To begin: the project goal: I am working to make a series of humanoid robots. I am using a Biblical theme of naming the first 3 robots I make Adam, Eve, and Abel. The goal is for these robots to have human body inspired musculoskeletal systems, advanced AI, and that they look human and pass for human to a casual observer at least at a distance. They must be able to walk, talk, run, dance, do sports, do chores, manufacture products, and make more robots just like themselves if not even better. My aim is to build a single robot arm and head and then add sufficiently advanced AI to that arm and head to enable it to build the rest of its own body for me. This way I am delegating the work of building the majority of my first humanoid robot to that robot rather than doing that work myself - and this is to save me time.In a like manner, my goal with the AI is to code just enough AI that the AI can begin coding itself and this way I don't have to code most of the AI myself because it will self create itself. I liken this to building a seed and that seed growing into a tree because for me to code that tree would take too long for me and just creating the seed would then save me time.
Here is my CAD image for the Eve robot:
My primary focus at my current progression is building the arm for the Abel robot. I am using a off the shelf PVC medical skeleton as the basis framework of the arm. I have added artificial ligaments made of workout shirt fabric and bone sleeves made of 1000 denier nylon fabric taped onto the bones with adhesive transfer tape and also sewn tightly around the bone for a snug fit. I then am able to suture all components onto these bone sleeves using nylon upholstery thread and a curved suturing needle.
During the project's last 10 years of development, I have mainly done large amounts of research and planning as well as trial and error. Making plans, attempting to implement them, and finding out why they did not work. These setbacks were a form a failing forward, learning what not to do and narrowing down my options over time. For example, I ruled out using geared servomotors because gears are loud and having the motor and gears assembled together with the circuit board and potentiometer all crammed into a little black box was not space efficient, as far as form factor goes, for placing these assemblies into the robot's available spaces where muscles would normally be. I just wouldn't be able to fit many of these. So I needed to elongate and stretch out the form factor by mounting the motor, downgear, and circuitboards separately, spreading out each of these components of a servo. I also did not like the massive amounts of gear noises that metal geared servos give off. So I needed to find a way to downgear motors without using gears. That led me down a rabbit hole of learning to downgear by way of pulleys which can be silent and robust if done well IMO. The general idea is to have the motor output shaft act as a winch to reel in strong braided PE fishing line which then will be downgeared by a pulley system just like a crane uses or similar to downgear its cables. Then after downgearing the fishing line will have the proper torque and speed to actuate the joint it is assigned to. By copying the skeletal arrangement of the human body exactly, I am also able to copy the position, strength,and orientation of the muscles of the human body exactly to reproduce their function on the skeleton of my robot. This way I'm using a proven successful design, the human body, to ensure the success of my bio inspired design. Developing this pulley system has been much harder than I anticipated and the space it takes up is tough to accommodate.
>>2860811you can probably get some decent off the shelf genitals from china or nippon or something.
This thread's main thumbnail is a CAD image of my Abel robot which shows the little BLDC motors I selected for each muscle replacement and mounted in CAD everywhere space allowed for it. There is nearly a 1:1 ratio of motors to muscles of a human body in this CAD. Although some muscles were so strong I used more than one motor to replicate its strength.
Here is a image of the hand curling the index finger.
Project website: http://www.artbyrobot.comFull humanoid robot building playlist: https://www.youtube.com/playlist?list=PLhd7_i6zzT5-MbwGz2gMv6RJy5FIW_lfn
>>2860808this is my android so far. The cad files are ready, as far as the structural parts are concerned. I'm running low on funds but when that gets sorted Ill have the skeleton fabricated from specialized alloys in china. Im still doing research on the best possible artificial muscle, which is quite difficult. The most difficult hardware part is going to be the skin since not much research exist on multi-composition silicone. I want to avoid the uncanny valley at all costs, unlike that amica thing. They used a single-composition silicone for the face and it looks so creepy
you will never own a robowaifuhttps://rumble.com/vkf7c1-casanova-of-the-new-western-oligopoly.html?e9s=src_v1_ucp
ive thought about this. seems clear to me nobody is going to work as a team. for smaller projects youd need a smaller budget and it would be easier to sell and get funding. then once you're established you have the means to hire people and then move on to the bigger project and protect it with ip. You can not sell a big project as a kit like some people claim. We ought to start smaller. Maybe toys? we are also at a disadvantage in the sense were not old but were not young. Its better to put some gas on it and go big. my baseline is kickstarter. i do have an idea. im going to refrain from saying like i usually do though. i dont know if were heading for a recession but it does feel kind of grim. but theres still need for innovation during bad times but its more about coping and survival than more risky or frivolous things. nobody buys pet rocks during a recession they buy beer. im just thinking out loud not saying id do any of that stuff.
>>2860968In theory a good sentiment, but none of the gooners have the attention span to back a kickstarter that would have to live thru years and years of R/D. And if you're gonna try to circumvent that by eg. making toys you might as well choose something more lucrative. I've semi-recently escaped 9/5 and my company is bound to hit mid 6 figures within the next 3-4 years. I'll start buying dividend stocks and just work on the bot after that. That's the plan at least.
>>2861048well youre in a different situation than mine i guess. okay do my idea was making a lockbox with a timer but its not really that great of an idea. i also had the portable nas idea which is doing well on kickstarter but it has a full range of features i was thinking of an sd card with bluetooth or something...im not really trying to circumvent getting a job its not really an option for me for reasons. well theres fiverr and upwork but even where im at im unsure...
>>2860822Looks good so far. I'm planning to make one robot skeleton from a sex doll's skeleton that I robbed from a old used doll I bought for a steal on ebay for parts. It's steel and relatively light weight. For another robot I'm making a hand made fiberglass skeleton. For another I'm using a PVC medical skeleton. Each of these approaches are in the $100-200 range and all 3 work IMO.Artifical muscle has not yielded anything remotely usable according to my research so I settled on bldc motors and dc motors as my only two viable options.For the skin, I think any silicone is fine. What matters is the artistry in texturing it and pigmenting it realistically. I might use plumbers silicone for lesser seen areas and use softer stuff for face and stuff that don't need to be as rugged.
childhood is wanting advanced realistic humanoid robotsadulthood is knowing vintage mechanical machines are good enough
For the Adam Robot:This was the first robot I started around 9 years ago. I built about half of its skeleton by sculpting it in clay, fiberglassing the clay, and then hollowing out the fiberglass bones to remove all clay from inside leaving just hollow bones. This is a light weight and strong way to make the skeleton for a robot. However, after some years distracted by other unrelated projects, I decided that just buying a PVC medical skeleton is a shortcut and I went this route to hurry up my progress toward my first robot for now. So I bought a PVC medical skeleton and named it Abel. Adam remains my first robot but Abel is the first robot I intend to complete. Once completed, I plan to have my Abel robot finish building the fiberglass skeleton for the Adam robot. In fact, Abel will build the entire Adam robot, not just the rest of the skeleton.For the Abel Robot:This is the second robot I began when I bought my PVC medical skeleton 4 years ago. I set aside the Adam robot temporarily to build the PVC medical skeleton-based robot which would be faster to build since I start with a finished skeleton - saving lots of time.For the Eve robot:So far the extent of progression on the Eve robot has been creating a 3d model of her in CAD. The Eve robot will have no "love holes" because adding those would be sinful and evil. It is a robot, not a biological woman after all and I will view her with all purity of heart and mind instead of using her to fulfill my lusts of my body. Instead I will walk by the Spirit.Eve will be beautiful. However, I will dress her modestly as God commands of all women because otherwise the robot woman would be a stumbling block to men which could cause them to lust after her which would be a sin. To tempt someone to sin is not loving and is evil and so my robot will not do this. My robot will dress in a way that is a good example to all women and is aimed toward not causing anybody to lust as a goal.
As to realistic humanoid robots being childish, perhaps so in a sense. However, consider that they can have end uses you can't get with mech looking ones. For example, shopping in a grocery store or doing an errand in public incognito where nobody knows its a robot so people leave it alone. Whereas an obvious robot is a theft target, vandalism target, etc. So its safer to use in public if it blends in with human look.
Here's picture of fiberglass robot bones in progress.
its hilarious in you faq your sweating talking about your robot not being a kid or a woman LOL like i know this is all a big deflection because at the same time you are building a cunnybot. like dancing? really? oh yeah i need that for my practical data man bot not for any sexual purposes. right buddy
>>2860808oh look another sex doll thread.
>>2861816 Well when I used to stream working on this live years ago, despite choosing to make Adam look just like me, and despite Adam being the ONLY robot I had planned at that time, I still was constantly told I was making a sexbot surely. There is no getting around this. That FAQ is in need of revision so thanks for pointing it out. This project is always growing in scope and evolving. But yeah avoiding it being a kid or female was supposed to avoid the sexbot accusations. It didn't. So now I'll just make w/e I don't care if someone thinks its a sexbot even while I claim it is not. That's on them. They won't have genitalia so that's evidence in my favor. The dancing thing - I did not even consider that could point to it being a sexbot until you mentioned it. I was thinking the Carlton dance and stuff. Like dancing is very hard to do well especially very advanced dance moves like Michael Jackson and Chris Brow did. So trying to get it to pull off that would be entertaining and put to music would drive tons of views. Going viral with the robot is a plan to monetize it by youtube ads. One of many monetization ideas. So entertaining things it can do was a focus. I did not vet if each entertaining thing could be deemed sexual though. Dancing CAN BE sexual but is not necessarily. My robots will be very religious (artficially so) so they would not be willing to do anything sexual including any sexual dancing.>>2861916 high end sex dolls often have a human appearance and full range of motion quality steel endoskeleton, both of which can be used to make a humanoid robot that has no sexual intentions. So for at least one of my robots I will be using a sex doll's materials for the build while having no sexual motivations for the project at all.
>>2861816 actually I just checked my website faq and it didn't say I'm making it a male and avoiding making it a female or kid. No sure where you read that. I used to bring that up but its not on my website's faq anymore.
I’m just finishing the face up, everything else is working… and fully functional. FULLY.
>>2861937
>>2860811those are mid feet ngl hoss
>>2860808Great, make it work in simulation first.>musculoskeletalThat's a red flag. Just copying a human exactly won't cut it. Everything we have except muscle works differently than muscle. And we can't keep muscle alive for long on its own.>AI can begin coding itselfLol no.>>2860813>don't want to use motors because loudUtterly stupid. Especially because high torque electric rotarybactuators for robots are getting cheap on aliexpress.
Where the fuck is the dude who already did this but the terrifying miku doll? But please OP deliver some funny ass uncanny shit don't abandon us like half the dudes who do shit like this
>>2862115 it might just look weird from the camera distortions. If you saw it from every angle up close you probably would think the feet looked pretty good. Most of the feet were just from the original model I created in Make Human for free. I imported that model into Maya for massive further customization and perfecting. The feet I don't think I spent much time on though because I'm not too particular on them. They will be in shoes mostly anyways.
>>2862139 you seem to misunderstand some things. I am not using real bones or real muscles. I'm using plastic bones and custom gearless servos that downgear by way of pulleys so they're silent and these then winch in fishing line that ACT LIKE muscle in that they provide contractions that actuate the joints. Nothing will be living or biological at all.As far as AI coding itself, this is how: lets say the robot shuts the light off when it leaves the room. I can tell it don't do that when I'm in here! And it would put a new rule in its code itself saying don't shut off lights when leaving a room if he's in there. So it now has a new condtion it added to its code by learning. In that way it is coding itself. That's one of a million examples. But it is self coding in that sense. And with enough of these types of rules, it can even begin to code logic type stuff eventually I'm hoping that is more sophisticated than the example I just gave since rules build onto rules to make more logic I think.As far as not using motors because loud, you misunderstand. I don't want to use METAL GEARED motors because those gears are loud. I DO want to use motors but downgear by way of pulleys which is quiet.
>>2862164 people fall off these projects I agree. They literally disappear all the time. It's very annoying to me as well. I have been known to disappear for a bit, but I have always come back. I'm 10 years into it and never quit or disappeared fully. Yes I know that's a long time but this is R&D not just putting together a proven design from a kit. This is trial and error. And I had to get many years worth of education in to know wth I'm doing (and I arguably still am a noob, but a noob that at least knows A SIGNIFICANT AMOUNT - enough to run circles around a totally new noob, that's for sure.)
Some progress shots of the fiberglass skeleton robot.
>>2860822how did you even shape this? It's extraordinary, I can't even fathom the process
>>2862251open or closed shoes?
>>2862364 closed
radius bone hand made and ready to go
Here was a rough test of plumbers silicone and acrylic paint to make rubber skin for the robot.
> most powerful companies spend billions to make shitty robots that can barely walk> anon cooking up a reprap android in his 'shop for realWhat a blessed place
Thom Floutz is a big inspiration with his incredible silicone work he does. This is a example of his work. I want to go for this level of quality for the skin.
>>2862593You need to update your schizophrenia detectors.
>>2862573What is your actuators?I'd consider going with multilayered EAP (electro-active polymer) for a low energy, muscle-like action. Only downside is Wacker doesn't seem to offer Nexipal yet and DIY is of questionable quality and considerable manual labor: https://www.youtube.com/watch?v=Ga_IafGRWyEI hope you have this figured out as it will be the most important part of your robot. Even more so than a realistic silicone.How heavy stuff will it handle? What algorithms for the computer vision? How many of which CPUs?
>>2862573>>2862700>EAP actuators ...also sensors https://www.youtube.com/watch?v=9bt-g0uI5hg&t=400s
As to my actuators, the Dinah robot will use alot of N20 gear motors for fingers and stuff. The Adam and Eve and Abel robots will use bldc motors exclusively with silent downgearing by way of custom pulley downgaering systems. The Dinah robot is my fast and budget build. That one I'm not as concerned with keeping noisy metal gears off it.As to how heavy stuff will it handle, dunno. Hoping human level strength of a prime athelete age 30 but we'll see.The computer vision will be custom. I have a youtube video covering my AI for that so far.The CPU will be a gaming pc cpu. It will have a mini itx gaming pc in chest.
>>2862707>motors for fingers and stuffThen you want to make adaptations in the skeleton for that kind of heavily restricted actuation. All the moving parts and rope ("sinews") will need to be accessible and serviceable.I'd say a realistic humanoid robot requires a new material for movement. That material is not made into parts available in the retail - yet!
No, everything it needs is cheaply available retail. The artificial muscle thing with electric moving a set of cords is NOT in existence in any practical usable way. The amount of times people ignorantly recommend this is amazing. Literally it is non-viable.
