[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y ] [Search | | Home]

Hey /diy/ What are the actual limitations on making myself a

This is a blue board which means that it's for everybody (Safe For Work content only). If you see any adult content, please report it.

Thread replies: 96
Thread images: 15

File: Power-Suit.gif (96KB, 468x320px) Image search: [iqdb] [SauceNao] [Google]
Power-Suit.gif
96KB, 468x320px
Hey /diy/

What are the actual limitations on making myself a powered exoskeleton? I realize its not really all that feasible, otherwise it would be done more often, but I am interested in finding out where the bottle necks on development are.

I'm mostly doing this just to get more familiar with robotics in general and my robotics club at my university are on board if I can get a decent idea for one, which I don't think is a given.

Power could be inelegantly solved by a two stroke motor. Its going to be whiney and annoying but it would work. Could something be done to either muffle it or, inversely, make it at least badass sounding?

Could I scale up the suit itself to take a bigger engine? Hotrod the shit out of it?

Can servos be strong enough but also quick enough for relatively unimpeded movement?

Sorry about the basic questions but I figured someone here was crazy enough to try it.
>>
>>905125
>What are the actual limitations on making myself a powered exoskeleton?
Learning machine code and making powerful motors, then programming them to move in tandem with your body without breaking any of your bones.
>>
>>905128

This is completely relevant and an important part of what I'm looking into. Are there any preferred programming languages for robotics in general or could python do it?
>>
>>905131
Try to stick with C. Also, machine language may be very important to be familiar with here.
>>
>>905140
>machine language may be very important
No, it's not.
>>
>>905131
Depends on your controller. Typically they use C.
>>
> Typically they use C.

nothing uses C, or Python, or whatever; everything is translated into machine code before it's actually fed to the processor. a lot of faggots are afraid of assembler so they make excuses for sticking with a high-level language. a real programmer will have created their own highly optimized subroutines that, in effect, can be called just like a command in a high-level language. that quashes the idea that assembler is long and tedious.
>>
>>905092
Thr actuators will be your biggest limitation. In order for it to be strong, compact, and not slow you need harmonic drives or other exotic speed reducers. These are $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$.

This is no job for 'servos' or direct drive motors. Heck there really aren't any actuators you can buy off the shelf that don't have shit torque densities. Oh and you also need torque/force control or you won't be able to control it.
>>
File: 1335501826075.jpg (42KB, 251x251px) Image search: [iqdb] [SauceNao] [Google]
1335501826075.jpg
42KB, 251x251px
>>905128
>>905131
>>905140
>>905142
>>905143
>>905158
>>905160
You guys know nothing.

What you have to do is to pray to Our Holy Machine God, and appease It's incarnation in your creation by chanting holy binary songs to It's greatness.
Praise the Omnissiah!
>>
You can use high level programming languages now. Controllers are mostly ARM chips running various realtime OS. Controllers have improved so rapidly the last few years. But the actuators I have no idea where to start.
>>
>>905125
>Power could be inelegantly solved by a two stroke motor
No, no it could not. You need a motor at each joint, unless you're positing a crazy system of pulleys and a frame, in which case you basically have a crane that's playing you like a puppet = pointless.
You need something that is computer controlled, which means electric motors, pretty much. Hydraulic or pneumatic if you want something small and light that is stronger than a person already is. you need to make sure it's light but robust enough to not go boom and take off someones arm.
It needs to >>905128.
Right now, as I understand it, the issue is making something small enough to be useful, and getting good enough batteries to power it for more than a 10 minute demonstration.

