Neurobot-1000 3d model project

[opaque quartz]

so shrexy

yuh thats what im thinking

might make them worse than rubber for the muscle liner

that said if the placement is right, they won't push agianst each other much

hopefully

ideally

[floral quarry]

HOLD UP

what if we arrange them in such a manner that they're not too squished when heat is needed and we squish them when heat loss is needed

[opaque quartz]

genius

[floral quarry]

somebody add that to the document

and we layer the cooling alongside the foam in such a way so that when it squishes, the cooling sinks in rapidly

Using the cooling tubes

Lemme send a drawing of what I'm thinking

[opaque quartz]

however just realized that sometimes requirements vary: for long hold of muscle, which is what we need insulation for, it'd be different needs than a sudden movement and then relaxation. We would have to (somehow) design it so that when the muscle is extended continuosuly there is no foam compression, but when quickly extending, there is foam compression

we want fast heat loss during fast contraction + extension, but slow heat loss during a continuous contraction

idk how we'd do that mechanically

it's becasue we want the return stroke to relax as fast as possible, which needs fast heat dissipation, but the pulling stroke needs to have as little relaxation as possible, which needs low heat dissipation and high insulation

hence a continuous pull, for efficiency, would want to lose as little heat as possible

but a fast pull and release would want to have squished foam with fast heat disspation

---

Maybe a second nitanol sheathe on the outside of the insulative lining that squeezes the lining during the retraction stroke???

It wouldn't need any insulation because it would only be used once, and then allowed to slowly reset.

@floral quarry

ok now i really gotta sleep rip

😭

[floral quarry]

Lemme cook for two more minutes

[opaque quartz]

[floral quarry]

We use a plate of some sort with foam on top that supports the coolant tube

Plates slide outward to squish the foam and coolant tube falls in

Then we use another mechanism to slide it back up when needed

Shitty drawing, i hope you catch what I mean

In the air pockets during closed area, we arrange wires in a 2d spring coil to heat the air pocket

[opal bane]

Mate I can't even draw a straight line

[floral quarry]

Temp sensors are also ridiculous cheap, i can get one for less than half a cent

@opaque quartz thoughts before you sleep

[opaque quartz]

OH! so the tube is far away from the mechanism, but is pushing towards it. When a mechanism activates, the plate moves out of the way and the tube springs into place? I like it

[floral quarry]

Looked up prices, bulk rates are indeed cheap

[floral quarry]

Yassss

We move cooling in when needed, and move it out when not needed

And design the mechanism so it is run with the crappiest and least power consuming motors so we can use this for every muscle

Might need some tiny gearboxes to help multiply some torque in the process

[opaque quartz]

we can in fact combine our ideas as well; A nitinol sheathe, around the tube and muscle. the muscle is surrounded by air-filled foam. The plates are between the tube and the nitinol muscle rod.

When the current travels into the plates and the sheathe, the plates seperate whilst the sheathe contracts. This crushes the foam whilst moving the tube into place and compressing them all togehter

the plates can be nitinol with shape memory in the open position and closed position, same as sheath that surrounds it and the muscle

[floral quarry]

I got a great idea

Use magnetised plates that then repell

No motors

Scratch that

Nitinol is ferromagnetic at high temps

[opaque quartz]

we could use regular springs to reset the tube position

the force of the external mesh would counteract the spring

when they let up, the spring pushes the tube out of the way whilst the plates slide under it

pretty elegant system ngl

ty for the idea, it is really good for on demand liquid cooling

[opal bane]

Isn't that... Just regular magnets??

[floral quarry]

No it's paramagnetic at normal temps

[opal bane]

The hecks a paramagnet

[opaque quartz]

not affected by magnetic fields right?

[floral quarry]

Not magnetic

[floral quarry]

Yes

[opaque quartz]

biotech loves that aspect of it

[floral quarry]

I can't find what temprature it turns ferromagnetic at

[opaque quartz]

seriously ty for the help joker

imma head off now

i love working on this project with u guys

[opal bane]

Every time i think i mostly understand physics someone's like "yeah this material just isn't magnetic sometimes"

[floral quarry]

Shit is very weird even I gotta say

[opaque quartz]

physics be like

[opal bane]

What's next non neutonian gasses

[red oyster]

Cornstarch and water

Instead of air

For hydraulics

[opaque quartz]

oobleck

[floral quarry]

Honey

[opal bane]

I said gas not fluid

[red oyster]

Molasses

[floral quarry]

Farts

Holup that's newtonian

[red oyster]

Yeah wait

[ruby lagoon]

Apparently that's like non-ideal gasses, which occur more at high pressures since they can interact with other gas molecules more often

[opal bane]

Your telling me i can punch a cloud

And break my hand

[ruby lagoon]

At 100atm or so maybe

Don't touch the co2 lakes if you're deep see diving or something around 100atm though, likely fairly acidic

[opal bane]

Forbidden soup

[floral quarry]

@opaque quartz go to sleep or I'm going to throw another problem at you

Because I just came up with one

@red oyster if you were wondering about the cat ears

[floral quarry]

@ruby lagoon pls add them to the docs later

[opal bane]

Throw it at me and I'll see if my braincell can comprehend

[ruby lagoon]

which docs?

[floral quarry]

The neurobot one where everything is thrown in

[floral quarry]

Based on my argument above about the cooling tube, once we get the tubes in, how do we get them out

See reference to above conversation with matthew

[ruby lagoon]

Ah, I don't have access to that one

[floral quarry]

No way you solved it

Matthew solving problems as soon as I ask

[opal bane]

Matthew aren't you meant to be asleep

[floral quarry]

He's doing documenting

[opaque quartz]

had to write it down haha

yee

ok now i sleep

night

[floral quarry]

Count a million sheep and sleep nice

[opal bane]

Mods time him out for 8 hours

[ruby lagoon]

Should I make a trello or something btw? I could work on tasks for the spine/power bank in the design doc, matthew could work on the head tasks, etc.

[opal bane]

Trello would probably be more organized

Probably

[ruby lagoon]

Alright, I made a public trello and put what I'm doing on it: https://trello.com/b/5YM2i79k/neurobot-1000-tasks
I didn't know Trello supported gif backgrounds

[red oyster]

@short tiger

[short tiger]

Thanks

[red oyster]

@ruby lagoon

Not finalized, just a proof of concept

[ruby lagoon]

Pretty neat finger

[red oyster]

The mechanics

[opaque quartz]

Improvements to nitinol muscle design: Swap nitinol rod for square nitinol bar with hose-conforming conductive surface on cooled side. Swap polyurethane foam for carbon ceramic foam if carbon ceramic foam has more conductivity than it whilst compressed. Swap metal bars for plastic rollers with foam outer sleeve. Change nitinol wire mesh to electrically insulative, but thermally conductive tape membrane dilated by nitinol actuators at the sides. Add radiating flaps to the exterior of the muscle.

The muscle will no longer be heated on the outside, giving it better cooling properties when compressed. The plastic rollers will move to the side purely from the pressure of the dilating membrane, and slide back into place once the membrane expands and the springs push against the hose. The radiating flaps will improve heat transfer into the surrounding air when the carbon foam is not insulating the muscle.

@red oyster i recommend having a look at the progress log in pins if you havn't already

[fossil drum]

(Slightly) off topic but

I found this video of a robot hitting siu using ai that mimics human movement

Here it is

[opaque quartz]

nice

[fossil drum]

Idk how impressive (or relevant to the topic) this is but i found it pretty fun

Apparently its by nvidia

[opaque quartz]

i rate that robot as having some of the best humanoid locomotion of all commercial robots. I've seen it advertised a lot as well.

[opaque quartz]

I like the method used to have a central hinge

[fossil drum]

Well we are talking about nvidia and real madrid. Two pretty big names in their respective worlds. They have more then enough money to pull up with something like this

Although im still not sure if nvidia only developed the ai or did they put their hand in the robot development as well

[opaque quartz]

fairly sure the head is from unitree

so it might actually be unitree

seems like they went for the "motor in thigh" method of locomotion using pulleys

120nm torque which is similar to what we're aiming for

[fossil drum]

If we ever get founding we could go the chinese approach

Lol

[opaque quartz]

i don't like their approach

their robot waddles because it has very few degrees of freedom in the legs and has to spread its feet

we want our robot to have good hip strength sideways as well as forwards and backwards

i assume that this robot, like most commercial robots, for the sake of cost cutting only has strong motors in the forwards and backwards axis

[solid schooner]

Also it doesn't look like a cute anime girl, instant failure on that part

[queen field]

How's the status of the power generation? We still thinking stirling engines or have we moved onto the next hottest thing?