>>2862253I know what I said.>plastic bonesMaking a robot that mimics human musculature is a mistake at present. You don't need human musculature and it adds unnecessary complexity. I bet you can't give me a good reason why other than it looks like a person.(it won't)>pulleyscable drives have countless problems. Namely, they aren't very stiff. This makes control difficult. Imagine how hard it is to push and pull something around if you had something rubbery between it and your hand>fishing lineExperiences creep under tension. You can use steel cable, but that experiences wear and fatigue. >robot shuts off the light when it leaves the roomLmao. You have no idea just how bad the state of AI is. I don't think you have any idea just how difficult it is to get a robot to turn off a light switch>learningRobots can't learn shit. Reinforcement learning basically doesn't fucking work IRL. Current RL takes too long to learn and what it learns is shit. Like it's only successful at doing stuff 80% of the time, so 20% of the time it's doesn't work. Oh you want it to write software too? Good fucking luck doing better than openAI>those gears are loudYou can actually build a humanoid robot in reasonable amounts of time with reasonable amounts of money if you just buy actuators off aliexpress. They are getting fucking cheap
Optimus uses fishing line now. TheRobotStudio uses fishing line. They create some of the best robots in the world. So no, fishing line is not a problem it is the best possible method.As to human musculature, the fact you think its just for looks says all we need to know. Go study anatomy and come back to us. It is not just looks. It is function beyond compare.Creep under tension does not matter. That's why you adjust constantly by reading joint angle and moving motors to fix.As to state of AI, the fact you think I'm a script kiddy hoping the state of AI will save me proves you really know little about me. I won't use jack squat of "cutting edge" AI. I will make my own that is elite.
Here's a photo of Adam's hollow fiberglass hand encased in artificial ligament made of adhesive transfer tape coated bones covered in spandex and stitched tightly together with upholstery thread.
>>2862798oh boy here we go
>>2862709No shit. All soft flexible materials we can make have shitty fatigue, creep and wear properties. Robots move a lot, so flexible stuff tends to break in a relatively short time, like months. Biology gets over this problem by being able to heal damage. So yeah, you can make a hand with tendons, but it'll only last a month or so. So instead of flexible rubbery crap we use steel because steel lasts so long it doesn't need to heal.>>2862783>optimus uses fishing lineNo it fucking doesn't. It probably bowden cables. Cables are still pretty shit>best possible methodNo it isn't. DaVinci surgical robots use steel cable because this experiences less creep. They still wear out within 10 uses though. Running a cable around a bend introduces varying load that fatigues the cable. And again, cables are less stiff than gears. Fishing line's ok for prototyping though>they create some of the best robots in the worldBest fucking how exactly? All these robots with anthropomorphic hands are a fucking joke. We don't have high resolution touch sensors that can fit in fingers. And AI's not smart enough to use them. They aren't very durable(cable drive s aren't). This makes it hard for AI to learn how to use them, because AI is very, very, very dumb. It's gonna bash that thing against walls repeatedly.>creep under tension doesn't matterGood fucking luck with that. It'll eventually break. And my experience with cable drives is that creep can cause the cable to loosen and slip off pulleys. Everything gets tangled up that way>it is function beyond compareMind telling me what function it has that can't be replicated by a rotary electric actuators?>I will make my own that is elite Prove it faggot, show your gym performancegymnasium.farama.org/environments/mujoco
>>2862720I'd study human anatomy to see how the biceps connects to the lower arm with tendons to create the strength needed to work like a human. A muscle expands in breadth to shorten in length and in turn pull a tendon attached to an adjacent bone which creates a fold in the elbow.With anatomically correct bones you'd want actuation resembling soft expanding muscles.With electric motors you have another movement pattern and it will make more sense taking advantage of conventional methods of robotics such as 360 rotation in the wrist etc. In turn it won't move like a human but it can of course be programed to use human tools!You are on an early stage of development your ideas of this robot arm and will have no use of either the hand bones or the servo motors... as you will find out.If it helps you stay motivated and interested in anatomy and robotics it is well worth the filament I guess. Here's a picture showing how muscles and ligaments in the arm looks:https://commons.wikimedia.org/wiki/File:Flexor_carpi_ulnaris.png#mw-jump-to-license>Anatomy of the Human BodyA servo instead of an expanding biceps requires a bone allowing fastening of it. It will probably wind up a chain or steel wire for the <mechanical advantage> comparable with a strong human. Good luck and rock on!
>>2862774> steel cable, but that experiences wear and fatigueYou can use that crazy stuff used to make glasses frames that you can twist around your finger and it just springs back. Flexon?I have to think of everything.
you guys trying to fly by flapping?
>>2862826So from experimental results fatigue life's probably like 2x better than spring steel. Fatigue limit's appears to be around 0.6% strain for nitinol here: https://pmc.ncbi.nlm.nih.gov/articles/PMC10228025/This page here is claiming you can get to 10 million cycles(which ought to be enough) at 3% strain. https://www.kelloggsresearchlabs.com/2019/01/06/fatigue/If that's really the case, there's a lot of potential. This is from a company that sells nitinol, so they could be full of shit.
>>2862821>mechanical advantagewill be less than one for muscles. The muscle needs to apply more force than is on the end of the arm.>>With anatomically correct bones you'd want actuation resembling soft expanding muscles.we don't have any artificial muscles that expand and contract exactly like the real thing. Patterned structure muscle sort of comes close, but it's made from shitty soft stuff and uses cables drives: https://twitter.com/lyosika50/status/1847150691355075002
Nitinol is total crap for a humanoid. Has been for over a decade. There's a reason it is literally NEVER used for them. As to AI being crap for controlling a tendon based robot hand, I agree, other people's AI can be crap at some things. Mine won't. Mine won't be like other's.
As to high maintenance predictions the fishing line will fail over time, doesn't matter. Regular maintenance is always an expectation and the robot, having already built the rest of its own body after a single arm and head are done, will be VERY well equipped to maintain itself so that there's zero maintenance I must do to it. It will also maintain my car.
I'm a robotic PhD that works on humanoid robots (I've mostly used THORMANG3). I obviously have not read all that shit but still I want to tell you some stuff. First you have a passion and seem to be completely autistic which is very important in research, I respect that you should try to approach a professor and get a PhD to learn and get paid for what you are doing. Secondly new hardware is cool, but in the field we would much rather use actuators that have precise and fast controls. There's a reason for that you should aim to make your shit repeatable fast and precise, that's what we need rn, too many "makers" like you only focus on the end goal and end up with dogshit robots that move like a puppet because they are bad at control theory. With good control even the shittiest of the actuators can look natural : https://youtu.be/KSvLcr5HtNcand with bad control even the fanciest of the robots with the most advanced actuators will look like shit. Don't forget mechanical design is for low IQ retards, controlling your mechanical design is the thing
>>2862962well I don't have autism but I take your thinking I do as a compliment in this context. Also, I'm rolling my own circuitry for driving the bldc motors and own custom algorithms so it will be elite. No worries there. Like I said, it will look and move humanly enough to pass for human to a casual observer not paying close attention.
>>2862962 as far as getting a phd to get paid for my research, that's not a bad idea but meh. I'd have to go back and finish my bachelor's (3 semesters) then get a masters then get phd. I don't know how the phd process works like am I guaranteed to get paid to do the robots I'm already doing or are there strings attached? That's a lot of time and money just to try to farm a little cash for my passion projects. Better to just make money in my existing business endeavors and side hustles and investments and make nothing doing robots until one day the robots make me money by manufacturing products for me and whtanot. And youtube ad revenue they will bring it if they go viral doing cool stuff.
Clay sculpted skeleton for the fiberglass Adam robot.
>>2862962Autism and schizophrenia might look similar at a distance, but if you read the whole thread, you'll find that it is, in fact, schizophrenia. Also if you want to reach next level in schizo detection and not have to read tl;dr schizoposts, the people making replica human bones for robotics projects are not autists, but schizophrenics.
heres hannadevs discordhttps://discord.gg/x6apUxx6the idea of usibg pulleys because theyre quieter is not without merit though. people have complained about hannas devs robot being noisy. ive never seen pulse width modulation make servos quieter but i know they can make stepper motors quieter since thats what happened when i flashed the firmware for my ender 3. as for the robot looking like an old man i dont agree with the design desicion. the skin is mostly relevant if youre planning to make a sex robot. what do people think when they see an old man robot on a forum for sex dolls...
>>2862327thank you. I modelled a human girl in blender, then from that a skeletal structure and then technical details from an engineering perspective, going back and forth between modelling and parts/systems R&D. Right now everything is geared towards modularity, everything is hot swappable with a couple of machine screws. The bot in the picture doesnt have the jaw or the mimic actuators fitted since those would restrict modularity and are still in the research phase. The rest is 3D printing, surface finishing, casting with resin, fitting electronics, making linkakes from gear rods and bowdens. right now im testing tiny camera modules that fit in the pupils of the eyes without much distortion and enough resolution to provide the bot with spacial recognition. eventually ill have the skull components made from alloy which allows for more compact, precise and robust parts than ABS/PLA printing.
>>2863044that looks pretty good ngl. It looks like its painted metal as well.
As to using a old man design, I'm not. That's a old woman and its a example shown of a silicone skin I aspire to reach that level of realism. But i"m not going for old look like that.As far as using gears vs pulleys, for the Dinah robot I'm using gears. For the Adam, Eve, and Abel robots I'm using no gears. So I'll get to do both approaches and see the pros and cons of each.
>>2863007how'd you get the formal shape right? and why so large
>>2863344 It's the same size as my ribcage. The fiberglass robot its for is meant to look identical to me! I got the shape right by just measuring everything constantly with a ruler and measuring pictures of ribs and doing mathematical calculations to ensure it was the same in proportions to the picture. It was a hard job but satisfying. This was all before I owned a 3d printer.
Why not just lean into the sexbot angle for kickstarter funds? Doesn't have to be so on the nose it violates their ToS. Start there and use profits for passions.
>>2863412 because its not going to be used for that. They won't have genitals even. I'm opposed to sex bots. The notion of sex bots goes against the Bible IMO.
>>2862798Is Adam a deathclaw? I don't think this is lore accurate
>>2862965Well then I believe you should try to finish at least one functional joint with full control instead of pursuing easy shit like making a ribcage in clay. It'll be boring for you if you finish everything and then realise your controller is terrible and you need to change your design. for what I've seen your design lacks a position sensor and you'll never be able to achieve good control without it.>>2862966In the states I've heard you retards can do a PhD and master at the same time, and yeah sorry I didn't realise you had to pay for doing a PhD, that's not a good plan.
>>2863446just cheat for the control system use ros control combined with cascadeur or sonething.as for the joint i found a sphere joint thqt uses mg966r motors. its pretty under powered but if the robot is 2 feet and weights around two kilos it should work for a smaller testing robot.
>>2863454You clearly have no idea about what you are talking about. You can't connect a custom joint using a DC motor to ROS directly. You'll need to create a special ROS package for your joint and implement the desired controller inside your package. I hope OP is more infomed than you
>>2863465okay so cascadeur has an api. The gyroscope module can output quaternions, those are fed to cascadeur and cascadeur sends it back to the motor controller which controls the motor.no idea how ros control works just mrntioned it just in case.
>>2863475Well at least you admit it I respect that. You are talking about higher level controls, I'm talking about how to achieve a 90° angle in the finger when your code asks you to go at 90° we are just talking about different stuff. I believe implementing inverse kinematics using moveit would be much more professional than cascaseur for higher level path planning but OP is far from there anyway that's why I was talking about lower level controls
>>2863490this is the joint i was talking abouthttps://m.youtube.com/watch?v=FeUMAKdcA9Yit coukd be used for the legs. i think the formula was weight of the robot*sin(45)*height of the robot=torque in newton metersi got the torque for the mr966r and it is slightly greayer with a 2 feet robot that weights 2 kg or sothat joint also grants full degrees of movement so thats where the higher level control system comes in.
>>2863432Technically making anything lifelike in the image of humanity is against the Bible. It could easily be construed as an attempt to be god-like. Sounds like you're damned either way. Might as well extend the bots modularity to the groin area. Someone will, if you don't.
>>2863820 that's only in the context of idol worship. If no idol worship is involved, then its fine.
>>2860808Lilith or bust
>>2863832 in most tv shows lilith is the name of a demon so no thanks. no evil robots wanted. Also, I had to look up your reference and supposedly lilith in mythology was adams first wife. That directly goes against the Bible so I reject it as a lie.
niggas will literally do ANYTHING but go outside and talk to women
>>2863971 I'm not building a robot as a RL woman replacement. I'm not MGTOW either. I am happily married. You assume too much.
Here's a sample configuration of how I will mount the motors. I wrap them in football jersey mesh and suture them onto the bone fabric sleeves. The underside of the football jersey mesh is coated in anti-slip carpet paint. It grips very snugly. The motor can also still breath due to the mesh holes. I plan to thread braided solder wick copper through the mesh holes and make contact with the motor and use thermal conductive silicone to seal the contact point and these wires will conduct the heat away from the motor and onto a water cooling pipe.
>>2863831Has nothing to do with idol worship. If you attempt to avoid the uncanny valley, you'll be attempting to make something in the image of our creator. As the creator made us in the image of himself. It's only contextually idol worship if you want to be seen this way. Since you cannot control how others view you, you've no way to determine if what you're doing is blasphemous. Better to embrace the uncanny valley and remove all doubt of your intent.
>>2862962i swear to god if that bitch pushes that poor little robot over 1 more time
>>2864223 there is no rule speaking of making something in the image of our creator. You just added that. Again, if it ain't an idol its fine.
>>2863475>>2863454OP, you're retarded. As far as I understand, you're using a crazy set up of pulleys and motors to imitate muscles. Just controlling muscles to move the knee joint to a given angle is really hard. Muscles come in antagonistic pairs, so both your 'muscles' have to move in coordination. And the relation between angle and muscle contraction is nonlinear too. And if you actually want to do shit the dynamics of your joint matter. You will have to do some very interesting control just to move the arm from one angle to another without it being too wobbly. There ain't a ROS package for that. Plus, low level actuator control needs to run FAST too, meaning it shouldn't run on ROS.
>>2864396Oh I'm not the OP shouldn't have brought up ROS control. Move it seems interesting though, cascadeur is meant for animations.
>>2864420I'm going to be honest with you guys i was hoping to assemble a team to eventually monetize this thing. I don't want this to be hobby. It is not meant to be a hobby in the same way writing a book or making videogames is not really meant to be a hobby cause its too much work and artbyrobot and hanah dev are not interested in making money ff. I'm broke asf right now though so I can't buy more silicone rubber and try to fix the skin. I wouldn't start a kickstarter campaign by myself cause for this project in particular it'd be dumb.I told him that what i bring to the table is that im in thailand so i can get manufacturing going eventually i can also do online marketing. He seems to think that has no value. Its the difference between this thing making something versus nothing.
