/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality.

LynxChan updated to 2.5.7, let me know whether there are any issues (admin at j dot w).

Reports of my death have been greatly overestimiste.

Still trying to get done with some IRL work, but should be able to update some stuff soon.


Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB


(used to delete files and postings)

Welcome to /robowaifu/, the exotic AI tavern where intrepid adventurers gather to swap loot & old war stories...

Walking Robowaifu Technician 09/15/2019 (Sun) 07:40:18 No.243
What is the best way to make a life size robowaifu walk? A hydraulic pump powering hydraulic cylinders in the legs with servo actuated valves?
Hydraulics are certainly one of the more powerful methods, and able to keep limbs narrow enough for most designs.

Some problems are the bulk needed in the torso, since you need a pump and a battery. You can put servos on each limb, but a centralized control block might be better logistically. Your call though.

[[5030 is a hydraulic system without the need of pumps. Rather fragile, but seems quite powerful and quick.

A cheap alternative is rack-and-pinion servo controls: hackaday.io/project/26021-two-leg-robot

What's interesting with this project is the use of Ethernet cables to communicate the servos and sensors to a network switch, acting as a sort of nervous system for the computer.

further reading on hydraulic bipedals
I'm unsure to be honest unless you give it large feet for stability. I personally think the arms, torso and head are massive enough hurdles. So we'll cross that bridge when we get to it.
I figure trying to replicate human, bipedal movement will prove itself arduous, so perhaps a more ostrich-like mechanic would suffice. After all, there are about 26 bones which comprise the foot alone. Of course, prosthetics have vastly improved upon this through a blade like design, but still something like an ostrich has had a longer time to evolve as bipedal than man. The robot girls of the future may very look like satyrs.
The best way to make it walk is pretty closely related to how inhuman you're okay with the end result looking. >>551 is probably the closest you'll get to a real person without crippling functionality. Most animals have four legs for a reason.
This guy has made a perfect walking motion with a few servos. Of course, I imagine a large scale robot would need gearing to accommodate the servos and shit. Also, a gyroscopic sensor is mandatory so the robot can detect when it falls down.
>What is the best way to make a life size robowaifu walk?
I'd suggest looking over the thread dedicated to this proposition as one good place to start.

Hydraulics might be reasonable for a big, gas-powered giantess robowaifu, but electrically motor driven solutions are probably better for most of our mobile robowaifus tbh. For a static, in-place robowaifu pneumatics might actually suffice (but ofc that defeats your whole basic premise of walking).
>I'm unsure to be honest unless you give it large feet for stability.
I'm no footfag, but I know there's enough of them that this wouldn't be a dealbreaker. Besides, any gynoid to come out any time in the near future would need to be relatively cartoony anyway, to avoid the uncanny valley. Proportions need not be realistic. In fact, it might be better if they're not.

While I advocate cartoonishness, I feel like this would look really weird unless, as you say, it's made to be a character that is some sort of satyr or something. But then I worry that limiting to that type of character would limit the audience. Better than nothing, however. That said, the difference is basically just that that's walking on what would be equivalent to a human's toes. So put the robot in high heels to justify it.