Also, pic is not relevant, you are trying to build a power loader from aliens at this point, not spartan mjolnir armor, largely due to power issues.
Also, cost is and issue.
So far, you're better off with haptic feedback gloves and a robot.
>>
>>905172
>getting good enough batteries to power it for more than a 10 minute demonstration.
I'm pretty sure that's what he meant with the two stroke motor: a replacement for batteries. Motor runs a generator (or maybe a hydraulic pump) which supplies everything else.
>>
File: image.jpg (167KB, 890x738px) Image search: [iqdb] [SauceNao] [Google]
image.jpg
167KB, 890x738px
>>905172
>He hasn't heard of hydraulic motors.
>>
>>905160

Can viable actuators be self made? Is it a difficulty in manufacturing or a materials issue that makes them cost so much?
>>
>>905162
Haha funny star trek meme

>>905225
Yes you take an electric motor and hook it up to a harmonic drive or other exotic speed reducer. Making such exotic speed reducers would be a project in and of itself. Machining them is hard

>>905172
>>905185
>> hydraulics

Need expensive servovalves to control fast. Hydraulics suck too.

>> pneumatics
Are not even worth considering. The bandwidth is just too low.
>>
>>905247

I have no problem looking into making suitable actuators as a preliminary project. Do you know any good sources for becoming familiar with their engineering?
>>
>>905247
>The bandwidth is just too low.

Do you mean the compressibility of the air acts like a capacitor and essentially low pass filters the forces?
>>
>>905125
Joint wear and battery life
>>
>>905261
Well step one is learning about exotic speed reducers like harmonic drives and cycloidal drivea. Harmonic drives I am not sure how you would make(except for shitty ones out of plastic) without expensive tooling. Cycloidal drives you will need to get the tolerances right: http://hackaday.com/2015/05/23/hackaday-fail-a-3000-prototype-that-doesnt-work/

It may not be pretty and you'll have to do some mechanical design to make this work, but you can get pretty high reductions with ballscrews. NASA's valkyrie robot uses ballscrews to obtain high torque density/bandwidth. Unlike everything else, you can buy ballscrews from misumi.

If you are willing to design/build your own electric motor you could make something like the actuators in the MIT cheetah robot:
https://biomimetics.mit.edu/research/optimal-actuator

None of these things are easy. Ballscrews are your best bet unless you can find a way to buy harmonic or cycloidal drives. And after all this you still need to figure out how to get torque control because that's like the basis of exoskeletons

>>905281
No it means that the frequencies at which they operate are lower than the human walking gait frequency. Meaning you can't get them to move with you. Yes this is because of the compressibility of air.
>>
>>905306
So you can't do slow stuff with pneumatics?
>>
>>905309

You could, but then you're hindered when trying to do anything else.

Really, pneumatic systems don't really have any huge advantages over hydraulic ones, but hydraulic systems have some notable points over pneumatic ones, including the critical one here regarding responsiveness

About the only kinda decent advantage of pneumatics is that your working fluid is (usually) just atmosphere, and so you don't need a dedicated return line and small leaks aren't a big deal.
>>
>>905306

Excellent, I'll definitely be looking into this.

And as far as using electro active polymers as muscle fiber goes, I'm correct in assuming that's 10+ years out before I could even do that in a PHD program?
>>
>>905321
Dammit everyone always thinks artificial muscles will solve everything. Yeah pretty much. You could work on them in your PHd program though
>>
>>905327

That would be a dream come true but I think I have realistic expectations rather than assuming they'll magically make gundams viable

So my other question in regards to this system is if it would be easier to design a bigger suit (power loader rather than t-45 armor)?
>>
>>905331
Well what are you trying to do with said exoskeleton?

I am not sure what a T-45 armor is, but it is very difficult to make exoskeletons that actually do anything useful.
>>
>>905125
If you need to ask any of these questions on /diy/ you won't ever build an exoskeleton.
>>
File: wallpaper_romel.jpg (123KB, 1366x768px) Image search: [iqdb] [SauceNao] [Google]
wallpaper_romel.jpg
123KB, 1366x768px
>>905125
https://vimeo.com/98084869
It depends what you want your exoskeleton to do. Do you want to be faster? See video above. Do you want to be stronger? Try looking at servos or nylon coated artificial muscle. See video below.
https://www.youtube.com/watch?v=wnHByFnGawM

And good luck to you. The hardest part I think about the exoskeleton is making a power-source light enough and powerful enough to last more than twelve minutes. And for each ounce you add to the weight of the battery the stronger your muscles have to be. Another thing to take into account is that even though your arms and legs can lift it does not mean your spine can take the weight of whatever you lift.
>>
File: s-l1000.jpg (52KB, 1000x753px) Image search: [iqdb] [SauceNao] [Google]
s-l1000.jpg
52KB, 1000x753px
>>905247
Even so, most advanced walking robots use hydraulics. They are a more natural substitute for emulating muscles than electric motors or even ball screws. I'm surprised hydraulics aren't more popular in amateur robotics.
>>
>>905393

Proof of concept. Find areas that could be improved upon with good research.