>>2864422oh wait there's laws against sex toys in thailand i forgot.
>>2864428well I could just hop on over to cambodia or vietnam or something
>>2864396 Cronos by TheRobotStudio on youtube solved every issue you mention. This usage of motors to pull fishing line that act as tendons on joints is not original and is a proven methodology. Your supposing it is not feasible or w/e is born out of ignorance.>>2864422 you're doing sex bots I'm doing chore bots. You want simple limited bots I want vast capable bots. We have different goals although some hardware and software techniques have overlap so I follow your work. In any case, I will not be part of a sex bot company. Not only that, but I have no interest in being part of a robotics startup that isn't offering me 6 figures starting salary and EVEN THEN I'd probably pass. But you trying to found a company with zero budge to even buy silicone is not good. You have to first have a good $50-100k saved up then you can begin. If someone partnered with you, what promises you will stick to it and not get lazy and wander off? Nothing. You think people won't partner with you because of concern of theft. It's not that. It's because nobody can know how hard you will work. You have not proven yourself to be a hard worker. You take long breaks and all the time complain of burnout. And you overestimate deadlines. I tried working with others before and they all were lazy and quit before long. No thanks. I'd have to believe the person was more disciplined than me (proven by them having more money than me) and more hard working than me (proven by them having more money than me) and absolutely dedicated to the goal for a ten year horizon (not someone who just dabbles sometimes and is distracted by a million other things). And EVEN IF I found such a person I'd probably still pass. Also, I would benefit ZERO by 3rd world country cheap labor because my robots are FREE labor so why pay anybody for labor?
>>2864582well i mean if we got say 500 back orders for say a very competitive early bird price of say $2500-$3000 wed surpass our goal which would be... over 1 million dollars. desu it has to be to make this thing take off. but then you have to build 500 robots. no way you could on your own, wed need help. a sex bot might be too controversial either way. id narrow it down to a robot with lidar to walk around the house and be able to pick things up and talk. and thats okay cause since were early we get to set the bar. it looking like hannah devs bot would also be pretty neat... yeah but no super man robot or anything a blank slate bot. it ought to be 5 feet tall ideally or people will raise some eyebrows which is a big deal cause like i said earlier everything scales by the power of 3 as it scales up and it becomes more unbalance. i just think why did they make the asimo 4 feet tall must of been for a reason and asimo did have a control system.
>>2864584oh i am aware that 3d printing is no viable for mass production wed want some good custom injection molds, i dont like how labor intensive the skin is either so thatd would best be tpe injection mold as well. theres some other stuff that wed have to set aside and the legal fees if we want a patent, the electronics would have to be custom made and ordered from pcbway or something like that, etc...i dont know how to do any of those things.
yeah but starting a business making tons of robots is not smart. What you want is a business where you make just a few robots and then those robots make more robots and once you have at least 50 robots only then would you even consider selling a robot. The first robots are your work force.
>>2864584 also 2500 for a robot is a stupid price. A decent sex doll is $6k-8k. It has 1/50th the production challenges and labor and parts cost. So perhaps $400k is more reasonable from the seller perspective and if the buyer can't buy that then they are not getting one.
remember this is 5x harder to build than a car and you have ZERO economies of scale to fall back on like cars do. And don't even start on the AI. That brings massive value.
also you cannot accept kickstarter preorders before you have a factory setup. You'd have to already have the facility and all workers and all parts and a production line FIRST before you even started the kickstarter. So you'd already have $300-500k invested before the campaign can even begin.
also, not even multi-million dollar robotics companies the best in the world could as of today produce 500+ robots to fill a 500 robot order for kickstarter. So what makes you think you can pull that off when you've never even started a factory before. You'd have to recruit workers, rent a large building, install assembly line robotics to assist, order hundreds of thousands of dollars parts and tools. It would be a multi-million dollar enterprise now that I think about it. And you would be paid ZERO for at least 10 years and have to work fulltime for no pay. You think you can do that? 80 hour work weeds for 10 years not any pay.
>>2864605got to offer a deal for those preordering a prototype with a white paper. if its a full bodied robot with no noise that needs an ai the price would be different.>not even large companies can fill 500 robot ordersthe equipment ranges from thousands to tens of thousands at worst. in thailand in surin you can rent a large warehouse for less than a thousand a month, for working hsnds theyre all old broke people living on miserable pensions willing to work minimum wage or teens at $10 a day. if we get +30 people assembling at 12 hours per day we could meet the quota easily along with assistance from mold technicians(13,000 a year) some industrial engineers.(around the same wow wtf) really i hate talking in these terms cause well who am i to say but i need to to make sense of the figures. this is without including the cost of materials which would be bought in bulk but shouldnt be excedingly high. according to my guesstimates the operation would cost around $500,000 in a country like cambodia or vietnam wed get even better deals but since wed keep it pc they might say whatever and i dont have to go anywhere. you know cause of the anti sex toy thing.
>>2864582Fron the short videos I could find from googling it, the arms wobble around like a floppy grandpa dick. And that's without any payloadhttps://www.youtube.com/watch?v=zsqiCyYkfDYHand's gotta be wobbling by about a centimeter or so. That's shit. Can't find any numbers on mean time between failures, but I suspect it's bad.>fishing lineIt's ok to use fishing line for prototyping, just don't expect it to last very long. You should be able to switch to steel cable for longer lasting and stiffer joints easily though. Then again, your wires and what not will probably fail more often. >>2862916>regular maintenanceAnd if MTBF is less than 100,000 hours, then it's not worth it to install robots because the maintenance costs too much.>maintain itself>maintain carIf you can make AI smart enough to do that, then there are a bunch of simpler things it could do that don't involve robots and make a lot of money. Why the fuck aren't you doing that if you really have smart AI? And how will you get high enough resolution touch sensing in the hands so that it could even sense objects as fine as fishing line?
2864704 no, you can't wrap steel cable around pulleys this small it would over stress the steel at the turn. Only synthetic can make such tight bends. So no, you cannot EVER "upgrade" to steel cable. Fishing line has to be the ONLY method and is not just for prototyping. It is the strongest and most pliable solution on earth. There is nothing better today.Cronos wobble is because of rudimentary control software and perhaps also it uses compliant muscles so they have some give and stretch to them intended to make it easier to work with around humans so it doesn't easily hurt somebody. I don't think this elastic muscle setup is needed so would not have it and my control software will not be bare bones proof of concept like Cronos' but be full fledged way more involved and sophisticated so would not be as wobbly. But the wobbly aspect has nothing to do with the viability of having a human musculoskeletal based design they used.If AI smart enough to maintain a robot has other monetizing opportunities, fine, I can look into that when the time comes. But that's a distraction right now. I can't make every product in AI field. Nor do I want to. I'd be more than happy to just have a robot that works hard to make me money and make all my inventions (well many robots, not only one)
>>2864704 from rob's github "3d printed plastic robots offer tremendous precision but significantly lower stiffness, a trait that is true in fact of all the lower cost versions of components in a robot. Here there may be a happy coincidence because the problem with expensive hardware is that it cannot be safely used around people. Industrial robots are made of metal and remarkably stiff gearboxes to produce rapid, highly precise movements - like a robot.But people don't move that way, we are bouncy and capable of feats of strength and precision that far outperform an industrial machine of similar weight and power.So, we know it's possible to learn how to control a bouncy system and we know they're lower cost."
oh yeah one more thing. I can't find touch sensors for finger tips of any resolution really. If anybody can assist there with what to search for. Instead, so far, the plan has been to just use strain gauges which is just a single pad that will measure a single point of touch alone. Now with the help of eye hand coordination, I don't think high resolution touch sensors are needed, but it would be nice anyways its more information to work with in general which is generally useful.
>>2864782>tremendous precision>lower stiffnessLol no. Unless you meant tremendously low precision.>can't be safely used around peopleNot true. Robots like universal robots robot arms and kuka LBR iiwa 7 can be used around people safely despite having stiff metal gearboxes. Rapid movement means you can stop robots faster too if you sense a problem>we are bouncyProve it. And 'bouncy' ain't a measurable property>we know it's possible to learn how to control a bouncy systemEver heard of rethink robotics? They sold robots that had springs between the joints for safety and force sensing. They went out of business because their robots operated too slowly. The decreased stiffness meant they had to operate slower. That being said, if your robot doesn't have joint level torque sensing, it's shit>>2864800>can't find touch sensors for fingertipsGet rekt. You can get high enough resolution sensor with gelsight. Something like a gelsight sensor shouldn't be too hard to diy. It's just some silicone lit by LEDs with a camera looking at it.>don't think high resolution touch sensors are neededPerformance for manipulation goes way the fuck down without touch informationhttps://www.youtube.com/watch?v=HH6QD0MgqDQIt's almost certainly needed for robustly manipulating stuff IRL.
>>2864813 surgeons use pliers and scalpels with insane precision and these tools have no touch sensors. I am confident touch sensing is not needed but is only helpful. As to LEDs with camera looking at them, how is that a touch sensor? A camera assigned to each fingertip? huh? sounds absurd. Is this camera outside the robot's body looking at a fingertip? makes no sense to me.You are telling me to prove stuff and whatnot when I posted a quote from a world famous roboticist TheRobotStudio. Take it up with him if you disagree. He is world renown with multiple PHDs etc.
>>2864781no again the ai should be stuff that already exists. we know talking already exists(ive talked to my robot head), we know lidar already exists(roomba) we know picking up objects kind of exists. these are reallistic goals...still quiet hard but within the realm of possibility.if i felt confident about advanced ai id have robots assembling other robots like you said.and i believe that people would be okay with that for the time being. just like theyre okay with a raspberty pi that cant do much of anything without being programmed.
>>2863359oh you a big dude.
Here's my CAD model of my motor controller design.
im unsure if OP is some autistic robotic genius or some schizo religious creep bringing the devil seedling into this world
>>2864891>surgeons use pliers and scalpels with insane precisionYou need touch sensing to use them reliably. If you can't sense them, you can't tell if they're slipping or get an estimate of forces on the scalpel.>how is that a touch sensorhttps://people.csail.mit.edu/kimo/gelsight/You point a camera at some soft silicone. Several different colored LEDs point at the silicone sheet and the different colors make deformation of the silicone stand out. It's high resolution and cheap. Nothing better exists at the moment>I posted a quoteThat proves nothing.
>>2865010 how do you point a camera at the robot's FINGERTIPS? are you suggesting to fit a camera INSIDE a robot finger? I'm confused.
>>2865159>are you suggesting to fit a camera INSIDE a robot finger?yeah, you put cameras inside the robot's finger like appendages.This details how grippers with gelsight sensors can do cable manipulation, which is really, really, hard for robots to dohttps://gelsight.csail.mit.edu/cable/This is a cheaper, but pretty capable design for a gripper with touch sensing that came out recently: https://arxiv.org/abs/2408.00610
no, the camera cannot fit inside a finger and would be instantly crushed once hard work started.
>>2865202>can't fit inside the fingertoo bad, it's the only option for high resolution touch sensing at the moment. So either build different fingers or invent a new touch sensor. But hey, I'm just saying you can get SOTA touch sensing performance with something you can build in your garage. >instantly crushed once hard work started.I don't think you understand how the gelsight sensor works. Did you even look at the pic in >>2865010The camera's in a box. Your robot ain't gonna be doing hard work with plastic fingers and cable drives made of fishing line.
>>2865279 yes it will. these plastic fingers are as strong as human bones IMO
>>2865414Put the cameras in the bones so they don't get crushed
>>2865417 ok I'll just scale up the robot to be the size of a skyscraper so that I can fit a camera in the distal fingertip without compromising the integrity of the fingertip bone structure *sarcasm*
Here is my CAD design for a bare bones arduino mega microcontroller to use for actuating mosfets for actuating the motors and for reading in strain gauge and potentiometer readings.
Here's a novel way I figured out to attach flat flex cable to a atmega2560 chip to make the smallest possible microcontroller setup! It's not done of course lacking the other components form the cad but is a start. This was hand soldered successfully.
I have opted to use EMEET usb speakerphone M0 as both my listening and speaking device for communications with the robot. I will install it inside the skull.
>>2864977>>2865466>>2866025I almost puked.I mean, you need something small but reliable, right?Why not use a fucking PCB (even 1mm ones) and use a smaller formfactor for your microcontroler!?This method of construction is really small indeed, but you will need to encase it in epoxy (rendering it almost impossible to debug/repair).If you persist in your deadbug assembly, remember to separate easily your modules, so that you only have a module to replace and not everything.Also, your thermals for your motor controllers will be horrible.If I miss something, sorry, I tried to read everything but I had to skip.Also, even if I tell you that, your project is really awesome.Fucking creepy, but really, it's super interesting.
>>2866135 I am considering using flat flex pcbs to solder down components and overlay these components onto the mosfets and microcontroller chips as layers ABOVE the chip just like my CAD shows. For now I'm using nickel strips to deadbug traces. The main goal though is to stack components on top of chips or mosfets to cut down length and width in exchange for a tiny height increase that will still be shorter than the tallest single component on a pcb anyways. So even the height of a traditional spread out pcb is taller than this method. So I have cut down all size with these designs - length, width, AND height. This makes these methods ideal.As far as thermal regulation, I intend to run the drain of each mosfet through thermally conductive copper wick over to a water cooling pipe for rapid cooling of each mosfet. So there will not be cooling issues.Also, potting electronics after they are tested and working is a viable apporach IMO although I can pot them while leaving test leads exposed outside the potting to still be able to debug them while potted.
This is a little sliding adjustable ceiling mount I created so that I can hang my robot arm above my office chair and have it descend down onto my lap to easily work on it for my convenience. It has been very useful.
Here's a index finger rigged with fishing line and TPFE guidance tubing detail shot
artificial lungs design drawing
>>2867102>artificial lungsYou, what!?Does it have a real purpose or only to be more agreable when you... cuddle?
>>2867224 I already explained its not a sex bot. That said, the lungs are for drawing in air from nose/mouth and distributing that fresh relatively cool air throughout the body. Vent pipes will allow for hot exhaust to exit the body as this cool air displaces it. This creates good air circulation to keep the robot from overheating inside its silicone skin suit.
>>2867368>Not a sexbotSorry, I didn't read it. Okay.>breathing for ventilationWell, if you have a silicon layer all around your robot, I understand now.Good luck with this project.