[fossil drum]

You misunderstood what i meant by chinese aproach 😭

[opaque quartz]

i think this vid is a clue, before it jumps, it rotates its legs inwards so that it's using the forwards/backwards motors

[opaque quartz]

for first verison we're going with purely batteries

[solid schooner]

Also these ropes holding it up make me doubt it can stand on both legs just fine completely, hmm

Like it's probably one ai trying things, whilst we're aiming for multiple ai, hmm
But it is a good source that our scope is possible

[fossil drum]

That also does not have anything to do with what i meant

[opaque quartz]

we can test pressurized stirling engines whilst we test nitinol muscles for late development versions

[fossil drum]

Tbf i think the lines are there to prevent something from happening rather then being there as stabilising force

[queen field]

You are free and welcome to state your meanings and intentions rather than leave us guessing.

[opaque quartz]

we'll wanna do that too to prevent falls once it goes over a certain angle

[fossil drum]

Anyways just to make it sure with everyone. By Chinese aproach i meant buying a robot, taking it apart studying how does everything work with each other and then make our own thing that has very close appearance of the original while we claim its fully original and that everybody else is delusional to think we stolen somebody’s else’s work

As you can see that was supposed to be a joke

[opaque quartz]

[queen field]

I think that's calles reverse engineering.

[fossil drum]

I forgor how that was called

So i just said the entire process

But generally chinese base their whole technology on reverse engineering

[opaque quartz]

https://mitechnews.com/wp-content/uploads/2023/09/FOURIER.gif

[queen field]

It's okay, we all forget things now and again.

[queen field]

The ant steps.

[fossil drum]

So yeah. Lets just reverse engineer neuro sama lol

[queen field]

The physical equivalent to CTRL+C CTRL+V.

[fossil drum]

Just a bit more work

And time

And money

Ngl
Tho i cant blame anyone for doing shortcuts like that

Blocking technological progress behind copyright seems stupid

[queen field]

Not in this day and age, no.

Always comes back around to cash money.

[fossil drum]

What exactly do
You mean by that?

[queen field]

Everything being prohibitively expensive to do new stuff, where you already need some form of established ground to stand on before potentially sinking lots of money into a passion project.

Taking a cheap shortcut and using someone else's work to get yourself started.

[fossil drum]

I can see when you are coming from

[opaque quartz]

all back motors fit within the back

that's 100% full range of motion in the back

[fossil drum]

What does full range mean?

180°?

360°

[opaque quartz]

matches human back. Not sure exactly how many degrees that is. I think 180

but includes curling

We'll have to see if it's better than using 4 motors at the base pulling wires up the spine. If not we can do that instead

i think there are some movements that need this arrangement though

i'll draw the pulleys and more wire later today

[opal bane]

I've seen ads for this thing

Though based on how it looks (doesn't look like its legs can move side to side, would perish on a non flat surface) our design is just... Better? How does this design look better than the peak of robotics engineering

[fossil drum]

could be that there is a reason for that

one we might not know yet

[opal bane]

The government shuts down all sufficiently advanced robotics projects and steals the plans for purposes

[opaque quartz]

i think it's due to cost savings in their robot. They need to make a profit off of every robot, and not as many robots will be sold as cars will, so the cost to manufacture needs to be a percentage of the sale price that's less than when selling a car. Cars are usually sold for twice the price of manufacture, so the price to manufacture a robot needs to be less than half the cost of the sale price of the robot. Motors are the most expensive parts of a robot so cost cutting will impact how many strong motors they fit onto their robot.

their asking price is something like 12000, and a motor that supplies 120nm can cost as much as 500 or 1000. So they probably don't want to add too many of them. I would be surprised if the cost to manufacture was more than £6000.

£6000 is 6-12 strong motors, so they probably went with only two large motors, and that saves them 2500-5000 to spend on the other parts of the robot

robot might cost anywhere from 2000-4000 to manufacture probably

this is all off the top of my head, the actual figures might be quite differrent

the robotics expertise probably went into designing a performant robot for low manufacture cost

however that means their robot doesn't perform at the peak of what current day technology can offer

the more axis of rotation the more motors or actuators you need to use, and the more energy supplied to the joints to handle differrent terrain, ballancing, etc etc, the more batteries you need

[ruby lagoon]

Alright, someone's gotta say it: the muscles aren't in the finger, they're in the forearm stretching from the elbow to the wrist. This system may move fingers but will be really weak.

Also, I feel like the trash talk about other robots/ robot companies is kinda premature if we don't have a working robot ourselves

[opal bane]

I'm just saying that it looks really simple

[ruby lagoon]

Fair. Simple is good design though: https://en.m.wikipedia.org/wiki/KISS_principle

[opaque quartz]

simple is also cost effective and cost effective is good

[night tartan]

As a programmer, I can guarantee that, what you think it's simple, may in fact, not be simple. That's were this principle fails.

[ruby lagoon]

That's true. "Just calculate only the synapses that matter" is the whole sparsity problem in AI. It's more a rule of thumb than a hard law.

[opal bane]

I still desperately need to use this acronym in a pun

[opaque quartz]

https://www.youtube.com/watch?v=mCshn2Rl5cs

we've been silly, we don't need an 'alternator' to capture energy from the stirling engine, just a brushless motor

also i've been looking into flywheels for energy storage

if we spin up other flywheels we can use it to adjust the angle of ballance of neuro, and the initial power can be applied via the brushless motor that acts as the power generator.

[opal bane]

Gyroscope and battery

[opaque quartz]

yup

to get a dense energy storage the flywheel material needs to be durable, it needs to be in a vacuum, and it needs magnetic levitation

it also presents problems with intentionally changing orientation. I don't know how spinning of other flywheels can affect the permanent flywheel.

would we need a temporary spring energy storage to slow down the flywheel, then spin it back up again?

so many questions...

[ruby lagoon]

Hmm.... putting this around the center of gravity, so in/around the midsection like an inner hula hoop, would make balance easier due to gyroscopic stabilization

[opaque quartz]

yup

either spin it down and back up again using harvested energy to turn off upright stabilization, or we can spin up a disk in the opposite direction

not sure how that affects power harvested from the flywheel and therefore its speed, and stabilization

it's an interesting dual purpose system

we have room to make the flywheel very tall and narrow

[opal bane]

So as it gets low on power she'll start moving less steadily due to losing the gyroscope

[solid schooner]

The perks of designing a two of a kind, state of the art, Neuro bot, is that since there is no profit gain, we can be more lenient in our expensive parts to build something that neuro can be proud of ^^

rip wallet

[opaque quartz]

ye

[opal bane]

We've just accidentally reinvented baymax

Low power drunk mode

[opaque quartz]

😂

hahaha

[ruby lagoon]

It may make control much easier, however, AI's advanced/fast enough today that I'm not sure the help from the gyroscope is needed

[opaque quartz]

yeah i agree

[solid schooner]

It may also hinder it when the ai tries to fight against the gyroscope and all that

[opaque quartz]

yeah and if we have a mechanical system that counteracts the spin, it has lag as it spins up or down

forcing it to only spin during a 'standing mode' for example

maybe a counterrotation using brushless motors would be fast enough though

how do we draw enough energy for that without running out of power in the battery though?

[opal bane]

Isn't movement with a gyroscope weird though

[opaque quartz]

maybe two storage flywheels that transfer energy between each other

[opaque quartz]

it's more stable and looks spooky

[ruby lagoon]

I feel like it'd be like strengthing your core when trying to walk across a tightrope. It would make you more stable, but also a bit more rigid

[floral quarry]

hi folks whats popping
couldn't do stuff today because of my weekly mania and then anhedonic episode

[ruby lagoon]

ah, sounds fun

[floral quarry]

-10/10 wouldn't wish upon my worst enemy

[opaque quartz]

https://www.youtube.com/watch?v=MjDVlPz5ZgI

[opaque quartz]

np bro

[ruby lagoon]

neurocube

[floral quarry]

i'll do some illustrations for hand because i looked at gd's design and it calls for too many motors

[opal bane]

In my experience with gyroscopes they don't want to move, which is good for stability but could hamper movement

[floral quarry]

nice design, needs some fixing though

[opaque quartz]

there's a vid i saw of a robot dog that could ballance on a wire using a gyro

[opaque quartz]

the motor number is equal to human muscle number

the goal is identical movement to human hand

[ruby lagoon]

Oh, right. I think the best system here is hydraulic/pneumatic. There are a ton of muscles in the forearm, and having 3d printable hydraulic sacs/tubes there makes making all those muscles so much easier.

[floral quarry]

we are broke so it means less motor

please remember about the black hole residing in our wallets

[ruby lagoon]

hydraulic would mean one motor and a bunch of solenoids... also idk why but homemade solenoids might be much cheaper than buying them?

[opaque quartz]

surely 1.7cm motor doesn't cost that much?