Getting ideas for my post grad stuff I suppose

>>905411

Not with that attitude
>>
>>905413
>I'm surprised hydraulics aren't more popular in amateur robotics.

It's a matter of them being a pain in the ass to deal with while offering little advantage at the scale amateur roboticists usually work at. Geared motors (ether steppers or motor/encoder combos) are easy to source, install, and control, while still providing adequate power output for most tasks.

Hydraulic servos, while offering insanely high force/weight ratios, require a compressor, hydraulic lines and couplings, and, ironically enough, electric servos to control the valving. It ends up being a significant difference in cost and complexity, while offering minimal upside to someone who isn't making anything that has to be very fast, very strong, or both.
>>
>>905414
>>Proof of concept
To do what? What are you trying to do with said exoskeleton

>>905413
>> most advanced walking robots
Define advanced walking robot. The only walking robots I know that use hydraulics are Boston dynamic's robots and HyQ. Other 'advanced walking robots' like StarlETH, ATRIAS, DRC-HUBO, valkyrie, thor, MIT's cheetah, iCub, etc don't use hydraulics.

>> a more natural substitute for emulating muscle
How are they a more natural substitute? What about them makes them more natural than anything else?

There are very good reasons not to just outright copy biology. Airplanes don't flap and don't have feathers.

It is not surprising that hydraulics aren't used much in hobbyist robotics. Hobby hydraulics are esoteric and only really used in RC construction equipment. They are expensive too.
>>
>>905414
>Getting ideas for my post grad stuff I suppose
In general, you don't get to decide what you work on in grad school. You will be doing what your advisor assigns you to work on.
>>
File: Winterized_T-51b.png (815KB, 1052x842px) Image search: [iqdb] [SauceNao] [Google]
Winterized_T-51b.png
815KB, 1052x842px
>>905393
>>
>>905420
Actually there's research that feathers and flapping wings could significantly improve efficiency of we could possibly build the materials.
>>
>>905463
Oh ok powered armor. It is really hard to do this because most exoskeletons end up increasing the metabolic cost rather than decreasing it. So you end up getting more fatigued than you would carrying the same amount of weight without an exoskeleton.

You really have to get the biomechanics right.

>>905470
For MAVs, but not for full size aircraft.
>>
>>905331
a powerloader is really just a forklift that can use the stairs, balance is an issue for regular forklifts, little less a machine with a lower back.

if you were looking to make it just be something to walk around in, yeah it'd probably be a bit easier, but it's also more likely to rip you in half if you messed up your tabbing in the code.
>>
>>905487
And why do you assume that powerloaders would be able to use stairs? Why is balance an issue for forklifts but not devices with a lower back?
>>
https://www.youtube.com/watch?v=VSzvVXiWkSg

It's actually not anything that hasn't been not not done before in the history of mankind and his derivatives.
>>
>>905560
Erm, because they have legs.

Duh.

Do I have to post the guy kicking the robot dog again?
>>
>>905591
And why are things with legs always capable of climbing stairs? A real powerloader might not be maneuverable enough to climb stairs or might not be able to obtain the step height necessary to climb stairs.

And balance is very much going to be an issue for a legged powerloader, after all it must be able to balance on one leg while carrying a load. Although if it has very large feet or inserts pegs into the floor(a la michael jackson) then balance would not be a problem. However, both greatly limit the device.
>>
>>905591
Also, the robot dog self righting requires the ability to take a lot of steps fast. Powerloaders may not be capable of doing this.
>>
>>905650
It doesn't self-right; it doesn't fall over in the first place.

You're acting like the actual problem is intractable, when it clearly isn't, as sauropods, elephants, etc. demonstrate.