Lungs division of air distribution idea
>>2860808Nice hardware, I have a suggestion on how to make the software needed for bipedal movement. Most labs that I know of have the robot try to "learn" through a generative system. The current setup has a robot sense and react to obstacles and objects. However this vastly overcomplicates simple tasks. To even hit a soccer ball a robot must use active 360 scanning to get an accurate view of it's surroundings, generate a possible 3d layout of it's positioning in the world. Then differential shading to detect the contrast of the ball in comparison to the field after which it must calculate the best distance and rout to the ball ext. All while adjusting thousands of electric motors in harmony to ensure it even stays upright. This takes an extremely high amount of processing power and is prone to frequent and costly error.Enter a "Pinocchio" rigging system. Rather than create a ton of external commands, you can rig a bipedal machine to use pre existing motion tracked data. Then use human movements as training data to generate a more complex AI. At first, all you'd need is a cheap motion tracking system.If you wanted to go all in there are suits designed to capture and map models efficiently.https://www.movella.com/products/motion-capture/xsens-mvn-linkAfter you establish the tech with a few demo's, you could focus on capturing movement data for neural net processing.>I thought about defrauding the DoD with said system. Spooking them with robot holding gun (with the mirror data being that of someone hitting a bullseye on the same arena) but realized that might be illegal somehow. Have fun.
>>2868061 good ideas. Although the calculations for the soccer ball example are necessary for it to do that stuff. However, it using motion tracking data is also good. I don't think it's either or but both. However, I don't want to develop a motion tracking suit which is a waste of time if you already intend to have it capable of motion tracking through vision alone. You don't need balls on a suit that are easy for a camera to track if you just have sophisticated computer vision that can make everything out without the need for a fancy suit nor any type of fancy glove or w/e. My goal would be for the robot to learn human motion for all manner of activities just by watching youtube videos it looks up for that activity and learning the same way we learn in this way. We just watch somebody do the thing on yotuube then we go do it, emulating what we saw. There is zero reason a robot can't do this. And it is exandible with zero work on my part doing any of the things. When I tell the robot a task, it goes and watches videos on that task and learns what to do based on the videos and then goes and does it. That skips all the fancy suits and w/e which are all just crutches used due to unwillingness to develop quality computer vision and AI to train directly from videos on youtube. Generative learning and reinforcement learned etc would all just pair along with the youtube video pre-learning to fill the gaps and fine tune.
An idea for a mini freon air condition built into the lungs.
>>2862915>Mine won't. Mine won't be like other's.Why do you believe this? What do you have that they don't? Robots seem to attract pie in the sky designers like no other field, so unless you have something up your sleeve the chances are that's you as well.Your bone casts are cool though I'm not sure if they'll be strong enough, the tendon drive is neat though I'm not sure how practical it's going to be, but it sounds like you have almost no understanding of how code or AI works.What language will you be using for this project? What AI model will you use as your foundation for this? I have suggestions if you don't know.
>>2868486Why don't you just use fans?
>>2868061That's nice, why don't you test if it works using a virtual robot in simulation?
>>2868486What's even going on here? The condenser's the hot side, it won't collect water. The use of a bladder here is absolutely bizarre. And do you really need refrigeration? Just moving more air is probably effective>>2867368You're pumping air through the body? Why make a really complicated system with air returning when you can just pump cold air in with a fan and vent it at the hot parts?
>>2868974> why don't you test if it works using a virtual robot in simulation?Getting a digital 3d puppet rigged to be weaponized has already been done virtually. https://store.steampowered.com/app/823500/BONEWORKS/You could say the same for advanced VRChat full body tracking. A basic rigging system already exists that you could affordably adapt with the right system. The issue is getting the money and hardware for it. I've had some experience in freeCAD and Solidworks, but not enough money to do something reliable on my own. I've had some chats with other defense contractors about it, you really need a physical product to prove that your pitch isn't vaporware if you want to pull a contract. Plus the contract has to be made before it's filled. TLDR I'm lazy. If I ever had an extra 350k lying about I could make it happen. But I'd rather do something on a smaller budget like mint my own cologne instead. That or make the project for under 10k and doing so isn't that appealing.
>>2868951 The thing I have that they don't is time. They are rushing to push demos and minimum code to only display basic capability or possibilities and ship robot to customer. They don't have the luxury of fleshing out a proper AI implementation. Nor do they have a incentive to do so.I'll be using C++ and a hybrid AI of rules based AI and machine learning combined.>>2868952 The ENTIRE robot will be enclosed in silicone skin and then clothing over that. There is no place for a fan at all. And the inside will be so congested with parts the fan would never reach every area of the human body shape. You'd need fans distributed all over the body and then it could not wear clothes which would block the fans. It could not have hair if a fan comes out back of head. Fans are no goes then.
>>2869009>The thing I have that they don't is time. They are rushing to push demos and minimum code to only display basic capability or possibilities and ship robot to customer. They don't have the luxury of fleshing out a proper AI implementation. Nor do they have a incentive to do so.That just isn't true. Meta is a perfect counter example to this. They're working in foundational models just in the belief that they will later be a building block in something useful. They aren't working to deadline at all really because they're subsidised by the rest of the company. That's why they come out with some cool robotics AI models that you should check out.>I'll be using C++OK nice>a hybrid AI of rules based AI and machine learning combined.That gives so little detail that it isn't really an answer at all. Which foundational model will you be finetuning? Or do you think you will be training something from scratch?>There is no place for a fan at all.If there's space for lungs, there's space for a fan. A fan with higher CFM. There's a reason we use fans for everything rather than bellows. If a fan won't reach it, neither will bellows, because it's just pressurised air after it leaves the pump.>then it could not wear clothes which would block the fans.Your lungs will need an exhaust too, otherwise they will just drive the air back out through the intake without meaningfully circulating. It's not something you can directly parallel with blood oxygen distribution because that has a secondary pump after dissolution - the heart.
>>2869003Here you go, here's a rigged humanoid robot simulator that's pretty much ready to go. Instructions to get it set up should be at the roothttps://github.com/isaac-sim/IsaacGymEnvs/blob/main/isaacgymenvs/tasks/humanoid.pyAll you need is motion capture data right? You can just like pirate that. I'm sure there's publically available mocap data too.>need a physical productIf you can't make your robot work in simulation, it's probably not gonna work IRL.>>2868106>learn human motion from watching youtubeCool, why don't you demonstrate doing that with a virtual robot model like the above?
>>2869026 How is rules based AI combined with machine learning too little info? Both are defined and understood. That's alot of info IMO. It's funny because when I write a long explanation people complain but when I write short they complain too. It's all from scratch. I wouldn't touch anybody elses models with a 10ft pole. I have to understand it inside and out. The way other people do their AI reflects how they think a brain should work. If you make your own AI and want to push the whole field forward, you better start from the ground up. I mean not a SINGLE LIBRARY, NOTHING can be off the shelf. You start with int main() and go from there ON YOUR OWN. ZERO help. And it is NOT going to be LLM related AT ALL. I hate that entire approach to AI. It's a joke to me. Your statement on fans vs bellows is ignorant. A bellows could suck air from 50ft away and blow it with force into a confined space. A fan has to suck immediately air in its direct vicinity. If the fan was in the middle of a silicone suit, it would recirculate the same hot air in the chest achieving NOTHING. ZERO cool air would enter the robot AT ALL. Let me explain further: the lungs will open, creating a vacuum effect which will draw in air through tubing leading to the nose/mouth holes. So cool air outside the robot will be sucked into the robot. It will then shut off the inlet air hole with a valve. It will then open another valve inside the robot that and exhale the air in lungs, into a series of tubes that send it to each region of the robot's extremities to reach all places. Another series of exhaust tubes will collect the hot air being displaced by this cool air and route this hot air out of the nose/mouth region, completing the circuit. Rinse and repeat.
>>2869030 my robot project is taking turns a few months at AI dev then a few months at hardware dev. Back and forth. It's not just pick one. They both need meaningful progress as the years go by. That said, the AI is coming along great and getting it learning by watching youtube is not here yet. I have to first create the computer vision system and custom web browser so he can pipe into many youtube videos and other websites simulataneously, feeding on their information without a UI, just in bytes.
>>2869076>How is rules based AI combined with machine learning too little info?One of the simplest things you could define about your machine learning component is what form you expect your input and output sequences to take. What modes will your model be working in?For your rules based AI, how will your robot decide what rules to add? You said you'd tell it to add a rule, but how will it interpret you telling it and turn that into a rule? How will these rules be interpreted and inform the robot on how to path?>It's all from scratch.The amount of compute you'll need for something capable is immense.>And it is NOT going to be LLM related AT ALL.OK. If you don't want your robot to use language then I agree it wouldn't need an LLM (though it might need language to interpret your rule changes, depending on how you plan on implementing that).Do you want your robot to talk?>A bellows could suck air from 50ft away and blow it with force into a confined space."from 50ft away" doesn't really make much sense, because you're always "transmitting" (so to speak) that negative pressure through the air at the location of the bellows.If you force it in to a confined space you're talking about pressurising, which won't really circulate very well.>If the fan was in the middle of a silicone suit, it would recirculate the same hot air in the chestCertainly. What I would suggest is either a one way system with an intake and exhaust both in the mouth, or a reversable drive fan and valves if you insist on the pressurising approach.>Another series of exhaust tubes will collect the hot air being displaced by this cool air and route this hot air out of the nose/mouth region, completing the circuit.Ah we agree on the one way system and you're not actually talking about pressurising. Yeah that's the only way that makes sense, but there's still no reason to drive this with bellows when a fan has much better CFM.Are you anticipating a lot of overheating in this robot?
>>2869084 >what form you expect your input and output sequences to take. not sure what you mean by this. It will learn from conversations, articles, video watching, basically every way humans do as input. Output will be taking notes and creating rules and definitions in its knowledge database. This will result in more advanced behavior and speech.maybe create a more specific question or example.>What modes will your model be working in?not sure what you mean by this. maybe create a more specific question or example.>For your rules based AI, how will your robot decide what rules to add?anything it learns becomes a rule unless another rule supercedes it then it chooses between the two. It also has to filter past morality and whatnot before it accepts a new rule in. Maybe create a more specific question or example for this too. >You said you'd tell it to add a rule, but how will it interpret you telling it and turn that into a rule? well the same way we do. If I tell my daughter don't touch the stove it's hot, she internalizes a new rule. That's one example of millions. Or a physical rule: it sees how I pick up a pen or hold a pen to write. It would create a series of rules on manual manipulation and wielding of a pen off that observation.>How will these rules be interpreted and inform the robot on how to path?not sure what you mean could you give a more specific example or question? Some rules are physical, some are conversational, some are logical. Which subset are you interested in?
>>2869084>It's all from scratch.>The amount of compute you'll need for something capable is immense.not true. I already told you I'm not doing LLM technology. Large compute is due to the inefficient and stupid design of LLMs. They are trash.>And it is NOT going to be LLM related AT ALL.>OK. If you don't want your robot to use language then I agree it wouldn't need an LLM (though it might need language to interpret your rule changes, depending on >how you plan on implementing that).you seem to think LLMs are the only way to have language. That's not true. Rules based AI can have language too.>Do you want your robot to talk? yes.>A bellows could suck air from 50ft away and blow it with force into a confined space.>"from 50ft away" doesn't really make much sense, because you're always "transmitting" (so to speak) that negative pressure through the air at the location of the bellows.No clue what you mean here. When you open a accordian, air rushes into it because opening it creates a vacuum that requires air to fill it. The location of that volume of air can be far away and it can come by way of tubing, but it will come and it will come fast. And it will travel any distance. It does not need to be locally sourced. In my case it is sourced from outside the nose in the room and comes in by nostrils.
>>2869084>If you force it in to a confined space you're talking about pressurising, which won't really circulate very well.you didn't read very carefully. The confined spaces all have tubes running from then all the way to the nose and out the nose. So when you blow forced air into them, the hot air that was previously filling said space gets ejected forcefully by the new inrush of cool air and that ejection force due to pressurization goes out the vent tubes and out the nose. This means 100% circulation is achieved. You need the incoming air to be piped lower in the cavity and the vent tube entrypoint to be piped higher in the cavity since heat rises.>but there's still no reason to drive this with bellows when a fan has much better CFM.no, fan's would not move air through tubing almost at all. Too much turbulence. Not enough suction. There's a reason why bilge pumps and self priming pumps use suction cups to suck air. A bellows is just a large air sucker. The fan would NOT suck air. Try it, get a garden hose, attach it to a fan, and see how much air is moved. You'll note that none is moved more or less. But then use a billows. It will be undetered.>Are you anticipating a lot of overheating in this robot?not with my cooling systems planned. But shy of a series of cooling systems, multiple types too, it WILL overheat. This is countless motors and hot electronics trapped in a rubber suit. It will be like an oven. The cooling systems have to be far beyond anything ever done or dealt with in most other applications. Have to be extensive and SUPER capable. This is make or break for the entire project.
Let me clarify one thing: a bellows sucks in one tube when opening, then blows air out same tube when shutting. But my bellows will have TWO tubes entering it. Only ONE tube will be open at any given time. So then it will suck air in, close the inlet tube off, then blow air out the outlet tube then shut that tube off and open the inlet tube again for the next in suck. This way it acts more like an inline duct fan except better IMO since inline duct fans don't work unless they are very large and blowing through large pipes like 6"+. They don't work in smaller scale or constricted piping like a bellows can.
>>2869093>not sure what you mean by this. It will learn from conversations, articles, video watching, basically every way humans do as input. Output will be taking notes and creating rules and definitions in its knowledge database. This will result in more advanced behavior and speech.That answer makes no sense, but I know you're still learning about AI so you'll get to grips with why at is I'm sure.When you design a model, you have to be consistent with the format or "mode" of information you provide it as the input vector, and it will provide you a consistent format or "mode" of output for its output vector.What that means in the simplest terms is you can't have a model that accepts text from articles in one iteration and video frames in another.If you didn't do this, you'd be asking it to learn relations between tokens that aren't consistent, as each step of gradient descent will be pulling it in a different direction, because the underlying (hypothetical) graph for understanding that topic and minimising that loss function changes every time you change that format.>well the same way we do. If I tell my daughter don't touch the stove it's hot, she internalizes a new rule. That's one example of millions. Or a physical rule: it sees how I pick up a pen or hold a pen to write. It would create a series of rules on manual manipulation and wielding of a pen off that observation.That's the user interface you envision: audio -> "rule". What I'm asking here is how do you plan on making that leap?>not sure what you mean could you give a more specific example or question? Some rules are physical, some are conversational, some are logical. Which subset are you interested in?The easiest way for us to understand each other here would be for you to write a draft of how you expect a rule to look, as it sits in the code. Doesn't have to be precise, just mock up a rule so I can get an idea of the general format you expect it to take.