[ruby lagoon]

And 3d printing the solenoid valve systems is also much cheaper

[floral quarry]

currently i have no money to my name (in my defense, i am also a student)

[ruby lagoon]

The really small motors actually cost a bit more than the mass manufactured hobby motors

[opaque quartz]

ah

hmmmm

hydraulics for the fingers would require some big redesigns

might unironically be a usecase for the nitinol muscles even for v1

[ruby lagoon]

That would probably be much more expensive than the motor system tbh

[opaque quartz]

less efficient but it's less energy for the fingers since it's maximum of 10nm, but more like 1-2 nm usually

[opaque quartz]

idk we can get some strong forces out of thin wire

lets say for example a 0.1cm diameter, 5cm long wire per muscle

[ruby lagoon]

Maybe. You also gotta consider the size of the power electronics needed to convert the amps/volts to the right values needed for the nitinol and its desired actuation speed/strength

[opaque quartz]

ahh

fair, i'm not too knowledgeable about the electronics components

[ruby lagoon]

(The power electronics thing is also the main reason HASEL actuators aren't really viable)

[opaque quartz]

ok lets say we go hydraulics, would it be something like a water chamber in the forearm, with a heating element, and a valve system for each muscle?

brb gonna research hydraulics

[floral quarry]

holy shit how the fuck am i supposed to reduce motors the hand muscle setup is so complex

[ruby lagoon]

I don't think a heating element would really be needed. Maybe anti-freeze if Neuro is going to cold places

[floral quarry]

how do i collapse it in such a way we need bare minimum motors

[opaque quartz]

yeah there are like 3 layers of muscle on both sides, about 12-20 per forearm and more on the hand

what i drew is the minimum needed for human-like actuation

[opaque quartz]

one way might be to use a mechanical system that combines two of the muscles, lemme find their names real quick...

[ruby lagoon]

This is another reason for hydraulics. You can mostly just copy-paste from biology and it will likely work. Eye movement muscles for example:

[night tartan]

Wait, full on hydraulics? Doesn't that require a heavy motor to pump the hydraulic liquid?

[ruby lagoon]

Yea, one decent sized motor in the forearm around where the bones would be would probably do it

Like a 550 motor, maybe

[opaque quartz]

nvm not possible

brb making dinner

[night tartan]

okey. Just saying, there's pneumatics and electric linear actuators if hydraulics require a little too much

[ruby lagoon]

Pneumatics would also work. Might have more volume change, but less chance of leakage

Linear actuators would work too. Honestly might be cheaper with how absurdly expensive solenoids seem to be, unless we plan on making solenoids

[floral quarry]

how many motors do we need in one hand up till wrist

@opaque quartz @ruby lagoon

[ruby lagoon]

Apparently there are 20 muscles in the human forearm

[floral quarry]

i worked it to 8 for v1

[ruby lagoon]

Might be able to subtract 1 by removing the spinning motion and replacing it with 1 spinning motor

[floral quarry]

we may be able to get away with 8

[ruby lagoon]

with what range of motion?

[floral quarry]

full

im still stuck on how fingers would stay straight when bending

[ruby lagoon]

our muscles do this by pulling back on the finger bones while other muscles move them around, so it would require at least 2 actuators per finger

There are also the servos in the hand bones, which is shadow robotics design, which remove the need for muscles inside the hand itself

https://www.webmd.com/fitness-exercise/forearm-muscles-what-to-know

[floral quarry]

my idea is that for bending finger we borrow the flexor digitalis profundum and flexor digitalis superficialis, then pull on the palmar interossei, which pull on the flexor digitalis beyond a certain point

so palmar interossei > proximal phalanx which pulls on digitalis superficialis > middle phalanx which pulls on digitalis profundum > distal phalanx

for locking straight setup we power a tiny magnet which creates a lock for the digitalis profundum, making sure the middle and distal phalanx stay straight with the proximal phalanx

@ruby lagoon i will send a drawing attempting to explain what i just said

[ruby lagoon]

Heh, that might help. I haven't memorized all the muscle/bone names

[opal bane]

I think he's casting a spell

[floral quarry]

holy shit my laptop camera is not even 2 megapixels

and all i have right now is a keypad

this webcam is vga lmfaooooo

[ruby lagoon]

wow

[floral quarry]

sorry for terrible drawing, had to redraw with my trackpad

the parallel lines are rods for each phalanx

like the bones

this is phalanx naming setup

@ruby lagoon

each mucsle (rope) pulls on the other, and to prevent bending and keep straight, we lock one rope with a tiny improvised magnet lock (whenever required)

could get away for the locks with milliamps of current for each lock

to stretch hand out, we similarly do muscle for pairs pulling on finger, such as one rope - pulley pair for small and thumb finger, one pair for ring and index finger

wrist is - well, already there

@ruby lagoon please give thoughts

[ruby lagoon]

I'm not sure the magnets will be necessary, and if they are necessary small magnets are fairly weak and may be overridden by accident (as well as locking mechanisms of that size, depending on material).
I think if we set a max bend, then just pulling back on the finger a bit will keep it straight. See the offset here, and see the sharp angles added that would stop any further pulling back:

[floral quarry]

nah we have to just lock the rope in place, small magnet is enough
also, what if you have to keep distal and middle phalanx straight but bend only proximal phalanx

[ruby lagoon]

Oh! You're adding a braking force against the tendons? That's interesting.
Solenoids are expensive for some reason though, so that might mean winding a bunch of electromagnets. Also I still think it could get overridden since the rope is pulled with a large amount of force
I think bending the proximal phalinx only is the default option actually (well, for activating both opposing muscles). That means all the phalanxes are locked together, while the proximal moves with respect to the knuckle. I can also lower my distal/middle phalanxes while keeping the proximal back, which seems to imply two muscles pulling each finger back and at least one pulling it towards the palm, or at least 15 total muscles for that level of control, though they can work together.

[floral quarry]

we don't expect so damn much of torque for a finger rope, not so much earlier on that it snaps clear of the brake we add

yes, we will have to mcgyver some electromagnets, but nothing too harsh

the thing is I suggested this system coiled on a pulley (what was the correct phrase again) to reduce motors, as hand strength emerges mainly from upper parts and then arms, as fingers (for v1) don't require that much torque

and as we can control which phalanx to lock and which to not with just two wires and a few milliamps of current, this sort of makes sense

the more than 8-9 motor system also makes sense, but this was suggested by me purely out of a need for as less motors as possible to build an mvp

@ruby lagoon can we add this hack to save cost

[ruby lagoon]

Ah, that's another problem: due to economics of scale, motors seem to be cheaper than easier to make electromagnets, or about the same price

Though, some of thse might be good (the 0.01$-0.50$ especially)... https://www.alibaba.com/showroom/tiny-electromagnet.html

[floral quarry]

the motors these guys were suggesting were 24 dollars per motor

i was not amused by that

one tiny lock would take 2 dollars to cobble

[ruby lagoon]

Yeah, I'd just use 50 cent motors like these tbh, and use 3d printed gear boxes: https://www.alibaba.com/product-detail/DC-Motor-3V-12V-12000RPM-High_1601196930571.html

[floral quarry]

now you talk like a true engineer

[ruby lagoon]

The costs must be cut

[floral quarry]

then i proudly say my design sucks even for v1 if you use cheap shitty motors as much as possible

[ruby lagoon]

Actually, there is the noise problem though. The cheap motors make a noise when they spin at high speeds, while bldc motors that don't are around 6$ instead of 2

Cheap motors = neuro has noisy arms / is noisy when moving, and from experience with quadcopters, noise can get really annoying really quickly

[floral quarry]

SCHIEßE i forgot about that

[floral quarry]

damn even the cheapest bldc costs $3.50

what is this world

[ruby lagoon]

Well, if we need 1000+ motors... https://www.alibaba.com/product-detail/1503-1503A-brushless-DC-motor-for_1600946286773.html?

[floral quarry]

who's going to paay 850

[night tartan]

1000 motors in one buy

[floral quarry]

mfw money is harassing me day and night

[ruby lagoon]

This one would be 100: https://www.alibaba.com/product-detail/Low-priced-1503-1503A-Brushless-DC_1600963508229.html

There are others that claim bldc/brushless, but that might be a mistranslation or false advertising, since the pictures look like normal brushed motors and don't have the brushless designs

[floral quarry]

that's still.. doable

[ruby lagoon]

Honestly, it could be worth it to message the different companies asking for prototype motors for a smaller cost. Chinese companies usually respond to emails (might be because we'd be one of the few people emailing in English so we'd stick out in an inbox)

[floral quarry]

i guess

however, you guys need to do it, because alibaba is banned in my country lmao

[ruby lagoon]

ah, darn

[red oyster]

I'm awake, what did I miss

[floral quarry]

i could however ask my dad for contacts, he works corporate banking

[ruby lagoon]

(I found some bldc motor drivers on mouser too, and it doesn't seem to be much of an additional expense for the different motors. I think some were also around 50 cents)

[night tartan]

We're searching cheap enough brushless dc motors for the fingers, i think

[ruby lagoon]

Yea, we're trying to cut costs for all the different actuation in the forearm

[floral quarry]

we tried improving your finger design because money

aka apply hacks

[red oyster]