I worry that walking is more difficult than most of the other things people would want right away. It's an important factor, but it also feels limiting. I think the best short term solution would be to find "narrative" reasons to not need the robot to walk, to justify its inability to walk. Say the character is disabled and put her in a wheelchair. The legs can still move to the extent current tech allows (probably enough for boning), but this would justify not being able to walk. It would make the product more marketable. Then we take that money and invest it into better robots, including the ability to walk. Eventually we'll get there, but I don't want to wait that long to have robowaifus. Demanding perfection will keep us from ever getting them. We just need marketable, and that will fund further advancement.
I think balance is key. Stand on one leg, only lifting the other one half a centimeter from the ground, move the one in the air a bit forward. It doesn't need much power. So either using a big foot, or fast muscles and sensors. The other thing is, how much walking can be separated in subgoals. Walking in high heals while carrying sth, ballet, gymnastics would be the most difficult. Walking on all four, or dancing with the hand to some pole or to the wall would be the other end of the scale. Walking while using the arms from falling would be in between. Carrying sth would be more difficult. Climbing stairs or going long distances is another story. I think walking at home should suffice. https://youtu.be/0ZBD2tcKOU4 https://youtu.be/OaSjF8_Az10 https://youtu.be/YvbAqw0sk6M https://youtu.be/4bYRqRIReuY https://youtu.be/xz9e5ovlCF0
>>4411 Yes you bring out goods points about the wide dynamic ranges of the types of bipedal locomotion the human form regularly engages in. People take it so for granted that very few actually appreciate the miracle it is, and the enormous kinematic complexity involved. >I think walking at home should suffice. Certainly to start with, yes. Thanks for the good video links Anon.
>>243 Only thing that is limiting us here are actuators. Once someone invents a good artificial muscle it's just a matter of using the machine learning meme to make wafiu learn how to navigate. Until then best bet will be monster waifus e.g.: Rachnera octapod type waifu. Bipedal is just not viable now unless you're boston dynamics or have limitless funds (and even there those state of the art bipedal robots are bulky and heavy - will break your bones if it falls on you). >>4411 Everything you listed can conveniently be handled by machine learning algorithms. Once you have your physical body you can create computer model of it and make it learn how to do those things. What's even better is that if one actuator fails or works poorly you can re-run learning procedure with new constraints, making waifu learn how to walk with "broken" leg. It's all about creating a reasonably accurate model of physical body (mass, leverage, damping, friction, torque curves...) and using it in simulated learning algorithms. I think now is the time to mostly focus on general AI (chat, personality) side of things and keep looking out for advances in actuator technology. I know that hardware side of things is more interesting and fun to do (I'm an EE and AI doesn't interest me too much), but technology is not quite there yet and is too expensive. Here is somewhat realistic progression of development 2020 - 2030: Text interface, desktop waifus, VR 2030 - ?: AR with introduction of 6G tech, this is where things will get interesting - realistic holograms ? - ...: Real hardware, if we'll still be alive by then (let's hope for life extension or anti-aging tech by then) 95% if not more of our time should be focused on AI research, other 5% on occasional hardware prototype (content ratios on this board are flipped now more hardware than software), but yes AI is generally boring to do because all that happens at the end of the day is you get some barely passable text response after weeks of work. It's girding your teeth against concrete fun. Well at least it's cheap to experiment with.
Open file (365.48 KB 2048x2048 IMG_20200705_235445.jpg)
>>4427 I agree that it's better to not focus on bipedal walking, but this doesn't mean we have to build monster girls nor to focus on AI. Building animated dolls which can move their legs and arms while lying in bed or move around on the couch doesn't need bipedal walking. I don't see us being hindered by the state of current actuators to build something great, there are other reasons. I also don't see better actuators coming our way. I'm certainly not going to wait ten years before starting to build bodies, I'm not that much interested in desktop waifus or VR.
I think that in order to keep cost down the entire robot should be done by capstan drive. that way you can 3d print all the necessary pulleys and locate the motors inside the body to keep the limbs slender. the cables can be routed through pipes at the joints, much like a brake cable on a bicycle. If you wanted to try push cost down even further you could use one powerful motor running all the time and a system of electric clutches to disengage and engage each capstan. however, this would mean that there is no speed control since everything would be locked to the main motor. it might not be a big issue since each capstan could me custom to allow more fine movement in the fingers where needed.
>>13460 You mean something like this?: https://www.youtube.com/watch?v=ENMZsPwCUcA It seems like a pretty good idea to me.
>>550 >>243 Intermediate steps aren't something to be afraid of. I for one would welcome a chonky boot Rollbot formfactor companion, they're cute
>>13462 That's right. I'm sure I mentioned the idea on the board before. I actually lost the thread because the board is so large. many of the threads duplicates, might become a problem. Those capstans are used on many similar projects, i think i saw them in a robot dog. now that i think about it,using one motor would be difficult because reversing the motor would mean only one direction of movement joints could work at a time. maybe that could be compensated for by using a reverser gearbox or second motor and twice the number of capstan clutches, like the old ford model T which had one clutch for forward and another clutch for reverse. the clutches would need to be actuated by servos so it might be just as cheap to use a stepper per capstan without clutches instead.
>>13480 >now that i think about it,using one motor would be difficult because reversing the motor would mean only one direction of movement joints could work at a time. I'm sorry, I don't understand what you're saying.
>>13481 suppose that you have a capstan for each hand whose function is to make a grip with the thumbs. if you want to grip a ball in each hand no problem, because the single motor can turn both capstans to grip, but suppose you want to release one ball while gripping the other then you're stuck, unless you grip with one thumb, then reverse the motor direction to release the other thumb.
>>13483 maybe use more than one motor? Its a robot.
>>13485 of course, but building a robot out of low cost parts will greatly increase the number of builds. Think of an automobile pressure plate and clutch, now install a pressure plate and clutch on each end of the capstan spool. then each clutch is driven by a shaft rotating in opposite directions. you could get away with using one servo to actuate both clutches, since only one capstan rotation direction is used at a time. if the servo had a spring removed so that it would return to center. I know auto clutches are not suitable for this design but if someone managed to develop a basic clutch, even an interlocking peg that could be 3d printed then that would be a huge breakthrough. however, the capstan would need to be braked while not driven so as to hold its position.

Report/Delete/Moderation Forms

no cookies?