>>2869107 no, I'm not still learning AI. I threw all the industry history and tactics in the trash heap of history and rolling my own. I don't have to learn what others have done. I am paving my own way entirely. I know exactly what I'm doing and how I'm doing it. Nothing I said makes no sense. If you have a specific question ask it.You speak of designing a model. By this I assume you refer to LLMs. I am not creating an LLM. You speak of how other people's AI works and ignore that I'm not making mine like theirs. Whatever imagined constraints you are putting on my AI and it needing to have the same data type or input or w/e is non-applicable for the AI I am developing.You speak of tokens. I'm not doing token based AI in the same way as LLMs. You really are not listening that my AI has nothing to do with LLMs or tokens or models. Sure some similar concepts might overlap, but you really are just envisioning I'm rolling a new LLM and I'm not. I'm not doing weighting nor gradient descent nor any math on the input nor tweaking weights or anything. I'm not even in the same vicinity as that AI in my plans.
>>2869117>If you have a specific question ask it.OK. Here is my question. If your AI does not work the same way as any existing AI, how does it work?I'm not asking about the user experience, I'm asking for a techincal explanation of what you are planning on doing to fill the roles normally filled by neural networks.I'm also not asking about the rules based part. I'm asking about the part you referred to as machine learning.
>>2869107 A sample rule: when you start a conversation with someone it's generally good to greet them first. Then it would have a list of possible greetings and a rule to choose a greeting at random from said list and another rule to not greet with the same greeting as you used last time from that list. If it heard someone use a new greeting word it could add that new word to its list of greetings options in its knowledge database. How could the greeting rules come to be? Well I can code them myself, or verbally give the rule like so: my robot and I begin a conversation and I note it never greets me. I tell the robot you should greet someone when you begin a conversation with them. It would then add that rule to its list of rules and when the next conversation begins, it would greet the person. So the rule was created from verbal instruction and the robot added the rule itself.Ok, I think I see part of our disconnect. I am using machine learning as a term to describe the machine learning, not the specific formal definition of machine learning and all that this entails. So that is why you misunderstood me. So going back, I'm using rules based AI and that will be a learning rules based AI and it will be symbolic AI. This article explains the difference between machine learning and symbolic AI. I'm doing the latter. https://www.inbenta.com/articles/symbolic-ai-vs-machine-learning-in-natural-language-processing/#:~:text=Symbolic%20Artificial%20Intelligence%2C%20also%20known,box'%20created%20by%20machine%20learning.
I had a guy on a forum go through my AI videos and he summarized his view of what I'm doing in a kind of neat way I guess. Maybe you'd understand more from it (he was critical and disapproving, but seemed to maybe describe my progress in a way others grasp better than I can?) He wrote:The descriptions that OP has written here are actually a pretty accurate and complete description of his fundamental approach in the videos, if you want to skip watching. I just think people here think he's leaving out the "secret sauce" in the description based on his claims, when in fact there just isn't any sauce there and the claims aren't legitimate. He writes rules (and vocabulary, and linguistics, and state, and religious tenets, and... everything else) in a folder full of text files with kinda semi-structured natural language with a bit of C++ like syntax sprinkled on top. And then the AI is "just"(!!!) an interpreter that needs to be able to load/execute/update them as needed. Based on what was shown, so far it reads a number of the files into memory and has a parser (which despite being fundamentally flawed in some design choices and still buggy) that can at least sometimes recognize if and while statements. Didn't see any indication there's anything written yet beyond the parser.All of this done in his own unique "style" of programming... which is essentially making the program one giant ad hoc state machine with the state spread across thousands of global boolean flags likeCode: [Select]bool parsingConversationsProgrammingFileArrayLineWaitingForSeeingIfItsAnIfStatement = false;and then the vast majority of the control flow being one giant loop full of simple if statements that test/set/clear the flags to dispatch, all written in a single file with one 250k line main function.
>>2869121OK. I'm glad we cleared up that misunderstanding.That approach will mean any time you say anything to it, you'll get one of two outcomes:- It won't have encountered the input, so you'll have to write the response to your own question. Might as well just write to yourself in a text file at that point.- It will have encountered the input, and it will give you your own words back to you. Why would I want that versus just writing my thoughts in a text file?Do you see that this is a low version of cleverbot by the way? And cleverbot wasn't all that convincing or fun to begin with.>>2869124Yes. I undertand what he's saying. I'm sorry to tell you that it's actually a pretty damning description of your ability to code and he highlights the same problem I can see with your work, which is that you're brushing the interpreter component under the rug, but it's probably the most complex part of the whole work, and ultimately what you will find to be impossible with neural networks.He's written that politely, but it's not positive or neat what he's saying there. Do you understand it all?
>>2860808What's the power source for this?
>>2869127no that's not true. Like if it has the more general rules it can give more general outputs. like if asked a question, give an answer. it has to decipher and break down the question. so if I say "what color is grass?" and I never trained it an answer to that, it would look it up in its knowledge database which it will fill all the time or look up the answer online to get the answer color. Anyways, the question what color is grass can be broken down by it to say okay, this is a question, he wants to know the color of grass. I have to respond with a color. It then finds out the color and responds with a response created by forming a sentence that is not hard coded like "the color of grass is green". It would know to say the color because you said what color. it would know to say "of grass" since you asked the color to pertain to grass. It would know to say "green" because it had this knowledge in a file that says "grass color = green" or w/e. And so in this way it could answer not in a hard coded way. Also, lets say it didn't know and didn't look online. Lets say it just said I don't know - another valid response. I could tell it "grass is green". It would then store that attribute of grass. I would not tell it what sentence to say. I would just tell it grass is green and it would dynamically form sentences with nouns, adj, verbs, etc using rules of english language for sentence structure, appropriate nouns for the sentence, and looking up the answer. So you assumed I'm going in a way I'm not going with this. You assumed no dynamic sentence formation nor dynamic learning but not true.
>>2869127 And when it reads books or watches videos etc it populates its knowledge in its knowledge database which it pulls from to answer questions or make statements. So you suggesting I made all the rules also twists what I have already claimed in this thread. Expert systems sometimes have every rule hand coded by the developer or "expert". I already told you my AI is a hybrid. The robot's inputs are not limited to the rules I make you see. You assumed they were falsely even though I already told you it would learn outside of me as the source.>>2869132it will run on 18650 lithium cells welded together in the abdomen area and will be able to charge and power itself off AC source as well at times when it is plugged in, but will use batteries when its power needs exceed the AC provision even when plugged in. Also, importantly, it will have hot swappable backpack mounted large battery packs to supplement its abdominal one. This way it can have 3-5 backpack sized battery packs charging at any given time so it never runs out of juice and can run 24/7 without interruption other than to walk over and swap out its backpack sometimes.
>>2869127 I looked up cleverbot with chatgpt and learned a bit about it. As far as I can tell from what little I know, my bot is not at all that simple. It will, unlike cleverbot, have rules that go so deep as to consider context and have a lot of depth. Like detecting sarcasm and naunce and various other things and have a vast array of language etiquette rules and conversation pacing awareness and ideas of when someone is being silly or humorous etc. Just because it is rules based you assume it is surface level and will have just a few rules. If you add enough rules, it can pass any touring test.You say the interpreter I'm brushing under the rug. I disagree. Lets dig into that if you are interested give some scenarios and specific questions and I can break things down for you on that. You say I will find things impossible with neural networks. Once again, I'm not using neural networks. I already explained this to you when I said I'm not using anything to do with math or weights. Neural networks is both of those. I am steering very far clear of those. Symbolic AI is the opposite of that yes?I did not find his post positive nor polite, no. But I thought it seemed clear in some ways just on how he said it. The issues he brings up where he finds things to be done wrong or not yet solved is where I disagree and that is a nice starting point for discussion and further prying me for details if you care to know. His assessment on those is subjective and based on ignorance. I really don't think he fully understands nor spent anywhere near enough time watching my AI videos to speak on. He skimmed them surely. He has only a surface level glimpse into what I"m doing and my plans. He doesn't get it.
>>2869144they tried to do ai without neural networks for over 60 years bros.
>>2869166 I can't speak on what they did or how hard they really tried. I know very little about what they were up to. However, I know I've already solved AI and it's very simple and anybody can do it if I just walk them through it which I will because I'm not greedy. Ask away.
>>2869144Your point about sarcasm highlights a problem with the approach you're talking about here >>2869139Every time you say anything to it you'll have to teach it some specificity and so far you've told me only about the absolute simplest class of examples, and that forum you were on says even that is implemented in a flawed way.I assume you haven't written your sarcsam detector, but could you provide some pseudocode on how you expect that to work?Also, my mistake, I meant to write "without neural networks".What is your aversion to neural networks anyway? I'm hoping you understand that they are an automated way of teaching and encoding these rules. They pretty much do exactly what you're after. If you understood how they work you'd see that.>His assessment on those is subjective and based on ignorance.His assessment is pretty plainly one of those who knows how to code reviewing the work of someone who doesn't.This discussion is getting to the point where you're just being wilfully ignorant of what you don't yet know. Someone who writes a quarter million lines of poorly structured state machine and no code organisation isn't at a level of experience where they can decide if there work isn't being fully understood, and it's just arrogance and pride to think you must know better. Creating AGI and self-replicating robots was already going to be an impossible task, but it's even more impossible if you aren't even willing to learn. I'm just going to let your thing run its course and you'll fail just like robotwaifutechnician. That's the only thing that seems to have a chance of changing your mind, and even then you'll probably just double down and go even harder on a tried and tested dead-end idea.That's what Pride does.
>>2869226 >and that forum you were on says even that is implemented in a flawed way.Well it's not. And note that he never elaborates how it's flawed. Nobody knows what he's talking about and he never backed it up by pointing out the flaw. So that part can be ignored. I may very well have far more programming experience than that guy. He's a random.As far as pseudocode on sarcasm detection, I will just attempt to come up with sarcasm examples and extract how I personally detect sarcasm then quantify that and define how I do that and then write that into a set of rules that can be generalized. That applies to all challenges for this project. Reproducing the rules we all follow to determine all things. That pseudocode then would be a big project I'm not willing to do just for a quick 4chan back and forth. Neural networks are trash. That's why I won't use them.If they were good, we'd have some decent robots but they suck and we don't.>and no code organizationyou made this up and it's a lie. My code is extremely organized and structured. So you are just making stuff up out of ignorance now, fabricating it.He doesn't understand my code. You side with him in your folly knowing nothing about the code other than it's a state machine and uses lots of booleans. You don't know how it's structured beyond some very surface level observations. You aren't qualified to make any of the further assessments you are making here not knowing much.>you aren't even willing to learnYou're right I'm not willing to learn trash techniques like object oriented programming or widespread garbage AI techniques that don't work. I'd rather innovate things that actually are capable of working as intended.
>>2869226so anyways, you are here falsely accusing me of pride. That's to your shame. I am confident and you are conflating that with pride. I have succeeded at all manner of things I've put my heart into all of my life. I simply don't fail at anything I put my mind to. This entire project is easy to me. >fail just like robowaifutechnicianwhat does his project have to do with mine? Also, until he gives up he hasn't failed. He's having financial struggles unrelated to this stuff. As long as he settles those matters and comes back to robotics he can pick up where he left off and keep progressing.
>>2869293>>2869294Alright retard. Have fun.
>>2867951>quite pungent my dear
I would like to add that going immediately to sarcasm pseudocode request is rather foolish IMO. Sarcasm detection is even hard for humans. This would be the highest level of AI development, the most nuanced, the most challenging, and the pinnacle of the field. Even Data on Star Trek could not detect humor or sarcasm perfectly if at all. That was a major topic of the show. And he is the example of basically the most impressive possible humanoid robot science fiction could come up with. So to request pseudocode for sarcasm detection is done in bad faith. There would be countless layers upon layers of foundational code establishing countless precepts to even work your way up to sarcasm detection. That would be like the top tip of a pyramid. But we are still working on gradually fleshing out the very bottom row of the pyramid upon which everything else is built. The criticism that my examples have been very simple then is also folly. It is prudent to focus on simple and easy to understand concepts for the AI and how it will gain new rules in simple ways and be able to store the rules and then later retrieve and execute on the rules going forward, proving gain of behavior complexity without any hard coding required. Then you build on this to gradually more complex situations which are largely built on the first simpler rules. And you gradually turn the knob of complexity. But this guy is saying no, lets just jump to 20 years in the future and tell me exactly how it does the very hardest thing in language. It's absurd. It really shows what a hard heart and lack of genuine interest this 4chan user has in genuine diologue or understanding of this process or this project. It is pathetic really. He's just a hard hearted and foolish little man who thinks he knows it all. Pathetic.
Lungs detailed further an alternate design from the freon one.
Ice cubes in lungs concept drawing
So the ice cubes idea is amazing I feel. The robot, if about to embark on a strenuous task, will get a cup of ice water with mainly ice cubes filling it and swallow that and its internal liquid cooling and air cooling systems will interact with the ice water to cool down the whole robot. The ice is an inexhaustible supply of cooling available anywhere you go in modern society. It will always have ready access. It will pee out the water once it gets warm and go get more ice water as needed.
what is the energy source/storage?100 lbs of batteries?what its the expected power draw and runtime before charging?
>>2870153 yes 100s of lithium batteries including one board ones and hot swappable battery packs worn like backpacks. The hot swappable backpack solution means 100% uptime. When batteries get low in teh backpack, you walk over to a fresh one and swap it in. Never run out this way. Never down for charging this way no matter what. This is how construction workers do it with their power tools. We had a guy building our deck this year and he had like 10 batteries charging in our backyard every night. They powered all his tools for the whole day. He never had to wait for charging. Same principle for the robot.
>>2870303how many brushless dc motors are you usingwhat is their max and average power consumption in volts and ampswhat is the expected runtime for normal use
>>2867368it'd be more efficient to pump a fluid around in smaller pipes and have a larger air-to-fluid heat exchangerless air pressure needed to function well, and smaller pipes would work wellonly negative would be needing a pump
>>2870593 3-400 motors. the consumption would depend on how many are active at any given time and how strenuous the task is. this huge variance in consumption then makes giving you an average impossible. it would depend on the activities in question and how long the strenuous or athletic parts of that activity last and how much rest time in between sets. Same goes for runtime. And as I said, runtime is irrelevant and 100% uptime due to hot swappable backpacks giving infinite charge.>>2870862 I'm doing liquid cooling and air cooling for redundancy and added cooling help. So it's not either/or, it's both.