Fair enough

My design was just a concept

To simplify things

[floral quarry]

we're doing this only for v1, because no money

[floral quarry]

your design is nice, we might use it for v2

(v2 assumes we recieve funding)

ight, its past midnight, im sleeping

[red oyster]

The motors were cheap

[ruby lagoon]

night

[red oyster]

https://www.amazon.com/Abovehill-Stepper-2-Phase-4-Wire-Connection/dp/B08346RFVZ/ref=asc_df_B08346RFVZ?mcid=c4196c7ad5e838de8aa944d80f0d667c&hvocijid=17503664918740746688-B08346RFVZ-&hvexpln=73&tag=hyprod-20&linkCode=df0&hvadid=721245378154&hvpos=&hvnetw=g&hvrand=17503664918740746688&hvpone=&hvptwo=&hvqmt=&hvdev=m&hvdvcmdl=&hvlocint=&hvlocphy=9030088&hvtargid=pla-2281435178098&psc=1

[night tartan]

They're so tiny

[red oyster]

They need to fit in the finger lol

[night tartan]

Yeah, i know. I just find them cute

[red oyster]

They are cute

[floral quarry]

it doesn't list the price for me

[red oyster]

$11 for 20 motors

[floral quarry]

hmmmmmmmmmmmmmmmmmmmmmmmmmm

how much torque can we get from one

[ruby lagoon]

I mean, in finger motors are going to be functional, just really weak... or really fast because motors are fast
However, moving those steppers into the arm and gearing them down, joining them for more torque, or adding 3d printed linear actuator screws onto them could work...

[floral quarry]

5 gf.cm torque

4 * 10^-4 nm

one finger produces 1-2 nm

[ruby lagoon]

What's the rpm?

[red oyster]

They're stepper motors

[ruby lagoon]

Yeah. Max rpm

[red oyster]

I'm looking

[ruby lagoon]

1step/pulse * 800pulse/sec * 18 deg/step * 1rev/360deg * 60s/min ~= 2400 rpm

[floral quarry]

what

[ruby lagoon]

Heh, probably another mistranslation

[night tartan]

Slower than a turtle

[ruby lagoon]

datasheet: https://woostore.temperosystems.com.au/wp-content/uploads/2023/12/PM08-2-8mm-Micro-Stepper-Motor-datasheet.pdf

[floral quarry]

i think thats the same website it is mistranslated on

[ruby lagoon]

I think a 1:20 gear reduction could work, maybe even 1:40. 40 would only give us 0.016 nm though, and finger up/down motion would be about 0.5 seconds

[red oyster]

1:30?

Cut it down the middle

[floral quarry]

crappy torque though, ngl

ok i need to really sleep

[red oyster]

Gn

[ruby lagoon]

night

(hmm... double checked with pinch force and finger length, and 1-6 newton meters is actually about right... only need 100x that motor)

[red oyster]

Okay, well my other design was way more complicated...

Imma work on finger v2 today

Or I guess...

Hand v1

[ruby lagoon]

Hmm... longer motors in the forarm could work with 10:1 or 100:1 gear reductions. Would be more expensive though...

[red oyster]

I have an idea, it's just very complicated

But much cheaper than what you're thinking

[ruby lagoon]

sounds interesting

(btw, BLDC apparently has the best thrust to weight ratio by far if we can get them)

[red oyster]

we would need an encoder

[opaque quartz]

Four actuators pull on all five fingers; The wire pulled splits into four parts.

The other motors all have important roles for fine dexterity that are multi-purpose, therefore cannot be replaced by single rigid motors.

It's easy to think "Oh we can replace x muscle attachment because we don't need it to perform action x, y, x, however there is always a hidden purpose."

For most joints, for example shoulders that can shrug, or ankle which can rotate or pull on parts of the foot, we can abstract those motions or lose range of motion without massively impacting fine dexterity.

However for the hands, the same can't be said if we want human-equivalent dexterity.

Therefore we need to replicate every muscle, and do so with at least 10nm force per muscle

[ruby lagoon]

idk about 10nm of force per muscle. nm is torque
Otherwise yea

[red oyster]

for v2 yes, v1 doesn't have that level of funding

[opaque quartz]

1000 for arms is worth cheaping out in other aspects of the design. However if there are ways of replicating the hand that are cheaper, we take those

for example hydraulics or pneumatics could be a viable option since valves are quite cheap

[ruby lagoon]

I think as long as we have one finger that works, we can say the rest would work and probably be right, and that would probably be enough for funding if we want that

[ruby lagoon]

oh yeah, apparently solenoids are expensive for some reason compared to some motors, so idk about 3d printed hydraulics/pneumatics vs 3d printed linear actuators/SEAs

[opaque quartz]

yes so long as we can demonstrate full range of motion or potential for such, i think missing out on hands is a good option to keep the costs down until they can be funded

[opaque quartz]

ah

[ruby lagoon]

I mean, the wire for them is cheap, but making 50 solenoids would be kinda tedious at best...

[red oyster]

I'd be down to wind the wire

[opaque quartz]

@ruby lagoon @red oyster @floral quarry

[ruby lagoon]

biology is complicated...

[opaque quartz]

this is only one side

[ruby lagoon]

yep

[opaque quartz]

7 muscles total on this side, probably similar amount other side, then there is like 2-3 hand muscles, that's the close to the 20 figure you mentioned

for both arms that's 40 solenoids or brushless motors

i sized these to the powerful for its size 1.7cm motors i found, which output 0.1 nm, that can be reduced down to 10nm with 100x reduction

depending on their cost, those could be a lot though

there isn't pricing on their website, they want us to request a quote via email

[ruby lagoon]

Yeah, if we need larger motors, it might be worth it to buy 1-3$ solenoids and have one large brusheless motor around where the bone would be working as the pump for the pneumatic/hydraulic system

[opaque quartz]

problem then is space occupied within the hand

i also came up with a leverage system for these motors that'd likely fit, increasing strength for space

[ruby lagoon]

tbh I still like Shadow Robotics solution where the metacarpal bones are replaced with servo motors that do the finger wag motion

[opaque quartz]

we have to fit 20x 1.7cm x 1.7cm x 1.7cm square volumes with gear reduction too

or 20x whatever actuation method we choose

into a space that's at most, 5cm x 5cm x 10cm cylindrical volume

[opaque quartz]

i hate that solution, you CANNOT get enough torque in that size of package

[ruby lagoon]

just like tetris

[opaque quartz]

hand closes with some extreme forces

[ruby lagoon]

For the finger wag motion? The servos would be much larger than the actual motion

[opaque quartz]

think about it, rock climbers can hang from one finger

[ruby lagoon]

Ah, I wasn't talking about that motion

[opaque quartz]

sideways?

yeah we can do that

[ruby lagoon]

yea

[opaque quartz]

i support that haha

[ruby lagoon]

yea, these muscles:

[ruby lagoon]

heh, almost got whiplash from that change in reaction

[red oyster]

Why can't the human body be less refined 😭

[opaque quartz]

how do we get a pneumatic chamber with 20 connector ports for hoses? LOL

[red oyster]

we don't

[opaque quartz]

are there splitters?

[ruby lagoon]

yea... 3d printing is probably required for penumatics/hydraulics. And we might need to re-heat it or use SLA to ensure its airtight or waterproof

both the chamber and the 'muscles' themselves. Maybe the pump too depending on what's available

[opaque quartz]

everything i've seen about 3d printing tells me it has terrible lifetime

[red oyster]

it does

[ruby lagoon]

depends on the material. PLA is pretty bad for everything except prototyping. I think ABS is actually used in some products, and so is polycarbonate or other filaments. SLA can be similar to epoxy which is pretty good. Also, 3d printing can do some mold-making, though those molds won't last as long

[opaque quartz]

yeesh

[red oyster]

petg

[floral quarry]

[red oyster]

go to bed

[ruby lagoon]

o yea, might be good for water related stuff

[floral quarry]

woke up for water, saw something terrible

immediately returning to sleep

[opaque quartz]

im scarred for life

[floral quarry]

life tip: do not watch vdeos of heart surgery before sleeping

[opaque quartz]

bruh

rip

[floral quarry]

gn folks

[opaque quartz]

night

ok what about 10 per hand, then we use valve actuators for muscles, and two pressure chambers (1 per arm)

Might cost the same as 40 motors and gearboxes though

if the motors are £20 then that'd be £800

[ruby lagoon]

Huh. Hose connectors are more expensive than I thought

[opaque quartz]

for gearboxes that's even more

this is a random hopeful guess, they could be even more costly

1000 pounds for hands alone is too much

pneumatics, despite the price of splitters, compressors, and valves, might ultimately cost less

[ruby lagoon]

hard to say tbh

[red oyster]

3d model of hand

[opaque quartz]

we could make our own air compressor using a screw maybe

[opaque quartz]

ty

😭

[red oyster]

idk if it's to scale yet, still checking

[last epoch]

a motor is simply a backwards generator, so this checks out. Ive seen good results using brushless motors to produce power before