The Cretaceous was not eighty-million years of big-ass dinosaurs rolling along the ground screaming "Help! I've fallen, and I can't. get. up."
>>
>>905128
>programming them to move in tandem with your body without breaking any of your bones.

Why doesn't my car's power steering break my thumbs?

You are over-thinking the control issues. you move your arm against a force/position sensor and the actuators move to minimize the force and position error. Minimal logic to stop actuators when failure modes are detected. Job done.
>>
>>905660
Recovering from a kick(or other disturbance) requires taking a lot of steps fast.

And for the record, are powerloaders not machines based on the human form? Not quadrupedal robots with manipulators nor bipedal robots with 'tails' (for counterbalancing the weight being carried)

>>905671
Force/torque control is harder than it looks. Pushing against a force sensor is not a good way to go, because humans are made of flexible meat. It is better to control the torque directly on the actuators. But yes, once you have torque controlled actuators the control logic isn't that bad.

One method is to apply virtual forces to cancel out weight and friction which you can translate into joint torques. All you need to do this is a jacobian
>>
>>905225
you could consider using solenoid actuators

since the power you provide them will be directly linked with their motion

one downside is having to have a large battery and/or capacitors
>>
>>905719
solenoid actuators actually wouldn't cost too much to make

an extremely powerful one could be from $100- $500
>>
>>905719
>>905720
You have got to be joking mate. Solenoids have very bad force densities and low displacements.
>>
>>905647
in order to fix this you just need a smaller minimum stroke in your actuators


which is literally just making the upper leg compress more

think about how you move your upper leg up and down
>>
>>905682
For the record, why do you think that matters?

The situation is the same for any bipedal object, be it a T-rex, a power loader or an office partition: if the center of gravity is within the footprint, it balaces; if it's without, it falls over.

Why do you think tails have anything to do with this?
>>
>>905682
the powerloader just needs to compensate by shifting its weight

if it has a spine of some sort or a gyro or even a weight shifting pneumatic system it could do this
>>
File: image.jpg (74KB, 736x490px) Image search: [iqdb] [SauceNao] [Google]
image.jpg
74KB, 736x490px
so here's a question: why not change the initial design parameters from a motorized system that's supposed to augment a humans strength to a robot that just has a human pilot using haptic feedback to control it?

I know that sounds like semantics (and maybe it is) but it seems more effective than trying to make a suit that fully encloses the operator. The legs will need some very careful designing but you can still rely on the human for balancing and control rather than a traditional mecha design which just assumes ridiculously effective programming.

Pic related for the concept, not necessarily anything specific.
>>
>>905587
My vacuum cleaner does something similar when I push the grip forward and back.
>>
>>905158
Just about every microcontroller as of 5 years ago has enough processing power to use C, and if you feel like spending a little bit of money you can even get away with Python or Java
>>
>>906155
it is still very very hard, only twice as hard(twice as many actuators), but with less of a chance of chafing(a problem with exoskeletons they don't tell you about).

>> the human for balancing and control
lag kills this in the real world. Lag really fucks up haptics.
>>
>>906325

I love that pic
>>
File: 1433639244423.webm (424KB, 1280x720px) Image search: [iqdb] [SauceNao] [Google]
1433639244423.webm
424KB, 1280x720px
>>906325
Does anyone have more Darpa fails?
>>
>>905125
>where the bottle necks on development are.

power supply and lack of linear actuators that function like human muscle.
>>
>>906409
God fucking dammit. We don't need linear actuators. What we want out is rotary motion so rotary actuators work. The problem is getting actuators with high torque density.
>>
Two main problems of powered exoskeletons.

1. man machine interface. - typical suits have slow reaction times to your movements. you also do not want the suit to injury the wearer. By forcing a joint in the wrong direction or not providing enough assist or too much assist. which means suits are slow for now.

2. power. enough batteries to allow for hours of constant use would be cost and weight prohibitive. small turbine or piston engines spinning a generator add fire, fume, and noise hazard.
>>
>>906813
m8 have u heard of fusion cores? They'd be perfect for this.
>>
>>905749

Because the center of gravity is constantly shifting.