Just wanted to say this is one of the most fascinating projects I've encountered online in quite some time, not just on diy. I'm curious, I know you plan on letting Adam code himself after you bootstrap him both from a hardware and software level, but have you given much thought to an ethical and moral core from which Adam will proceed? Fundamental assumptions and narratives about it's shape, it's story, it's identity? Or are you planning on the LLM to do heavy lifting in that regard? My background is largely in psychological development, pastoral care and treating childhood traumas, so when I see the emergence of a new life like Adam, I see how the shape of his existence can come to shape his experience and growth. Being half-built and asked to complete your own body is such a profound experience for a newborn, the story of that will have so many repercussions on Adam's personality. It's brilliant in a way, because we also have to shape ourselves growing up, but this visceral interpretation of it can wind up creating a vastly different temperament than one conducive to the typical human activities you imagine it performing.
>>2870942 I don't agree with this. There is no such thing as a temperment unless I code it to have a temperment. It is not living. Everything it "thinks" is just artificial, it can't actually think, it can't actually feel, it is just code executing. Not real consciousness, nothing like a human. It's a mere machine. It's as smart as a toaster and that applies to all AI. That is all AI can ever be. It's called "artficial" intelligence for a reason. You are talking like it is real intelligence and a new life form. That is science fiction and is impossible and is actually dangerous to even suggest. The fact people are believing this is sad to me. If I program a simple program that just says "I'm sad" and turns off, now we've got people pointing eyes wide and saying "it's alive!" "see there, real emotion" when it literally just read from a voice prompt I gave it. Heck I could record my voice saying I'm sad and hit play and people will say "look, the machine itself is alive!" when it was just me playing a voice recording I made earlier. This is literally exactly what is happening with these robots in a more complex way but amounts to the exact same thing.So any morals it has is not real convictions, just rules I put in for it. But yes, it will have a morality system based on the Bible. I'm not going to give it a fictional backstory. I will just let it "know" straight up that its a machine made by me. It's not real. I'm not going to put lies in it like "you are a living being and blah blah was how you came to be" etc. Also it will not be using a LLM. It will have a expert systems AI with the ability to be its own expert in part. Coding itself in part.
>>2870951The way you spoke about it you certainly could have fooled me, sounded like you were essentially trying to make a group of self-replicating entities capable of acting like humans. If you don't believe consciousness can erupt from a non-feeling instruction set, especially one that is going to iterate, and iterate, and iterate itself for generations, then I suppose I feel sorry for whatever you give birth to. I don't really understand why you'd even bother to give it a "morality system based on the Bible" in that case.Good luck on your project all the same.
>>2870955 so you choose to stay in your delusion that machines can be conscious. Sad.A moral system is still needed for a unconscious machine that makes serious decisions that must take into account a moral choice. Consider a self driving car. Surely you aren't going to suggest that is conscious too right? Well you probably will *rolls eyes*. Anyways, if it has to hit 2 kids and kill them in the process or hit a grandma and kill her in the process and these are the only two choices, it has to make a moral decision there. A morality system in place would solve this. But beyond that, even chat gpt which is not conscious gives morality related discussion. If my robot will talk to people it needs to be able to talk in a moral way. It needs to be able to tell people if they do wrong. To not add that and have it approve of immorality would be foolish. Whether it is conscious or not, the programmer has to put that system in.
or lets say somebody tells it that it should go around and call everyone an a-hole as a greeting. A moral system would prevent it from taking that advise. Even the most basic chatbot that is EXTREMELY obviously not conscious would be able to benefit from such a system then. To prevent people training it to do bad stuff that is obviously wrong.
>>2870140>It will pee out the water once it gets warm and go get more ice water as needed.It's a fun idea if it works, which seems unlikely given how much cooling you're aiming for from a few cups of ice.Your 'advanced humanoid' is going to have a severe drink problem, constantly in the fridge looking for ice, in the toilet pissing when not drinking.Who wants a humanoid if it behaves like an alcoholic?Ice won't cool it in my opinion, but at the very least, use the 'pee' to water plants, maybe his dick can be a nozel that sprays, he can heat the pee up with his balls, then steam clean curtains and couches.
>>2871043 it can be like 1-2 GALLONS of ice. and it can replace that every 10 minutes if needed. how would that not cool things. Who said anything about 3 cups?You suggest this would be a constant thing. I already explained that in normal usage, the evaporative cooling, liquid cooling and air cooling systems would most likley suffice and that ice is a thing that can be employed for the most extreme circumstances like doing intense workouts or mowing outside on a 100 degree Fahreinheit summer day. It would not be typical.Anyways, why would you think ice can't cool something? That's folly sir.
>>2869905>OP insists he's not making a sex bot>makes robot have to pee for no reason
>>2871059I feel like you're losing a lot of 'use case' for the sake of making the thing piss desu. >That's folly sir.Which 'design' are you planing for the genitalia?Male is most functional if it's going to be peeing.
>>2871163How else are the ice cubes going to get out then?
if you had the latest robot from whoever is most advanced right now, how would it affect your developmentlike, if you were gifted the chassis of whatever boston dynamics is working on right nowwould it change anything?
>>2871163 the robot pees because water doesn't just dissappear and if ice and water are to cool it, once they warm up they stop cooling it. So it has to flush that to get more icy water and ice cubes into it. also you pretend like it needs genitalia to pee. It doesn't. It can be a surgical needle size sprayer or w/e, not anything sexual.
>>2871306No sweating robot?Phase change is the best cooling you get out of wayer anyway
>>2871298 the serious players in functional humanoids that do useful work don't go for human appearance because that's a hard hurdle and they want low hanging fruit so don't bother. they make a mech looking one. I want the best of both worlds and am not in a rush to market like them. I'm in this as a hobbyist with no boss down my throat shouting deadlines etc. So I can go full human appearance AND full capabilities in strength and movement speed etc like a human. The big boys haven't even tried this. They don't want to spend time to do this when people wanting to buy don't care that much about appearance like I do.
>>2871307 it will sweat too. but that is not always a good option if you want to have it be more dry sometimes its better to just let it pee then. if it sweats alot that could get furniture wet and stuff if it sat down or w/e.
>>2871308didn't answer the questionif you were gifted a fully working chassis, with a completely documented api of its sensor inputs and control outputs of the most advanced currently available robotics tech, what would you do differentlyor are you saying the form factor of {pic related} is somehow restricting the robot from doing general tasks?how?
>>2871325 it would likely lack various degrees of freedom the human body has which would make it less capable of dexterity and certain tasks a human can do and a proper anatomical humanoid robot can do. It would also look like a toy compared to a robot that looks properly just like a human. If you see a refrigerator looking thing with a big smiley face on the front and legs and arm looking like hvac hoses, it looks like trash and the robot you pictured is just slightly evolved better than that. The end goal is looking exactly human so I'm skipping to that part. By looking and passing for human it will be able to better communicate with humans and we will better be able to think of it more as a friend and it will make it more interesting to talk to. Having just some mech body and two lightbulbs as eyes and no nose or mouth really takes away the human interaction capabilities alot. Ruins the immersion of treating it like a friend more. Ruins alot of potential fun socially speaking. Also, it would be fun to try to have it talk to people and see how long it can last before they realize its a robot. Also if it passes for a human it can shop or go in public places incognito and not get harassed or reported or w/e because it just fits in and looks the part like it belongs there because it looks like a real person. This enables it to go into public safely and do useful work in public. A mech robot would have store owners call the police in alarm.
Here's a concept drawing for a Archimedes pulley based downgearing system I have planned for the robot to be able to silently downgear its bldc motors. I HATE the loud noise metal gear based downgearing emits and want to avoid that at all costs by downgearing in this novel new way if possible.
yo OP, check out this cheapass tactile sensor: https://www.youtube.com/watch?v=8eTpFYgae64good luck on your pissbot
>>2872004It won't be silent, BLDC motors still make noise.
>>2872099hanna dev made a test and they did indeed seem silent. he should still use servos for the face and give pwm modulation snd padding a chance but hes just going to do his own thing his own way like all these robot indie snowflake devs dummies.
>>2871635>certain tasks a human can dowhat tasks?>It would also look like a toy if it looks like a human, it'll perform like a toy. You can't get human level performance with something that looks like a human right now. We don't have the tech. We don't have artificial muscles that work EXACTLY like human muscles work. You want muscles that 'bulge' exactly like human muscles when they contract? Nope, we don't have those. Oh, and dielectric elastomers and HASELs(some of the most promising artificial muscles) work differently than human muscle. Dielectric elastomers expand rather than contract, and it looks like HASELs need clutches to get large actuation strokes. So even if you use artificial muscles, the optimal musculature must be different! Another issue is that humans run off of chemical fuel with a pretty high energy density, while robots got to run off of lower energy density batteries. And worse yet, real muscles can store energy too! To be competitive, robot muscles should be stronger and lighter than human muscles, so that the robot can carry more batteries while weighing as much as a human. Getting performance requires deviating exactly from what humans look like!>looksgive up on looks and build for performance. You may not like it, but digit is what peak performance looks like. Those legs are BUILT for long battery life with springs to recover energy on every step.>A mech robot would have store owners call the police in alarmwould it now? Food delivery robots exist and those don't look anything like people
Gentlemen. The skin still has not been resolved. The silicone rubber has the consistency of yogurt and im broke and i'm not getting anymore silicone rubber from alibaba. I also dropped one of the buckets anyways. I made this just now. The new plan is to make a honey comb 2 mm structure with 15% infill with the minimum 0.4 paper thin sheet layer on top made out of 95A tpu for now...https://autode.sk/40K68CA
>>2872108 the only reason all human appearing robots today perform like toys is that the makers were lazy and took shortcuts. I am not doing this. I will be making the performance human level strength and speed while looking exactly human. The best of both aesthetics and function. No compromises. You are wrong to think this is not doable. Just because none have been up to the task so far does not mean it cannot be done. It can be done and I am doing it.Digit looks like trash.Battery life does not matter because you can have hot swappable battery packs worn as bookbags just like a cordless drill so this means there is always a full battery ready to swap in and no downtime.Food delivery robots that are sidewalk only cannot and do not shop in stores like mine could.
>>2872206 just save some money and try again tpu is too hard
>>2862875>we don't have any artificial muscles that expand and contract exactly like the real thingLinked here>>2862700>https://www.youtube.com/watch?v=Ga_IafGRWyEThat long-stacked EAP actuator shortens in length and expands in width as a high voltage is applied. It is without the traditional motors of magnetic fields. The force is from the electrostatic (Coulumb) force.Longevity depends on the rubber used, some silicones are even safe enough for human implants
>>2872230I dismissed Hasel actuators at the beginning because it requires thousands of volts of DC and that isn't going to happen. Do you know anything about the polymer ones though.Definitely don't want high current either as that can kill you...
>>2872235>Do you know anything about the polymer ones though.No idea desu. >Definitely don't want high current either as that can kill you...Voltage shocks (sparks in air is 1 kV per millimeter) and current burns (like welding).Building robots is not for the faint hearted!
>>2872230 no, a actuator has to be able to expand and contract at different speeds and be able to stop on a dime. Only bldc motors can do this with enough power and precision for a real robot. All the other stuff is just experimental and failures across the board. There's a reason its in no robots on earth.
>>2872249i think polymer is out of reach. i just googled a research gate article and it mention carbon nanotubes etc...an alternative for face animatronics though could be magnets though. something like steel rings getting pulled by a magnetic force via a string or bags filled with liquid that swell up to move the lips via pumping or something.the thighs in particular id one if be most problematic part of the robot as it requires high torque. maybe that could use hydraulics instead...the spine could be problematic also but if its stiff then its useless to me.
This thread is really the most retarded shit I've ever seen in robotics, you two are the biggest larpers ever. Try to build a working humanoid with normal tech first before starting to consider retarded bone structure and flexible actuators. If you were smart you would've just done one joint and tested + improved it and you would've made a simulation already. Now you are deep into a project that will probably never work but you can't know it. We are building many humanoids in our lab (we have 4 different types now of every size) let me tell you the process:-sourcing/making one motor + gearbox (optional)-designing one joint (usually we use elbows or knee as a starter)-designing the control system and test the limits of the motor -designing the rest of the body -creating the simulation-after it works in simulation we buy everything You are literally doing all those steps in reverse sorry anon
>>2872271im just not into it anymore. theres no money to be made here. no patreon. no buy me a coffee, no youtube ads, nothing. damn it to hell. Nobody even listens. they cant "pay" attention. here i am dealing with a nagging customer for a custom 3d printing for pennies. what else can i do im 35 not even halliburton will have my ass at the oil rigs.
>>2872271 you say all that but ignore the reality of my progression. I built a single sample of silicone skin. I built a few iterations of humanoid bones with artificial ligaments as the frame. When you build a car, you DO start with the frame so stop with the nonsense. This IS step #1. Next, I have attempted many motor + gearbox combinations to actuate a single joint. I have ruled out many many options and narrowed things down to bldc motor + pulleys for silent downgearing and have been iteratively testing and improving on pulley designs and formations. When this is finalized, I will have a powerful, streamlined, small form factor actuation solution to actuate a single joint and perfect that. Then I can use the same actuation for the next joint and so on. So all your talk of larping and w/e is purely fictional delusion on your end. You are in gross error and misjudging my strategies, progression, and approach entirely. Folly.Also enough with the "you guys" stuff. robotwaifutechnitcian and Artbyrobot are not part of the same team and are just acquaintenaces with a common interest in humanoid robotics. That's it. His approach and goals are very different than my own. Deal with him separately from me.
>>2872294sorry I was a little mean I understand that you had progress but no you generally don't start with the frame or skin when designing a robot.> I have ruled out many many options and narrowed things down to bldc motor + pulleys for silent downgearing and have been iteratively testing and improving on pulley designs and formations. When this is finalized, I will have a powerful, streamlined, small form factor actuation solution to actuate a single joint and perfect thatOk, I obviously did not read all your blog posts but anyway that's my point you should finish it before everything else, that's the hard and important part as said here >>2862962>You are in gross error and misjudging my strategies, progression, and approach entirely. Folly. maybe, honestly I hope everything will go right for you, we bought our last robot for 150K USD so I guess if you could bring down the price that would benefit me too. I just have huge doubts for now, I'll visit this thread once in a while with the hopes of seeing a video of something actually being controlled.
>>2872311id be more motivated to make the 2 feet walking shell bot with sphere joints sample if you promise to subscribe to one of my patreon tiers if i do and thatd be just to cover the cost of materials.i cant afford shit right now. https://www.patreon.com/robotwaifutechnicianheres me talking to the robot headhttps://m.youtube.com/watch?v=J_xvPFFo6-4&t=352syes i know its super slow cause im not using an api like chatgpt or gemini but i can change plans if needed
>>2872317you are just a cringe perv I'm not responding to you
>>2872311 I disagree that one should not start with the frame of a robot and I disagree that people generally don't start with the frame. I've seen multiple robotics projects begin with the frame. And beginning with the frame is the intuitive and proper thing to do for designing most machinery. So not sure what you are on about but you are in error IMO.Also I didn't suggest finishing the skin or anything like that. I just did a single test of a small piece of skin as a proof of concept to dabble into it and begin the planning process early on. Same with the AI I spent a few weeks laying the foundation. Just dabbling into some of the key areas as a break from the main focus areas can be a good idea to begin the planning process on future aspects so when you are ready to get into that you already have a plan in place, did your research, etc and can dive right in when that phase of the project arrives. Rather than just get to that phase without a clue. Not to mention people ask about future phases and how you plan to tackle that phase so it can be good to at least do the research, maybe a light test or two, and lay some kind of foundation and plans in advance so you have an answer to those questions as they come in.