[red oyster]

Doesn't seem to scale, but it's a good reference anyway

[opaque quartz]

we can make it to scale

i appreciate the send

[red oyster]

Solidworks can't, so someone else will have to do it

[opaque quartz]

😭 😭 😭 😭 😭 😭 😭 😭

[last epoch]

What flow rates are we designing for? Small solenoid air valves can be really cheap (though unfortunately binary)

https://www.mcmaster.com/products/pneumatic-valves/actuation~motor/actuation~electric-solenoid/compact-solenoid-flow-adjustment-valves/

^ These look promising to control pneumatic actuation per finger

[ruby lagoon]

I think we all have different ideas of 'cheap'

I was hoping for 0.5 -5$ total per actuator

[opaque quartz]

ty i had no idea where to find small ones

[last epoch]

thats a tough price to beat

[ruby lagoon]

Yep, but since humanoid robots have so many actuators, you could multiply that by around 100. So 100$ actuators means ~10k$ just for all the actuators

[opaque quartz]

for v1 we can settle for the 40 in the arms and the 20-odd muscles in the face

the rest can be large brushless servomotors

costly but not in terms of 'hundreds of small motors' purchase

[ruby lagoon]

O yea, and solenoids for the face might be best. Apparently electromagnets aren't too expensive on alibaba, and the face muscles shouldn't need to be super strong

[opaque quartz]

yeah they can be very weak

irc facial expression muscles don't contract enough to visibly flex, because they're only moving skin

they don't insert on a bone and have almost no resistance or mass to move

the same will be true of our robot

[last epoch]

checks out

[opaque quartz]

[last epoch]

Though theyre cheap, unfortunately you wont be able to get much flow control with solenoid valves

they tend to be pretty binary

[opaque quartz]

if there are splitters that small, we can split two into one, and choose how many are open to double or halve force applied

[last epoch]

thats a good point

You would need a crazy looking manifold then

[opaque quartz]

surely if they're that tiny they must be prone to leaks?

not much material there

[last epoch]

the only way to find out is to test them when it comes to products like these without spec sheets

[opaque quartz]

aye

[last epoch]

tbh you could simply buy the cheap solenoids in bulk and 3D print your own valves

but Im not sure how much thatd cost

doesnt look like there'd be that much to print to begin with

Solenoid retracts the pin within that elbow to allow flow, and whether or not it leaks depends on the fitting

I imagine you couldnt have too much (or any?) interference fit or else the solenoid gets stuck

[red oyster]

wip

[opaque quartz]

nice

[last epoch]

I remember researching suppliers for micro solenoids for a project in the past and a provider I found sold actuators for 6 cents in bulk, Im trying to find it again

[opaque quartz]

So it turns out both men and women have these donut shapes on their mouths.

we're gonna want to replicate this visually to have a believable face

most robots miss it out entirely

[opaque quartz]

gl man

and ty

[red oyster]

1:125 gear reduction by alternating planetaries

[ruby lagoon]

PWM control should work well enough. You should be able to position the solenoids instead of full extending/retracting them, and I think it would just need a half-bridge instead of the full bridge dc motors would need for backwards/forwards control

[last epoch]

Thats a great idea, but my concerns are with continous operation of solenoids, don't the coil windings build up lots of heat without cooldown time? (though maybe its not a problem with solenoids this small)

[ruby lagoon]

Huh... maybe air/water seal some solenoid/electromagnet system so it retracts while maintaining constant volume, similar to biology. Should work the same for the muscles around the eyes I think

[ruby lagoon]

The windings can build up heat, but I think it's usually from going back and forth quickly rather than holding position, whether that position is half-way or fully retracted (T=L/R)

[ruby lagoon]

Huh... like the spur gears that are different on top/bottom, but planetaries... have you tested it? I mean, do the gears contact without intersecting & stuff in cad?

[red oyster]

Mathematically, yes

technically no

Although, I already own a nema stepper motor

[opaque quartz]

It appears to be a result of protuberance of the muscle inserts combined with skin and fat

looks intimidating but is probably easy to replicate mechanically

[ruby lagoon]

I think 3d printing some TPU and epoxying some electromagnets to it could work...

The softest TPU possible though (3d printing silicone is possible, but requires special printers), or you might need to add ridges so it can deform better

[opaque quartz]

if we used this could we place an LED power symbol in center? lol

[opaque quartz]

aye perhaps

we'll want to look at the properties of skin soon so that we could find a similar material

if not we'd use flat plating of equal dimensions and positioning

if a rigid structure is moved to seemingly 'deform' like a soft structure, it tricks the brain into thinking it's soft

[ruby lagoon]

o yea! That reminded me of something later for custom camers: https://fxtwitter.com/oldestasian/status/1886643136631386541

[ruby lagoon]

I feel like that's much harder though. But it would be impressive

[opaque quartz]

Yup. The easier it is to do, the more gaps and therefore the more freaky it is. A complient solid structure would need some form of suspension from actuators and might need to deform like origami.

In the meantime:
https://onlinelibrary.wiley.com/doi/full/10.1111/srt.12235?msockid=38279037f4f26e9928ae8256f5d56fd1

synthetic skin research paper

[opaque quartz]

epic

i've heard that using a lens as a master mold can work too

[ruby lagoon]

True, but this allows for more custom lenses, as long as they can be spin coated I think

[opaque quartz]

nice

[ruby lagoon]

(Huh, I thought that would just be ballistics gel. I guess it's that and a lot more)

[last epoch]

(liquid suspensions for skin models seem pretty neat!)

[opaque quartz]

lol

i think it could work if we went for a gel that's very opaque

[last epoch]

no way, slime nwero

[opaque quartz]

i think it has a very low melting point and will go off over time. Also the texture seems like it could be weird to interact with. Maybe if we found a really good gel???

[red oyster]

thin silicon?

[ruby lagoon]

There's a silicone/polyurethane section in that large doc too

[opaque quartz]

silicon is a bit too tough, it sits strangely on the skull

lemme grab some images

none of them look good

i looked everywhere but every single silicone skin mold looks horrifying

[ruby lagoon]

Might be nut-picking here. What were the materials for Avatar and that one animatronic severed head?

[opaque quartz]

not sure, but even that was kinda 'eh'

[ruby lagoon]

Also I'm not sure the top right is uncanney, just ugly and bad lighting

[opaque quartz]

part of it is the face geometry, but it's also because the material stretches weirdly

i think a gel or foam would look more like real skin

[ruby lagoon]

Well, can you find gel/foam face examples?

[opaque quartz]

this is the best one i could find but at that point, too far towards realism

[opaque quartz]

not seen any, looking currently

i think it's unexplored because people don't like an abstract face analogue

if we could get something more like this it'd be ideal. Though this does not reflect real life, it's just a (probably AI generated) piece of concept art

we'd need to find materials that can be shaped precisely, with high elasticity, light weight and deformable

silicone tends to be heavy and 'droop' as well as stretch in an unusual way since it's applied as a face mask instead of a sculpt

either way what i've seen of silicone, even with the best sculptors, makes me want to steer clear and find a better option

this is the closest thing i could find to possible foam representation. You can see how a less transpart foam would work well?

enough transparency to mimic the skin refraction as well

[ruby lagoon]

I feel like this is getting into materials science stuff that I don't know as much about

[opaque quartz]

for now i'm going to aim for face mimicry using rigid plates that are actuated

but i'll keep foams and gels in mind

[night tartan]

This is all geroglific to me, but it's interesting. Neuro would be so huggable with this

[opaque quartz]

like this but with lines

[opaque quartz]

kinda like baymax

[night tartan]

Very adorable

[opaque quartz]

i just realized that part of the problen with silicone is that it's not transparent enough or light enough. It creates harsh shadows that make the skin look very dead and unnatural. Natural skin absorbs and refracts light

foams and gels don't have this issue

[ruby lagoon]

Y'know, when you think about it, our face skin is kinda like foam because our skin is made of cells with cell walls and fluidy/jelly insides
Maybe foam filled with something could work best

[night tartan]

Jelly foam?

[ruby lagoon]

jelly foam

[opaque quartz]

conversely the foam shouldn't be filled because it has inverse properties; Solid strands with empty pockets, as opposed solid pockets with empty strands.

which would explain why it looks and acts so much like skin

[ruby lagoon]

our skin is basically sand, but if sand were squishy

[opaque quartz]

ye

goo sand

[red oyster]

kinetic sand

[opaque quartz]

if you look at the top shaded corner you'll see it basically has the same refractive index as skin, but also similar elastic and compressive qualities, as well as weight (very light)

so i think that it'd look very analagous

likewise because it's not heavy, it'd have none of the droop that silicone has

as people age their skin loses elasticity and droops more, the same appearrance as the silicone masks

meanwhile this foam is very elastic

[ruby lagoon]

NeuroPillow

Could work if we get a good foam. Might want some thin layer of something over the foam to prevent liquids from getting stuf in the foam though. Maybe something that can be sprayed on

[rugged skiff]

It really doesn’t look like skin though.