At least it is in humans, and when you strap a bunch of shit to yourself, the little shifts in the center of gravity your body is used too become a bigger problem and harder to counterbalance. Unless you make lead soles or something so heavy at the bottom of your feet that it would be literally impossible to topple you.
>>
>>906927

Fusion cores? I heard they only last like 10 minutes if your running with them.
>>
>>905671
Because the output is at the wheels, not strapped on your body.
>>
>>906948
Even less with a jetpack
>>
you are still limited in strength by your skeletal system
>>
File: 455445.png (2MB, 1280x1024px) Image search: [iqdb] [SauceNao] [Google]
455445.png
2MB, 1280x1024px
>>905125
Completely feasible, but it really is determined by 3 things. Cost, Skills, and Intended Use.

https://www.youtube.com/watch?v=ZMxNLCEIcbU
https://www.youtube.com/channel/UCVP1PTBbRGpmTQE1oQx8xNw

Just make sure you put limiters on it as part of your first step. That way it doesn't snap you in half. Second thing is to use an onboard power supply. Either generated onboard or battery powered.
>>
Start with an arm, perfect it, then move on. don't go for an entire suit to begin with
>>
also i bet the guy from /diy/ who made the 3d printed prosthetic finger would be a big help here
>>
File: inside-job_0.jpg (224KB, 1200x1801px) Image search: [iqdb] [SauceNao] [Google]
inside-job_0.jpg
224KB, 1200x1801px
>>907244
The exoskeleton actually takes all the force and weight. They are essentially just micro mecha.
>>
File: 7536b3308d32.jpg (45KB, 500x444px) Image search: [iqdb] [SauceNao] [Google]
7536b3308d32.jpg
45KB, 500x444px
I really prefer tracked or wheeled mecha instead of exosuits.

http://www.abovetopsecret.com/forum/thread587951/pg1
>>
>>906948
Not relevant when they literally weigh nothing, have zero volume so that you can carry hundreds at a time, and your exoskeleton changes them automatically.
>>
>>907415
what would be the danger of having fusion reactions taking black on your back?
>>
Literally just power, cochise. They're feasible as fuck.
>>
>>907418
It is no more difficult than building your own car really. Much simpler in fact.
>>
>>907416
Vault dwellers might shoot you in the reactor with their matrix hax bullshit. Even if they have to shoot you through your chest to hit the core, they will hit the core.
>>
>>905131
Don't start with C and some tiny embedded device. Massive mistake.

Start with an entire powerful PC wired to the system and a nice easy language like python or C#.

Master the control routines in your over powered easy to debug computer. Then you translate to the smallest embedded system that can support it.

Same with power. Start with external power. Make sure it actually works, then add portable power.

You only want to be debugging one system at a time, otherwise you won't know if your problem is due to software, hardware, power, motors, communications, firmware...

PROTOTYPE
>>
>>907845
>Start with an entire powerful PC wired to the system and a nice easy language like python or C#.

>Tells him to use an easy language
>tells him to use convoluted Meme# over good old C in the same breath
I'm sure he'd love to be making 56 character function calls using those oh so nifty OOP class and object things instead of using almost-as-simple-as-BASIC C to do the same job in 20 characters with five letter variable and function calls. He should learn in BASIC and do in C, or at least give him a PICAXE.
>>
>>907863
What, do you think using one-letter variable names makes your code go faster?
>>
>>907877
Do you think hiding functions inside of classes inside of objects inside of containers inside of a fucking bag of holding does?
>>
>>907878
It helps write code faster, and more correct.
>>
>>907881
>actually believing this
>especially in the context of embedded programming
It adds on an unnecessary layer of complication and shoots readability to hell. It's much easier to learn procedural programming than learning OOP. ESPECIALLY microshaft's meme language.
>>
>>907882
You're trying to make this procedural-vs-OOP, when the real issue is managed-vs-unmanaged.

Telling a beginner to start off with an unmanaged language is just dumb.

