also high voltage drive circuitry isn't cheap. At least off the shelf, there's no reason why you couldn't build a little flyback circuit to do the voltage control for cheap.
>>1562458
The entirely hypothetical actuators I mentioned wouldn't either. Piezoelectric actuators can potentially provide high torque/force at low speed while providing high power. For dealing with impacts one can add a spring in between, or just measure force and move really fast. Yeah, pneumatics and hydraulics hide a lot of the costs. Proportional valves aren't cheap

>> No.1562556

>>1562507
The half bridge is a tricky thing to build regardless. Don't think there are even any 10 kV isolated drivers off the shelf.

I wonder if AC drive wouldn't be easier, use two transformers in anti-phase and drive the electrodes with them. At 100 kHz it should be way faster than the muscle can respond, so it will just see a relatively constant force.

>> No.1562765

>>1562556
Digikey sells them. They aren't cheap. Some chinese researchers made almost the same thing using a flyback circuit.
>>AC
doesn't sound like a good idea. I mean these things are capacitors and capacitors have impedance, so wouldn't power be dissipated? I think the flyback circuit I mentioned above was half wave though and it seemed to work fine.

>> No.1562935

>>1562424
>In fact, soft materials tend to be less durable than hard materials
Like silicone rubber vs glass

>> No.1562937
File: 255 KB, 1536x2048, 2470661238385946777.jpg [View same] [iqdb] [saucenao] [google]
1562937

>>1561324
I've got a bunch of this shit if anyone is interested, gotten duty cycle to around 30 seconds on thinner material. I think if you had nickel doped nylon 6, spun on the nano scale you could have something that's highly effective as a form of artificial muscle. Additionally recent research has found that you can 'train' these types of nylon actuators to behave like SMAs but they cost almost nothing.

>> No.1562963

>>1562937
>anyone is interested,
Can you start with how you made them and the required volts/amps?

>> No.1562983
File: 219 KB, 1536x2048, 5860125250114737707.jpg [View same] [iqdb] [saucenao] [google]
1562983

>>1562963
First I suspend nylon fishing line from a small 12v motor with a hook attachment a weight keeps tension on the line. Then I wet the nylon with some water and then I pin an equally long piece of nichrome wire at the top and bottom of the nylon. Then I spin the motor which causes the nichrome to wrap around the nylon, I visually watch it until the wire seems covered. Then I unpin the nichrome which stays via tension alone now. Finally I continue to spin the nylon until it doubles over on itself and they are done. I put two strands together to make a helix shape in the pictured example. Checkout this video for a non-nichrome version tutorial.
https://www.youtube.com/watch?v=1A2LUbJjDQ0

These have a duty cycle of about a minute, they take 16V at 4amp. Here are some pictures of producing hydrocordial muscle which though weaker are very fast.

>> No.1563007

>>1562937
>>1562983
What kind of contraction/expansion can you get on those bad boys? Do they move a good % distance?

>> No.1563022
File: 147 KB, 648x334, 3-Figure6-1.png [View same] [iqdb] [saucenao] [google]
1563022

>>1563007
With the 8in helix muscles in the original picture you can get about 17% contraction, I'm trying to get them to 20% through experimentation. With the hydrocordial you can get +30%, but I have not messed with them as much as I've started looking at combination muscles with planar motors like in this image. Additionally you can make hydrocordials that push instead of pull.

>> No.1563026

>>1563022
Shouldn't you be able to multiply distance using gears or a cam mechanism or something in conjunction with these muscles?

>> No.1563103

>>1562935
>Like silicone rubber vs glass
different strenghts so to speak
I'd like to see a pane of glass survive being folded in two, for instance.

>> No.1563256

>>1562765
Resistance burns power, an ideal capacitor can not.

A true flyback generator just generates DC, if you wanted to use that to drive the actuator you'd need to add a bleed resistor to turn it off again.

HV enthusiasts also often call a Royer oscillator flyback, because they use a flyback transformers with the rectifier diodes taken out or burned out. That generates AC though.

>> No.1563268

Does anyone know of any cheap torque sensors/encoders? Position control is ass for robotics.

>> No.1563307

>>1560888
I think I saw a video of some simple synthetic muscle experiments using twisted nylon strings and heat. It looked pretty neat and easy to make.

>> No.1563346

>>1563256
>>bleed resistors
are standard practice for HASELs and dielectric elastomers
>>1563103
>>1562935
all soft stuff we know of exhibits stress relaxation, meaning all soft actuators eventually get saggy.
>>1563268
A spring and some way of measuring distance the spring has been deformed. You need to learn a bit of control theory or its really hard to deal with the compliance. There are also resistive torque sensors for steering wheels, not sure how good they are.

>> No.1563855

>>1563022
If this isn't clear enough I'll make an info graphic with pictures or diagrams. Just let me know, if you look up material actuators I have a hacakday. Io project.

>>1563026
I have not considered this but it's a good idea.

>>1563307
>>1562937
See

>> No.1564888
File: 976 KB, 854x480, 1460671761471.webm [View same] [iqdb] [saucenao] [google]
1564888

>>1560888
Well it would be fairly easy to make a cable driven robot that could be actuated through a series of poses by a small handful of remote servos

>> No.1565638

>>1560888
As long as robotics are separated from biology we will never have what you are looking for. What will end up happening is cloning and gene editing for the sole purpose of limb harvesting or servitude. Obviously this is a moral issue for most and also why it would be quite some time before it would happen. The idea of having a full metal gf is a meme and always will be. Androids will end up being more like synths from fall out than anything