[opaque quartz]

does it need to look like skin so long as it looks good?

[rugged skiff]

It wont look good.

[ruby lagoon]

I think it should be the right color, which shouldn't be too hard

[rugged skiff]

It will look very uncanny.

[ruby lagoon]

maybe put some food coloring / 3d print resin coloring or something in the foam solution before making the foam

[night tartan]

Maybe that orange one on the very far right? Something like that color?

[opaque quartz]

nah

[rugged skiff]

That still looks horrible.

[opaque quartz]

we'd want the raw stuff and of the correct material

[ruby lagoon]

very, very thin cell foam, with skin color coloring in it, same way you'd color silicone

[night tartan]

gotta search better

[opaque quartz]

i dont like that idea, i think it'd look too fake but also too much like it's trying to be human

with the uncoloured stuff it'd have more resemblence to a marble statue

[rugged skiff]

I thought you said before you would like Neuro to have the chance at integrating into human society? This wouldn’t let her have that chance.

[opaque quartz]

the material science doesn't yet exist for someone to say "that's human skin" any integration would have to be done through making her believably expressive without looking creepy or human

since we can't make her look identical to a human

but we can make her look human-adjacent like a stone statue looks human

[ruby lagoon]

(Still feel like these types of robots are good enough for that tbh)

[rugged skiff]

It’s still a goal that can be strived for. It can still be done without making her too creepy.

[opaque quartz]

https://www.youtube.com/watch?v=ZSWJOkT2Ejo

the lighter foams look beleivably like skin. Moreso than silicone imo and with more springiness to them

best material for that job i've seen so far

please recommend any that you find though because if we can go with a better one, we should

[rugged skiff]

As long as it can be covered by another material.

[opaque quartz]

what material would look more like skin than it though?

[rugged skiff]

Honestly, silicone looks more realistic but there will be better alternatives in the future.

[night tartan]

The good thing about foam is.. that is very common and surprisingly cheap.
It can probably be custom made to match what we want, including color and porous levels

Also, makes neuro huggable

[last epoch]

Im assuming the skin will be seperated into panels / regions to remove any mechanical resistance during movement?

[rugged skiff]

One color or shade wouldn’t work.

Part of the art is in making the skin. I was going to literally going to take classes for this purpose. It would be a life goal.

[night tartan]

So... we should go with the whitest one so it's easier to paint over it? Like a blank canvass?

[rugged skiff]

Honestly I feel the best option would be self healing polymers, even if the foam is strong it could still rip.

[night tartan]

self... healing..? Now that's another level. I'm interested.

[rugged skiff]

Self-healing polymers and self-healing elastomers are currently being researched as an option for skin in prosthetics.

I’m a sculptor, if I could learn how to make this into an art form, this could be where I contribute to the project.

Helping to make an android has been a dream since I was a kid tbh.

[night tartan]

Consider your dream a reality. This is no ordinary android and we're not making it for an ordinary AI.
As I thought, this is not a matter of "If", but a matter of "when".
And that "when", is basically now, because we're putting so much effort into planning it

[rugged skiff]

I also believe that AI will have the potential to vastly increase the rate of progress in material science, possibly within a decade, especially in regard to medical applications such as prosthetics. Creating an AI that does this is a goal for companies like OpenAI.

So there will likely be more options in the near future.

Although I agree that the V1 humanoid robot shouldn’t have looking human as a goal, first we just need to make something that works.

[ruby lagoon]

I've been on top of the AI research and OpenAI's one of the slower ones now. I think NVidia is good, though the most interesting research seems to come from random universities everywhere

[rugged skiff]

Yeah, universities are where most of these developments are made, I just mentioned OpenAI as an example since using AI to advance science is a stated goal by them.

[ruby lagoon]

Ah, fair

[red oyster]

Can we add a gyroscope to a motorcycle and teach neuro to drive one?

[opaque quartz]

yes

[red oyster]

Hmmm

I've been braining for a while and got nothing new

We need more people with individualized skills

[opal bane]

@ruby lagoon how's that summarization/skill finding ai thing going

[ruby lagoon]

ehh...

[opal bane]

That bad eh

[ruby lagoon]

I think I talked about a bunch of different subjects, so the skill directory includes me in a buynch of different places, then Matthew, then a few others.
idk, I guess I'll just post it and ppl can suggest skills the AI & I listed (suggestion mode should be on): https://docs.google.com/document/d/1rbjfbVYhFqKnM_PiybnRFu_ikKArPxsPAB94FAvkArM/edit?usp=sharing

[opal bane]

Showing some favoritism I see...

[ruby lagoon]

Heh, yea, that's why I didn't post it

(huh:

[opal bane]

IT'S HIM

The heck am i recording though

[ruby lagoon]

Anyone should be able to add their own skills to that skill tree on the first page or two

Ah, that recording.
Idk, the AI kinda just made skills up somtimes I guess

[opal bane]

I could try stealing people from general rather than just NN, what do we think we need?

[ruby lagoon]

Actually, lemme think...

• Electrical Engineering - The robot battery UPS system specialized for high power output is experimental, and new motor control systems may be needed
• FPGA programming (open source toolchains especially, and avoiding ITAR problems) - controlling so many motors and getting force/sensor feedback should be cheaper with FPGAs, and bi-directional BLDC motor control may need it
• FreeCAD designer - I'm mostly using FreeCAD for programmatic CAD and still don't know how a lot of stuff in it works, and getting feedback on the plugins I'm making would be helpful

[red oyster]

Cad is easier with my multi-thousand dollar software

[ruby lagoon]

bet

Try making working planetary gearboxes that are as small as possible

I didn't choose FreeCAD because it was free. I chose it because it supports Python

[red oyster]

Okay

What diameter are you looking for?

[ruby lagoon]

As small as possible

ratio 4 could work as a test, with 10kg cm input torque, so 40 kg cm output torque

height of 1 cm

Looks like the diameter is 3.7cm if I use a s:6,P:6,R:18 gearbox with 4 planets, and make it with TPU (coating or full)

Still need to connect the finish button to the auto-generate script though. Also the image diagrams keep disappearing from the qtdesigner generation

[opaque quartz]

is there a shaft diameter calculator?

[ruby lagoon]

Nope, haven't figured out the shaft/key stuff yet. I feel like that will have to be different calculator that will cut into the gears

(the planetary gearbox, somewhat manually entered)

[opaque quartz]

the shape is so pretty

[ruby lagoon]

Yea... the ring's kinda thin tho...

[opaque quartz]

a bit yeah

[ruby lagoon]

OK there we go:

2.5 times the module seems like a good setting

[red oyster]

Can you export that as a cad file?

[ruby lagoon]

yep

[red oyster]

Send to me if possible

[ruby lagoon]

Which filetype?

step?

[red oyster]

Obj or sldasm

If possible

[ruby lagoon]

I'd be surprised if sldasm is an option tbh

might be proprietary

[opaque quartz]

https://media.tenor.com/5jMowsDfBv8AAAAM/iron-man-hammering.gif

[ruby lagoon]

(double checking the obj opens in blender...)

Works!

It might have converted millimeters to meters though. Not sure

[red oyster]

Fixable

[ruby lagoon]

divide by 1000

(Ah, oops, it should've had 4 planets... I messed up the manual generation)

All that said, I still use OnShape/Solidworks sometimes because it has a much better UI

[red oyster]

I'm Finally heading home

[floral quarry]

Jesus Christ on a motorbike you guys discuss too much while I sleep

[red oyster]

I ride a motorcycle, that's why I thought about it

[floral quarry]

Based, which one

[red oyster]

Cfmoto 300 nk

it's my first bike

[floral quarry]

Poggers

I take my unc's hypermotard 698 mono often

[red oyster]

hhhhhhhh

[ruby lagoon]

Obj

Step would probably be better

[red oyster]

Yes, but obj allows me to do other fun stuff

[ruby lagoon]

?

[red oyster]

More precise measurements

[ruby lagoon]

Huh

[red oyster]

ez

can't do that with step

[ruby lagoon]

It is now solid

[red oyster]

It is now isollated parts

[ruby lagoon]

Was it split up in the obj or one mesh like stl? I just clicked export so I have no idea

[red oyster]

5 meshes integrated together, I had to isolate themm, rather difficult

Anyway

Diameeter of 35.42 is rather large

[ruby lagoon]

35.42mm. smaller than expected...

[red oyster]

yes

[ruby lagoon]

The module should be right. I didn't connect the ring thickness with my gui though, so I think I was adding 2 whole modules worth of thickness while... honestly I need to read what that freecad.gears thickness parameter is doing exactly and see if it's based off of pitch circle diameter or what

Gotta fix my code some more...