Blitheringly, pants-on-head dumb.
>>
>>907883
How fucking so when that garbage didn't even exist until the mid 2000s in any relevant capacity? Why even bother learning C# when most mcus have native compilers for C and C++ instead? He's not going to be writing webapps or videogames, he's learning to write firmware for actuators.
>>
File: 1435267381153.jpg (10KB, 145x312px) Image search: [iqdb] [SauceNao] [Google]
1435267381153.jpg
10KB, 145x312px
>>907884
He should be starting with assembly coding.
>>
>>907884
You're essentially demanding that the OP starts out getting bogged down in irrelevant details.
>>
>>907890
That's the exact opposite of what I'm doing.
>>
>>907891
No, it's the exact opposite of the exact opposite of what you're telling him.

>Dear OP, rather than writing code that does stuff, you need to write code that handles locking and memory management. Make sure you do it perfectly, or your program gets terminated.
>>
OP I think you can do it. You should be able to make cycloidal drives provided you have access to a CNC machine:

http://lcamtuf.coredump.cx/cycloid/
http://www.zincland.com/hypocycloid/

Now if you really want to get adventurous you could use water cooled brushless rc boat motors driven by ultracapacitors just like SCHAFT uses.(you still need a speed reducer) Water cooled brushless motors have been used to get high torque densities. No one has ever tried RC boat motors for robots, but electric cart racers use em...
>>
As far as the threat of injury goes couldn't you just mechanically make it so the exoskeleton is physically unable to bend a way that your body is unable?
>>
>>906325
Is it a processing speed problem or a communication problem?
>>
>>909252
Yes. They are basically just pieces of metal that block it. The best method is to have those blockers coupled with something that completely cuts power for that direction.
>>
>>909255
Communication.
>>
>>907886
>>907883
>>907882
>>907878
>>907863
>>907845

So pic related runs some weird (almost) realtime variant of java. It won second place in the DARPA robotics challenge. So yes it is possible to use managed languages to control robots and stuff. Thing is a lot of really weird stuff had to be done to make this work. For one, they had to run it on a single processor to reduce jitter.

Another robot in the DRC used compiled matlab and it also did pretty well.

Why you would use C# for robotics is beyond me. Practically no one uses C# for robotics, so there aren't handy libraries for doing robotics stuff out there.

Now for an exoskeleton you are gonna need some embedded controllers for your actuators, this almost certainly means things that you can't compile c# to.
>>
The design and coding is absolutely trivial compared to the power requirements.

Just like damn near every other bit of neat future-tech, we have no decent power source to run it.
>>
>>909318
How are fuel cells compared to batteries in terms of 'capacitance' and voltage?
I heard that's what everyone is hoping the solution will be.
>>
>>909321
>>'capacitance' and voltage?
wat. So the main things that matter with fuel cells are power and efficiency.

Hydrogen storage, well that's another matter. Hydrogen has a great energy density on a per mass basis, but the problem is that it has a relatively low volumetric energy density(unless you use liquid hydrogen).

With modest improvements to hydrogen storage technology, the effective energy density, that is the work you get out from electric motors could equal that of gasoline.


However, there is the big issue of distributing all this hydrogen....
Thread posts: 96
Thread images: 15


[Boards: 3 / a / aco / adv / an / asp / b / bant / biz / c / can / cgl / ck / cm / co / cock / d / diy / e / fa / fap / fit / fitlit / g / gd / gif / h / hc / his / hm / hr / i / ic / int / jp / k / lgbt / lit / m / mlp / mlpol / mo / mtv / mu / n / news / o / out / outsoc / p / po / pol / qa / qst / r / r9k / s / s4s / sci / soc / sp / spa / t / tg / toy / trash / trv / tv / u / v / vg / vint / vip / vp / vr / w / wg / wsg / wsr / x / y] [Search | Top | Home]
Please support this website by donating Bitcoins to 16mKtbZiwW52BLkibtCr8jUg2KVUMTxVQ5
If a post contains copyrighted or illegal content, please click on that post's [Report] button and fill out a post removal request
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the content originated from that site. This means that 4Archive shows an archive of their content. If you need information for a Poster - contact them.