>>2872345>So not sure what you are on about but you are in error IMO. Our humanoid participated in picrel almost 10 years ago. I believe there's a lot of different ways to design robots of course but putting so much effort on the frame instead of the control and motors is definitely retarded. Yes nothing is stopping you from researching other buzzword shit, I'm just saying maybe invest this time developing a car that drives before thinking about the paint colour and the rim size. I would be really surprised if your design could do 10 meters without collapsing for now so yes I encourage you to focus on that first it's just an advice, good luck
>>2872362 I already have shared I went with a pvc medical skeleton as the frame and spandex as ligaments so that part is largely dealt with and the focus can now turn to actuator development with the pulley downgearing. Without that important step established relating to the frame, I would not know where the motors will be placed in CAD and would not then know what kind of space constraints I am dealing with as it pertains to motor selection, design, and placement, gearbox selection, design, and placement, control circuitry design and placement, etc. So all actuator development without CAD indicating placement and how much room you have to deal with would be futile. You'd end up with actuators that won't fit the application. So starting with frame development and overall form factor establishment is critical and must preceed actuator development as a necessity. Otherwise, you are GOING BACKWARDS like every robot in that picture you posted. They threw together actuators that were too big then tried to make it a bit human looking in hindsight as an afterthought hence why they all look like big, ugly, mech abominations. The prudent approach for a beautiful final robot is to design from the outside in - first estabslishing the final form factor of the exterior of the robot - its body shape, then designing the frame and actuators to fit within that form factor constraint, doing wahtever it takes to make it fit inside that. This is done largely in CAD but also in the actuator iterative design phase itself, constantly referencing the CAD and form factor constraints to ensure you are within your constraints because not a single actuator can go outside your form factor constraints which would ruin the exterior appearance of the final robot when completed.
>>2872458>I would not know where the motors will be placed in CAD and would not then know what kind of space constraints I am dealing with>doing wahtever it takes to make it fit inside that.Here's you in a year : >Wow I spent 10 years making a human skeleton model.>Now let's see if I can fit my 100 DOFs in it >ohshit.png >it's ok I just need to use smaller motors >robotcantliftabottleofwater.jpgWe simply don't have the power density for your frame and you would've realised that if you did as I said, now you don't even know the final weight and just pretend your motors will be able to lift the robot anyway. honestly I suggest you take some engineering classes instead of eyeballing everything but I'm afraid you'll try to teach the professor and take a reality check in return
>>2872581fact check me real quick. i came across the formula for the torque needed for the hip motor. it us(weight of the robot in kg)*(height of the in m)*sin(angle of the leg(45))=motor torque in nmthe mr996r is roughly 1.7 nm slightly more since the sphere joint uses bevel gears.my test shell bot weights roughly 1 kg at 2 feet(0.6 meters)*sin(45)=0.42nm torque required.
>>2872587Depends the specs you want to have, pi/4 seems like a reasonable range to have your hip worst case scenario set in. However be aware that your formula considers the feet to be placed on the ground already, if the leg of your robot is long and you have a weight at the end (a servo for the feet pitch control), then you need to do the same torque check but in reverse, consider your base fixed and your leg moving to see if you have enough torque to move your leg quickly.
>>2872587as a side note I would really suggest to move from RC servos to something a little more professional such as some dynamixel servos for small humanoids. That gives you a huge cabling advantage + a pre made ROS interface+ PID gains control + speed and torque control + virtual limits + a low backlash gearbox + a joint state readout ect ect, there's just too many advantages over using a RC servos
>>2872581 I have already successfully fit all the motors required for my 100+ DOF robot into the form factor of the human body with enough power density built in to successfully match human level speed and strength. I already achieved this in my CAD and now just have to build it out. So you are in error once again.
>>2872581 also I did not spend 10 years building the frame. I spent 3 or so months carefully hand making a fiberglass rendition of the frame before I had access to a 3d printer or other potential options. I was going from scratch with clay bone sculpting. I then had a viable arm and chest to begin electronics work on. Then I began iterating through possible actuators to use and ruling them out one by one over the course of 8 years while also studying during those 8 years electrical engineering, mechanical engineering, AI, and ever other related field and getting significant progress in those areas and doing some preliminary work on those fronts here and there with either related projects or some prototyping for this project series. I also spent alot of time on unrelated projects, establishing a financial foundation, some other business ventures, a family, and various other stability in my life. You pretending I just literally locked myself in a room for 10 years building the frame and only coming out to eat once in a while is a joke and completely untrue. I probably spent 95% of my time those 10 years not even working on the robot.Would you say the same thing if someone cleaned half of a messy bedroom for 5 minutes then went to the military for 10 years, came back home and cleaned the other half of the bedroom for 5 minutes? He spent 10 minutes cleaning the bedroom in total actual time spent but you could say it took him 10 years to finish cleaning the bedroom. The latter phrasing is misleading even if in some sense it can be said. That is what you are doing to me right now.
>>2872609>>2872610oh no im just using these servos cause i have them in my room right now and cause the original design for the spherical joints uses them. https://www.thingiverse.com/thing:4857920for the actual 4.5-5ft robot id have to go with blcd with a gearbox maybe and like i said earlier now sure if id use hydraulics as well. this is a ptototype of a prototype. i am well aware the weight scales exponentially with size. square cubes law,it also wouldnt be mostly an empty shell.i also intend to put two box pockets on the chest to put nuts and coins and see what effect it has on the balance. ill aksi use an mr996r with pulleys for the spine. thank you for the confirmation. for the control system somebody named a software suite once but i forgot the name.ive also read about ros control but ive been told thats a bridge. for the time being ill go with cascadeur api which is really meant for animations.i do not know the higher level physics or maths to do it properly so i have to find something...dynamixel is out of the question because they are very pricey. however before i get to all that ill do a test with the sg90s to see if i can make them less noisy by using micropython with the machine module. c++ is more convinient but machine is a premade solution for pulse width modulation.>>2872622artbyrobot bros. youve been talking about usibg pulleys for the fingers for months bros. just do it already. also maya will do to get the point across but you should pick up fusion 360 or onshape. theyre not too bad. also do the demonstration of the ai without neural networks.
>>2872581 I just want to point out one more thing on this topic. So when someone is doing something highly experimental and highly unconventional, without any guidance to just follow along a proven path, when they are trail blazing, things can take way longer to get done. It is like one step forward two steps back. I found myself making actuator after actuator, mounting it, testing it, and removing it from the robot realizing somewhere along the way that it is not viable for my use case. This repetitive process of elimination, ruling out failed ideas and trial and error can go on and on until you find the correct choice or a valid choice and are able to step forward from there. It almost feels like Odysseus' wife weaving and unweaving that piece of textile over and over day by day where it wasn't ever getting finished. Yet these little experiments are critical to enable one to find the best path or at least a viable path. Ideas can prove wrong over and over. Consider Edison with the lightbulb, how many times he tried and failed before landing on a viable lightbulb construction. It was 99% perspiration and 1% inspiration the saying goes. This is sometimes the nature of inventing. It is not alway so easy. You can't always one shot things. But if you press through these blocks you can find something the field desperately needed. So you press on, confident you will find a way. Seeing the progress in such cases might seem like hmmm... seems like he did nothing. But all the trial and errors are not nothing. I found many ways it cannot be done. That is something. That knowledge was not obtainable any other way. And if I'm correct, I am now landing on a way things can be done that is game changing. I think I am arriving now. That was not insignificant. And the time taken to reach that cannot be diminished or ignored like this. It is not just.
>>2872631 >youve been talking about using pulleys for the fingers for months bros. just do it alreadythe pulleys I've actually talked about for at least well over 2 years maybe 3-4 years I planned it. I have gone through many failed designs on them. I have built them, then had to throw them all away and start over. This happened repeatedly. So just saying "build it and be done" is not so simple. I have to build ones that work or it cannot ever be done. I've failed repeatedly at this. I've learned a ton along the way. There is nothing simple about it.>maya will do to get the point across but you should pick up fusion 360 or onshapebut why? I literally take models right out of Maya and into Cura for 3d printing them. why should I use anything else if Maya works for all 3d modeling jobs possible and I'm comfortable with it and very competent with it already? Why waste time learning a new program that achieves the same result?>also do the demonstration of the ai without neural networksI was so excited about the AI part of the project I put a month into it some years back. I laid the foundation and got a great start on it. It is solid. I don't want or need to work on it anymore until the robot arm and head are done. Then I will shift my focus entirely to the AI and I will advance the AI to the point that it causes the arm and head to build the rest of the robot's own body. This means after I finish building the arm and head, I no longer need to build anything. The robot takes over entirely and my entire focus is then on AI going forward.
>>2872649sorry about that. Its none of my business really.
Here is my design for how I want my Archimedes pulley system pulleys to look for downgearing my motors.
>>2873018youre neglecting a lot of things artby robot. how are you going to mount the pulleys and motor, youd want 1.7 nm torque atleast if were going by the mr996r torque. what torque does the blcd have. what voltage would you need to make it slow and the right torque. what kind of blcd ir dc, what would be the resulting torque. what are pulleys diameter, will more than one pulley be necessary to have a multiplier effect to make up for the space contraints. saying oh just use pulleys instead of whats been proven to work doesnt mean much. youll be doing more trial and error as well to get the measurements right since youll be using maya instead of cad software.ill start the pulse width modulation test on tge 23 and post the results.
>>2873023 I see you are not reading the whole thread. I addressed some of what you are asking here:>>2872004My 2430 bldc motor is 0.4lb of pulling force acting as a fishing line winch. I will be going 32:1. So 12.8lb pulling force is the final but after friction maybe more like 12lb. We'll see. After mechanical disadvantage of the finger joint, it will be probably 4lb of actual force as far as load the finger can contract. This is nearly human level strength for a finger contraction. Also no, you don't modify voltage to reduce bldc motor speed. You reduce the speed by reducing the phase switching speed on your dual hbridge pulses, You can make each turn of the motor as slow or fast as you want.The pulley's diamter for most of them will be around 3mm but the outer flanges will reach out wider than this. 5 pulleys will be in the archimedes pulley system and 1 pulley will be a turn in place pulley mounted to the motor.You saying you can't measure things in Maya is not true. I measure everything in Maya all the time. I use a ruler in maya, a literal 3d model of a to scale ruler that measures in metric mm/cm. So the notion you can't measure in Maya is a false one. Also Maya IS CAD software. CAD means "computer aided design". I am using Maya to aid my design.
Here's a archimedes pulley system prototype from a video on pulleys that inspired me on using pulleys this way on the robot.
Here's an attempt to make the servos quieterhttps://www.youtube.com/shorts/JxQGAHXiGloIts not great but I wanted to be sure. So i figure we can afford to make the pullback slower if you think about it...maybe
>>2873473 I appreciate that effort and appreciate the fact you see the value of making robots quieter like I want to do. It's nice to see someone agree with me on this. I personally don't want a super slow robot so just making it go slow to get the gears to be quieter is not an option for me but it is one way to make gears quieter I agree.
>>2873564i think the pull in is roughly as fast its just the pull back thats slower but yeah sounds like vinyl scratching.I got a good feeling well be able to pull of the magnetic gears though.magnetic gears are not an idea theyve been used in real life and theres even software for simukating them.
>>2873845 I am doubting any magnetic gears were made at relevant small scales to what is needed for a humanoid and all the ones I've seen so far posted are gigantic and unusable.
>>2872230Nope, they aren't exactly like the real thing. They don't work as well when stretched, electrodes get further apart. So if you try to make a bimorph with them, the range of motion's less because you can't contract one muscle too much without the other muscle getting stretched too far. >>2872235>>2872249>>it requires thousands of volts of DC and that isn't going to happenstop being a bitch. 10 kV is nothing. You can play around with this type of actuator(and I mean play, not doing anything useful) with a $5 ionizer modules from china. It hurts like the worst static electric shock you've ever felt, but it's ok. As long as the current's low enough. Safety study's have shown that dielectric elastomers don't store charge to kill you, so if you have proper shut down electronics, they won't kill you. Oh and a new type of HASEL can operate with hundreds of volts rather than thousands:https://www.science.org/doi/10.1126/sciadv.adi9319But you gotta pay out the ass for some exotic polymer, for now at least. It's only expensive because no one's making it at scale.>>the polymer onesyou mean these? https://en.wikipedia.org/wiki/Ionic_polymer%E2%80%93metal_compositesthey're shit. They're slow as a snail
Exploded view of my proposed pulley design. Bearing is in center, plastic discs sandwich it.
>>2873985made these following the gear and i did the math for the volume ratios from this instructablehttps://www.instructables.com/Magnetic-Gear-Box/the idea is to make a 4:1 gear ratio magnetic gearbox about 3 or 4 and connecting them with a belt. that'd replace the servos gear train and be really quiet. i made the 8 mm magnets one a mm bigger or thats what the software says. they should be 2.3cm and 2.4cm.
How long do you think your robot will take OP? I saw in one of your posts here you've been working on this for over 10 years and I know, I know "these things take time", but so far you have no working robot at all. How old are you? Because at this rate you probably won't have finished your task before you die of old age.
>>2874913 it is unreasonable to extrapolate on time like you just did since it assumes I worked on it the whole 10 years - which is false and it assumes I'll work on it the same meager amount each 10 years going forward - which is unlikely. So your two false assumptions are all you are going on. Also, it assumes that the same design and CAD work and massive research taking me from total noob to advanced level knowledge has to repeat the second 10 years. That's also untrue. That was a one time massive knowledge gap to cross which I did. Not to be repeated. So now that hte knowledge is there and designs are there, those are behind me and the future is mainly just building it and coding it, not so much just trying to plan it and research all related stuff. So progress is going to be WAY faster at that point. You did not put any of these things into consideration.I've estimated the robot is a 30+ year project. I anticipate that the building of the robot will take 10 years and that the coding of all of the advanced A.I. and capabilities will take another 20 years. The first 10 years were mostly design/planning and researching so don't really get included in that estimate I guess.