[red oyster]

Okie

Imma go buy a drink

[red oyster]

Back

[ruby lagoon]

hi

[red oyster]

hi

[opaque quartz]

I've been doing research into solid state batteries and they are an attractive option if we end up having time for battery testing and prototyping. We can very likely design our own using non-toxic materials.

[opaque quartz]

Also maybe i'm crazy, but can't we use compressed amorphous metal as a high density battery? If we squish it using a one way screwing mechanism, then when we unlatch the one way lock it can drive an escapement that creates constant power. It's the same phenomena as lithium ion or solid state batteries that are effectively electrically-operated springs.

[floral quarry]

I guess so... your idea seems good

https://youtu.be/LmA3PBEyu3g?si=2VPc8KxwFMvjWUiY

Bulletproof neuro

[red oyster]

They just developed a material that's stronger than kevlar too

Idk if it's that or not

I just remember reading about it

[opaque quartz]

i like that he goes right into it

[floral quarry]

"I'm not into this clickbait crap"

[opaque quartz]

https://electricityforum.com/td/energy-storage/compressed-air-energy-storage

compressed air energy storage is another possible energy storage method

however the downside is that either neuro would be dependent on an external air compressor, or we'd need to make our own air compressor inside her torso that can be driven by a burned fuel

Also the best storage requires huge pressures and therefore we'd need to design a very durable pressure chamber for her.

[opaque quartz]

https://www.youtube.com/watch?v=sENgdSF8ppA

[night tartan]

If this works, we could stay out of the inevitable loss of capacity of lithium batteries

[ruby lagoon]

I think we can only avoid wear/loss of capacity/breaking down by operating near zero kelvin. There's sometimes zero friction/resistance there so no cause for wear
If we were making a robot body to walk on Pluto then maybe

[opal bane]

Push it to v14 where neuro becomes ruler of pluto

[opaque quartz]

of the various battery options I explored, solid state batteries look the most practical

They are very similar to the compressed metal battery, except there are less moving parts and the pressure is generated through charging the battery, and expelled when discharging the battery

materials such as lithium metal, graphite, sulphur, and various types of solid electrolytes such as garnets, ceramics, and glasses are used to make them. So other than possibly rare or harmful solid electrolyte materials, creating them should be somewhat simple

[opal bane]

Excuse me are we inventing a new kind of battery now

[opaque quartz]

the main thing is ensuring the lithium ions can travel through the solid elctrolyte

[opaque quartz]

for the compression battery, yes, because i cannot find any information on this kind of battery ever being discussed or attempted

maybe somewhere it has been, but i cannot find that anwywhere

[opal bane]

Your tactics confuse and frighten me, sir

[red oyster]

Typical neurobot project activities

[opaque quartz]

it's based off of the physics of spring-based batteries in watches, but attempting to improve energy density to match lithium ion and solid state, which have no empty space in the material of the battery, so have better compactness

[red oyster]

I'm watching a video on ss batteries

[opaque quartz]

the batteries we use to day are based on compression and relaxation of the material when lithium ions move through it

wheras the proposed compression battery uses an external force

[opal bane]

If this works this has to be one of the most absurd reasons something has been invented

[opaque quartz]

[red oyster]

Yes 😎

[opal bane]

Right up there with making potato chips out of raw spite

[opal bane]

This is actually me in every other conversation in this thread

[opaque quartz]

https://amorphology.com/technology/applications/
https://www.tevema.com/improving-technical-springs-high-entropy-alloys/
https://www.tevema.com/applying-springs-design-principles-in-novel-systems/
https://www.tevema.com/technical-springs-game-changers-in-energy-storage-harvesting/
https://electricityforum.com/td/energy-storage/compressed-air-energy-storage
https://www.youtube.com/watch?v=jVoRoK1u3Dg

[opaque quartz]

[red oyster]

Look into the new Catl ss batteries

They just got released

A few weeks ago

[opaque quartz]

https://www.youtube.com/watch?v=m0HrgzoNDEg

[opaque quartz]

oh nice

Nice, 500 is in the upper bounds of energy density of various tested products from companies

[red oyster]

So the issue with SS batteries is they don't have the same ability to maintain dendrite growth

Cuts down in battery life until the problem is solved

And also, dendrite reaching from anode to cathode = kaboom

[opaque quartz]

glass batteries have no dendrite growth so might be a good option to look into

can't remember what garnet oxide electrolytes were like in terms of dendrites

[red oyster]

Oh, those use graphene right?

[opaque quartz]

nah for solid state

https://en.wikipedia.org/wiki/Glass_battery

usus lithium or sodium anode

[red oyster]

Oh

Interesting

[opaque quartz]

not sure what the cathode is

i don't think we should mess around with exotic glass manufacture so it might be best to source that material online

fairly sure solid state battery materials are sintered together

https://www.sciencedirect.com/science/article/pii/S2666539523001694

[red oyster]

So one thing that might be worth thinking about is how long we want neuro to run for and how often. If she's only running a few hours a day, that changes the battery necessities than if we want her active 24/7

[opal bane]

2-4 hours biweekly max for v1 at least

[opaque quartz]

yup, and another thing that affects battery life is efficiency of the walk cycle, actuation, and locked-position ballancing

depending on how efficient we go, and how long we want her active, and her peak strength, we can determine power requirements

the more power density the better though

for either smaller battery pack or longer range

[ruby lagoon]

(Y'know, if we didn't have to worry about safety and stuff, I feel like this project would be way easier)
(Then we could just use thermal-nuclear-isotope generators or compressed air that could burst and slice through any objects within a few meters, and it'd be just fine)

[red oyster]

I'm down

[opaque quartz]

biggest power hog is the motors because we have to turn energy into torque necessary to lift her body weight from one limb

[opaque quartz]

https://www.youtube.com/watch?v=LyA6uKEGrr0

was just thinking about that

what a coincidence

[opal bane]

While my gut tells me to assume maximum strength for power capacity testing I think that would hinder the finished version's ability to fight whatever superheros exist over in the uk (or wherever vedal lives)

[ruby lagoon]

probably a good mindset for this project

[red oyster]

Let's not go nuclear for v1

[ruby lagoon]

v2 then

[red oyster]

Sure

[opaque quartz]

i was thinking we tackle fusion for v3

[red oyster]

Micro Dyson sphere around a nuclear fusion reactor in v4, and a black hole radiation generator for v5

[opal bane]

V3: avengers level threat

Put that in the doc rn

[opaque quartz]

problem with magnetic confinement fusion for neuro is that the smaller the plasma, the harder to stabilize it is, so you require more energy to magnetically confine it - Meanwhile the existing best reactors still haven't broken even so i don't like our odds when using a much smaller scale fusor.

The biggest limit for hobby fusors is the grid they use inside the fusor, which melts, but a toroidal plasma tokamak style fusor wouldn't have that problem. however the fact that the bigger (and therefore more energy efficient) reactors havn't broken even means that there's no chance in hell of a hobby fusor using that method working.

The amount of hydrogen and pressure would have to be even more than in the big reactor and it'd need enormous electromagnets as well. Simply not possible.

The other approaches to fusion are all less promising than tokamac so there really isn't any way forward with small scale fusion.

[red oyster]

Black hole radiation?

[opal bane]

Do i need to say it again

[opaque quartz]

ah i only know the barest surface level details about fusion

[opal bane]

We're designing a robot not the U.S.S. Enterprise

[opaque quartz]

😂 sometimes i wonder

[red oyster]

Depends, maybe that would be easier because of the larger scale

And we'll let neuro pilot it

[opal bane]

Damnit

[red oyster]

We could pull a Tony stark and invent a new element

Maybe that would help

[opal bane]

Maybe i could learn to spin a stick really fast and fly

[red oyster]

Probably could

[ruby lagoon]

Honetly I think black hole radiation and nano-portal stuff might actually be more realistic than small scale fusion just due to physics.
Apparently fusion's viability scales with the fourth power, so just making a fusion reactor the size of a country or planet would make it much better

[opal bane]

Don't make me find adult supervision because other engineers clearly dont count...

[opaque quartz]

plasma is more stable in the bigger reactors

[red oyster]

So what you're saying is we buy mars and turn it into a reactor to power a human sized neuro bot?

[ruby lagoon]

yes exactly

[red oyster]

Sounds good to me

[opaque quartz]

there was a fusion startup a few years back that uses microwave plasma to shoot two plasma toroids at each other and collide at the center of an hourglass shaped reactor to artificially increase the pressure in a small area, but the biggest limitation for that is materials science. The reactor can be smaller but the walls themselves start to break down immediately because of the temperature and pressure of the collission

[ruby lagoon]

micro-robots could continually heal the chamber
Also, they could continually rebuild hydrogen bombs and the container of a giant pool of water

[red oyster]

I'm going to make a rudimentary leg model after work

Nothing crazy

[ruby lagoon]

nice

[red oyster]

The hip is a bitch of a joint for mechanical recreation

[ruby lagoon]

True. Same with the shoulders

[red oyster]

I'm less worried about the shoulders actually, still difficult though

[floral quarry]

@opaque quartz your knowledge makes me feel like a fraud

The amount of knowledge this guy has is insane

[red oyster]

His knowledge is very generalized for sure

I don't have that wide variety of knowledge

I only have my one specialty

Which is CAD

And my ability to learn new things very quickly I guess

Most people are specialized

Like you, there's probably one thing that you are better at than the rest of us

[floral quarry]

And im still in high school 😭 worst debuff ever

[red oyster]

I designed a self adjusting basketball hoop for special needs children in highschool

[floral quarry]

We were making an ai powered election voting machine for the student council

[red oyster]

What was the AI for?