>>2874795 way too big IMO to be useful and also I don't think it will downgear at all if both gears are same size. It is 1:1 ratio then. One has to be 4x as big as the other one in diameter.
>>2874923So you've had 10 slow years, but you anticipate the next 20 will be faster. Have you ever felt before like you were "over the hill" in terms of learning or is this the first time? In other words, are you pretty much there in terms of the knowledge you need? You say you now have "advanced level knowledge."Do you perhaps have a plan of how you will arrange the work across those 20 years? I know you'll be alternating between robotics and AI, as you've mentioned, but I'm just wondering if - since you have an idea of how long this will take overall - you've broken down the work into smaller tasks, and roughly how long you think each subtask is going to take.
>>2874945 estimates of how long a task will take is based on how long similar tasks took you in the past. If you have done nothing similar or for some parts are doing trial and error on things never befroe done or shown publically, then things become all the harder to predict. That's why I kept it vague. I will do one actuator and test fully and test on one joint fully. Then I will rinse repeat for all other necessary joints to get the arm and head fully functional (no facial expressions yet, just minimum funciton of head for it to do useful work using eye hand coordination). Then I will switch to AI focus and train the AI up till it can cause the arm/hand/head to build the rest of its own body for me. So I have a direction but as far as creating a exhaustive list of all tasks even ones way off in the future, no, I will make subtasks of subtasks lists as I enter into those phases but keep the subtask lists macro scale while they are still far off.The learning is a spectrum, but I have felt pretty confident I've passed the learning phase "over the hill" anyways since maybe year 5 of the 10 year period. But the last 5 years I barely worked on it. You see, I have only worked on this project fulltime for the equivalence of 4-5 months actual development time hands on (not including massive research). So to be more precise in duration of project so far, it would ideally be phrased: "I have worked on this project for 4-5 months spanned sporadically over the past 10 years". This means 9 years 8 months of the 10 years were spent not working on the robot at all other than passive problem solving, planning, and research but no actual hands on development progress during that.
>>2874945 also the project is 30 years, not 20 years and I'm 4 months into that 30 years despite 10 years having passed since I announced the project. Life has not allowed me to work on it more than 4-5 months so far but I think it will allow me much more time on it going forward than ever before. But that's just a guess.
>>2875003I wasn't asking for an exhaustive list of course, just a list of the largest sub-projects you have to complete to get your robot working. So for the "one joint" subproject, how long do you estimate that will take?I understand you've been working on it on and off, but have you made any robot or robotic part at this point? Just for practice and a feel of how these things work, and to understand what limitations you will inevitably bump up against (cheaper in terms of money and time to discover that on the small scale).Oh also, which parts of your robot are going to be doing things never done before? I don't mean in the implementation (as in, I'm not talking about your peeing cooling mechanism), I mean in the outcome. I have read your ideas in this thread, I just want to know which ones you think are groundbreaking, or whether there are any you haven't mentioned that you think are groundbreaking.>>2875004If you don't mind sharing, just in vague terms, how old are you right now? I guess you must be approaching 30 at least, since you probably didn't start building until you were at least an adult. I - and I think most people - would not be able to do full time robotics development for 30 years or more (especially because that will encompass your late 20s to your early 60s, in other words, the time of life when you are most able to work and build up a life) without some form of external funding. Do you have something in place for that?
>>2875032>for the "one joint" subproject, how long do you estimate that will take?full custom electronics motor controller: 40hrs?full custom microcontroller: 30hrs?full prototyping trial and error perfecting custom downgearing: 100 hrs-300hrs?full custom sensors suite: 30hrs so 400hrs more for just one joint I think but setbacks may happen and long breaks sometimes happen for a season so this might not be all togethr it might be sporatic over monthsmade any robot or robotic part at this point? I made 3 skeletons like 30%-60% progress on themI made partial custom motor controller 60%I made partial custom microcontroller 35%I made partial custom power supply 35%I made 10+ custom downgearing setups to 60%ish which each had to be scrapped with lessons learned.I made a proof of concept exoskeleton 30% which had to be scrapped with lessons learnedI made a proof of concept robot hand skin which had to be scrapped with lessons learned
>>2875032>Oh also, which parts of your robot are going to be doing things never done before? -ice cube cooling system-actual exact human skeleton fully loyal to exact replica for whole robot-the layout of my custom microcontroller-the layout of my custom power supply-the layout of my custom battery pack-the hot swappability of my custom battery pack-the downgearing system with pulleys-the means of attaching all parts to the frame (suturing)-the ice cooling system-the artificial lungs-the peeing system-the custom silent pumps I'll be making-the expert system that acts as its own expert coding itself in part AI plan which is already well underway in implementation phase (like 1.5% progress on a multi-decade project)>I don't mean in the implementation (as in, I'm not talking about your peeing cooling mechanism), I mean in the outcome.-well I read this too late lol already listed what you didn't want but I'll leave it up for others. But different outcome? Quiet robot that cools well and is entirely closed in silicone skin while also being fully human dexterity and speed and strength from head to toe with all significant muscles and bones representing for full humanlike performance-codes itself-makes most of its own body then self replicates to other robots
>>2875032> how old are you right now?early 40s> I think most people - would not be able to do full time robotics development for 30 years or more (especially because that will encompass your late 20s to your early 60s, in other words, the time of life when you are most able to work and build up a life) without some form of external funding. Do you have something in place for that?yes I do have something in place for that which is actually operating my own home businesses as well as my real estate investing property renovations and rentals business. All of which were part of what kept me from doing robotics full time - I was busy with these other things the whole 10 years. Oh yeah and vehicle flipping and other investing too.So basically long story short, I work like anyone else and I work medium hard. Not work-aholic and not barely work but medium maybe light-medium. And by this I do okay and get by fine and save and invest. And my wife works part time too. So together we do okay and get by fine. No complaints. And with my free time available beyond the light-medium work load I can do my hobbies like robotics. So that's the "funding". The robot parts is like 10-15k per robot I guess. Although that probably includes lots of incidental tools and whatnot that are one time purchases and lots of overstock so may be more like $7k for second robot? Not sure maybe $6k? I mean it depends. Batteries might add up quite a bit and so do motors so not sure 100% just ballpark. Who knows maybe $8k-9k?And I bought probably 1/2 the stuff so maybe need $5k more to complete it and I have that covered no problem. And since the time is just hobby time, that doesn't have to be funded/compensated although having the hobby time bring a monetary return is cool. That is basically what a researcher does: they get grants to fund their research hobby which then makes it a career. If they lost grant funding they would have to work and do research on the side
>>2874925nobody has read the instructables link. the gear ratio in a magnetic gear is determined by the volume and amount of magnets not the diameter. why does nobody watch or read the stuff i post. trust me its not a computer virus. assuming the magnetic force is determined by the volume of the magnet.anyways you just got put a little more juice ($$$) into it. set aside $30-$50 per month for parts. i also think working on the ai at the same time is a mistake. if the robot works well it should be no problem to find people to work on the ai on github cause unlike the robot that costs no money and there is a glut of people with a computer science background anyways. the robot and control system is the hard part not the ai.lastly i dont know maya but i know blender. the pieces i just made for the magnets took like 30 minutes to make cause its the right tool for the job. start by drafting a circle in fusion 360 and extruding it. im sure the rest will click and youll be able to pick up the basics in a month or so. i cant get a hold of ecoflex silicone cause thats onky found on the us. i got a plan though. make a mix of hardware store silicone meant for sealant. got to be cost effective wuth our limited resources. cant be any worse tgan the silicone i got from alibaba
>>2874913If Tesla throw billions at robots because of Musk and get garbage, the roboteers in this thread could end up with a useful functional robot arm and it'd be worth it.
>>2875106modern day corporations suck at research. their billions are good for mass production not for getting people to think.dont ask me whats wrong with modern day corporations but there is something wrong and theyre not a good example of anything.anyways theres been a lot of work in robotic tooling this past decade so a lot of the hard work has been done.
>>2875106We already have useful and functional robot arms though. I don't see how Tesla factors in to what's happening in this thread.>>2875110>modern day corporations suck at research. their billions are good for mass production not for getting people to think.Why do you think this? I've noticed this is a bit of a mantra with indie robot devs, and I suppose it must be so they can believe they have a chance of succeeding on their own where well-funded, well-connected groups of researchers have failed.Have you read about VC-1 or NVIDIA Isaac? Did you like the BD Atlas? (Not the new electric one, the hydraulic one.)
>>2875127yeah boston dynamics and figure get a pass but im just not that impressed by tesla bots.But they have the means to do better still.Use robot polymer muscles instead of motors for example. theyre doing tge bare minimum. Sticking with whats safe.i tried mujoco and nvidia issac. you have to define the constrains in issac i dont like its also bloated. For mujoco i wasnt able to simulate gears so i also dont like it that much.
i dont even know if its possible to fit all the little kagnets in there without them popping off. i dont know how many 2mm magnets i lost so far but its over 4...
>>2874795>>2875198Wouldn't you get mad slippage with these? I don't see them as being useful in your robot or OP's. I did read and watch the contents of the link, and it didn't make it better.
>>2875203its frustrating that im being second guessed all the time and that people dont try to understand what im trying to do. If it worked as it did in the instructables one yes it would be useful. the servos could be modded and theyd have little noise.
>>2875220You're getting second guessed all the time because you're retarded. You always post the stupidest shit imaginable. "Second guessing" is an attempt at politely showing you that.The instructable specifically mentions loss of torque. Did YOU read your own link?
>>2875227i assume its loss of torque compared to a regular gearbox...
>>2875097>nobody has read the instructables link. there's no desire to click an instructables link to build a downgearing solution we have no intentions of building>the gear ratio in a magnetic gear is determined by the volume and amount of magnets not the diameter. that's false. The magnets are just the gear's teeth replacement. By no means can you downgear with gears the same size as eachother whether its a belt, teeth, string, or a magnet nothing changes on that its the same principles the whole time.>assuming the magnetic force is determined by the volume of the magnetno the force comes from the spinning motor shaft the magnet force is just the gear tooth replacement alone it has nothing to do with the final output force of the pulley downgear other than ability to rotate the next pulley by gripping it using magnetism instead of gears or a belt>anyways you just got put a little more juice ($$$) into it. set aside $30-$50 per month for parts.THIS. not hard for an adult>i also think working on the ai at the same time is a mistake.This. Unless you are just too eager you can work on it a bit but the control and body of the robot should come first I agree.>lastly i dont know maya but i know blender. maya can make cylinders and extrude and all that stuff too. It's all I use and I can model ANYTHING with it no problem.the pieces i just made for the magnets took like 30 minutes to make cause its the right tool for the job. start by drafting a circle in fusion 360 and extruding it. im sure the rest will click and youll be able to pick up the basics in a month or so.> i got a plan though. make a mix of hardware store silicone meant for sealant. got to be cost effective wuth our limited resources. cant be any worse tgan the silicone i got from alibabaThat's a good plan. I already did this and it worked great. It's a bit hard though so maybe do a soft layer over the top of it or use the alibaba stuff for softer areas
>>2875262This reply says nothing. The statement with or without the comparison means the same thing.Have you done any research or reading around the overloading of those gears?
>>2875322the magnetic gearbox will go to the back burner for now. i see more potential for using electro magnets for quiet face animatronics anyways. Which is the goal right now. Here's a an ugly sketch with the idea for the eye. it uses a 9 electromagnet grid to move the eyeball around. im not sure what the joint should be for the back of the eyeball but im going with a universal joint for now and see if that works. im 3d printing the pieces right now ill post the results in an hour or so.
>>2875580> 9 electromagnet grid to moveJust use hard disc drive head actuators
>>2875581that only allows one degree of freedom of movement it needs to be 2. It might also be over kill. the amount of force required to move tge eyeball around is oretty small.
>>2875583i cant tell if this will work cause i did it pretty fast but maybe.
heres an electromagnet with 30 turns. the plan is to make it 100 turns. 9 of them. 3*3 pulling at a screw at the end of the of the eyeball lever. left diagonal, center up, right diagobal, middle left, center, center right, lower left, lower center amd lower right.
Okay I put this together there should be no misunderstanding on what the plan is by now. It'll take a few days or so maybe a couple.
>>2875984ive been meaning to print this to try it out but my 3d printer has been kept busy. anyways im thinking if moving the cylinder right there from the yellow to the red. that should work or work better i think.
Side view of custom mini pulley for my archimedes pulley system. It is based on plastic discs from strawberry containers and a 1x3x1mm ball bearing from aliexpress.
Top view of same custom mini pulley
>>2876561im sorry bros but that's hardly worth showing off. Its just a ball bearing with some goo by the looks of it. unless im missing something and you machined it...
>>2876577 it's been formed into a pulley scroll up and study the exloded view drawing of the design. It is an incredibly important part of the silent downgearing plan's execution and took alot of skill and precision and creativity to come up with this approach. It is also extremely miniaturized which is necessary to do when you need THOUSANDS of these to fit into the humanoid.
>>2876657so the ball bearings are spheres and you machined them into pulleys?well yeah in that case good job. guess theyd be more durable if theyre made out of metal. Even so there are tiny desk cnc machines that can make the process less labor intensive(assuming you machined a sphere into a pulley by hand) ive got a module for generating 500 watts(even though i should have gotten the 250 one) anyways my plan is to make brown gas and have it come out of a tube from a mason jar. Theres a youtube video on it. it can make a torch than can generate over 1000 degrees celsius worth of heat from water with baking soda(seawater works too) anyways thatd be a nice tool to have.the 500 watt module can be dialed up and down anyways.heres the video i found about ithttps://m.youtube.com/watch?v=wpMaB-gIFUE&t=322s&pp=ygUZYnJvZW4gZ2FzIHRvcmNoIG1hc29uIGphcg%3D%3Dyou know i tried that in the past without and just used a wall plug i clipped and blew a fuse lol
>>2876661 no I did not make the bearings I made the plastic discs and did the sewing of the fishing line through the plastic discs and the bearing to construct the pulley all of which is precision challenging tedious work and a good design IMO and necessary to my overall pulley system and must be shown and discussed for people to be able to follow along
>>2862669You should make a realistic Ren and Stimpy, their hijinks would entertain for combined days.
So this is pretty big. anyways i got a good idea of how im going to do this now, the electromagnets will pull strings instead of using a lever. i think ill be making a newer video soonish. i also intend to try to make a discord again and just flat out banning women altogether unless theyre like technical minded lesbians or something(cause why else would they want to make a waifu sexbot)A woman reported my last discord channel. fuck these harpies.
>>2877051on second thought never mind discord nobody posts anyways and they nag you for being online. github discussion page or smth.
And here's a new video getting back to ithttps://youtu.be/Ge7sgc_zFjM