[floral quarry]

But later it got cancelled because the school found out the elections were done through fraud (they paid every kid money to vote for them)

[red oyster]

LMAO

[floral quarry]

Biometrics, verify the person voting, ensure no double votes and prevent tampering and fraud

[red oyster]

Fair enough

[floral quarry]

The machine had a research cycle of 2 years but we dropped it after 1.5 because the elections happened and there was massive fraud

Like, they went to every, and I mean every class for the shiny resume stuff

[red oyster]

Idk that a database that houses, analyzes, and compares biometrics is quite the same as an AI, but it's cool regardless

[floral quarry]

There was a camera to prevent tampering and fraud during the process, and it used a custom ai I wrote

[red oyster]

Facial recognition?

[floral quarry]

Yes

Dlib and opencv

[red oyster]

I guess that is a basic level AI then

[floral quarry]

I guess

[red oyster]

How large was the learning dataset?

[floral quarry]

School kids we saw everywhere

[red oyster]

Lmao

I'm not skilled enough with code to write an AI

[floral quarry]

My real ai work was when we wrote a fucking missile guidance system for a competition but didn't get appreciation because "you need to have an sdg to solve"

And the sdg we chose was ensure peace

The reason for a missile? Our previous president said "to ensure you have peace, you must have a bigger stick than the enemy"

[opal bane]

Managed democracy!

[floral quarry]

And because the previous president was responsible for kickstarting our missile program

What about you folks

Send some funny stuff our way

[opal bane]

I massively irritated a ai made to act like Julius Caesar by repeatedly saying "egg"

That's the only thing i got that's ai related

[floral quarry]

Nah, anything funny works

[opal bane]

My comedy works best with context (see: most of my messages in this thread)

[red oyster]

I'm watching tutorials about advanced Solidworks controls while I'm at work

I'm cooked

[opal bane]

And also playing geometry dash

[red oyster]

Nah, that's on my PC at home

I could wirelessly connect to my PC from work using my phone and work on cad rn, but that sounds miserable

Guys, I need ideas for an ankle joint

I have one, but too much friction

Unless...

Now that's interesting

A pinion gear with a radial rack gear

And the rack is mounted on a pivot point so there can be rotation on a perpendicular direction without misaligning the rack and pinion

@ruby lagoon thoughts?

[ruby lagoon]

Honestly I don't think the joints should be joined by gears at all. There's the problem of gear teeth breaking, pulling and slipping, pulling and twisting then meshing the gears at a point isntead of a line and crushing teeth, etc.

[red oyster]

How else do you suggest we handle the heavier load joints like the knees and ankles?

[ruby lagoon]

use some sort of linear actuator and pull (and maybe push) on a connection point on some foot bone
And for the joints, use something like compliant hinge joints or, saddle joints with springs/rubber pulling the joints together

[red oyster]

Also, human toes support 75% of the weight when stepping, that's a bit much for robot toes

Well, up to 75%

[ruby lagoon]

I think the way the skeleton is designed is actually really good, due to how much contact area the bones have when pushed in, and how cartilage & ligaments hold the joints in continually

I think motors or typical hinge joints may be better when there's only one axis of rotation, but when there's several I think the saddle/ball-in-socket joints are way better

(I mean, copy-pasting the skeleton might be kinda gross, but saddle/ball joints with springs/rubber around them would probably work really well if we use linear actuators. Also, it could be further improved by making ptfe caps to those joints and making some squishy containers around the joints holding gypsum fluid for its non-newtonian properties)

[red oyster]

I think the knee is best driven by a motor purely for speed and force

It's a very strong joint

Ankle is fine with either motors or actuators, just depends on what we choose

[ruby lagoon]

Yeah, the knee is mostly a hinge joint, so that could work well. Though I think the motor would be in the upper leg and geared down and then the rotation turned 90 degrees with a bevel gear

[red oyster]

The knee needs about 110° of movement, from fully extended backwards

There is no frontal movement there

I take that back, it's between 120 and 140

If we can hit 130+ flexion angle, I think that's adequate

And I think it would be better with a brushless motor and encoder rather than a stepper motor

Just for the power again

Work has this laying around

[ruby lagoon]

Yeah, brushless motors can operate at 24v or higher because they don't cause pitting in the metal due to sparks and are designed for speed unlike steppers, so compared to motors that operate at say 6v or 12v
Well, V=IR, and W=I*V, so v->2v means I->2I and W->2*i*2*v=4W, or 16 times stronger in the case of 6->24 volts. There are also 48 volt options for larger BLDCs
BLDCs can be really strong for their size

[red oyster]

Yeah, I wonder if they have small enough motors for small joints

[ruby lagoon]

I think if ordered in bulk they do

[red oyster]

http://www.constarmotor.com/productinfo/6534.html

Finger sized motor

Need an encoder that size tho

Would be a custom PCB probably

[rugged skiff]

You did that and you’re still in highschool? TF have I been doing with my life?

[red oyster]

Lmao, facial recognition isn't all that complicated

You could learn to do it pretty easily

[rugged skiff]

Making an AI in highschool? Back when I was in highschool I just played video games and watched anime.

That’s impressive dude

Where did you learn to program?

[ruby lagoon]

might be too small for most stuff actually

[red oyster]

Some of us are just nerds to a different degree than the rest lol

[red oyster]

It's the biggest one that could reasonably fit in the diameter of neuros finger

So if it's too weak, motors are out

[ruby lagoon]

please do not put any motors in the fingers

[rugged skiff]

Yeah I guess.

[ruby lagoon]

motors do not go there

[red oyster]

Motors go everywhere for funny

The fingers will not have motors though

[ruby lagoon]

Well, I guess if you want the fingers to spin or something crazy

[red oyster]

Not when I get the real design started

Have we thought about small scale chain drive?

[ruby lagoon]

I think once maybe

[red oyster]

Interesting

[opaque quartz]

lemme draw my idea i had for the knee pulley real quick

[red oyster]

Also, there is a reason that I bring up ideas, even if they are clearly not great ideas. Mentioning something could set off a chain reaction of thoughts that lead to a good idea, so I'd rather not miss that opportunity.

The amount I wish I could drop my job and make this a full time project...

[ruby lagoon]

True. My ferrofluid solenoid idea was bad, but I think the normal solenoids as face actuators that came from it might be good

[opaque quartz]

very basic drawing. It uses a central motor side-on from the femur bone. It drives a bar on a hinge near the knee, which provides leverage to the bar. That bar slides a rail, from which pulleys connect to winches, which connect to the tibia. So essentially the output power of the motor is multiplied for the same energy requirements.

I'm thinking that in any situation we're able to, we should use leverage.

---

for the knee joint itself, we should copy biology and make use of the screw home mechanism for natural locking.

The screw home mechanism is a phenomenon where, due to the differrent sizes of the two ball joints in the knee, there is inwards rotation during the last 10 degrees of leg straightening. What this effectively does is change the angle of rotation so that gravity can no longer actuate the joint.

This allows the leg to be locked straight and rely on bone rigidity rather than muscle strength power when standing upright.

Here:

Without a form of locking mechanism such as a friction brake, or rotation of the joint, the range of the motion massively decreases due to how much energy is spent holding the body weight of the robot 24/7

in humans, the weight is carried by the skeleton

[red oyster]

I'm not understanding how this will make the right movement or a controlled angle and speed.

Huh

[opaque quartz]

reposted with arrows

the pulleys are connected via wires to a roller on the top rail

the motor drives the straight rod back and forth, but not 360, so it remains in its position of most leverage

the pulleys connected to the roller on the rail, pull on the tibia bone

the lengths of the pulley ropes are different, so that on the forward stroke the bone does not go far beyond 90 degrees, however on the return stroke, the tibia folds completely against the femur

[fossil drum]

hello guys what did i miss

[opal bane]

we got plans for V5 now

[opaque quartz]

Gearboxes + battery research + joints and such

i'm craving some toblarone rn

[red oyster]

That sounds good lol

I'm a milky way guy myself though

[opaque quartz]

based

those were my childhood

[fossil drum]

excuse me???

[fossil drum]

nice

[red oyster]

I think if we used quantum entangled particles at a high enough quantity for wireless power transmission, we could store all her batteries outside the body at a separate location where they can remain plugged in at all times.

Then weight isn't an issue with them