Hand: open-source Robot Hand | Hacker News Reader

Hand: open-source Robot Hand

github.com

444 pointsvineethy104 comments

Open articleView on HN

What jumps out at me is the $135 bill of materials.

What a time to be alive!

Most hand related jobs upwards of $100 per 30 min

If this Robot hand can do those jobs we could see some industries take a hit

https://m.youtube.com/watch?v=6pbgOmaBx34

People have been making end effectors using hobby servos for ages. These servomotors are designed for use in an RC aircraft, they're light, cheap, and expendable.

Industrial needs care not about weight, care less about cost, and care a great deal about capability, repeatability, and reliability.

This is a cool project for a hobbyist but it's not meant to be a serious industrial machine.

Edit: what is with this thread? Lots of very generic positive comments here but not much thinking about what this is actually useful for.

Woosh

oh....oh no....

I think the confusion stems from the fact that you're responding to a joke innuendo thread.

They just haven’t watched the Big Bang Theory episode with Howard and the robotic hand:

https://www.youtube.com/watch?v=OYryogNE8Ys

Ooooh. For _that_ application you better have either enough elasticity or reliable force/torque sensors and a good control in place.

...that's how it was in year 2020.

What you're missing is: today, we're nearing the point where actual general purpose robots become viable.

Which means: the purpose of a robot is no longer to sit at a factory line and precisely execute the same exact motions on repeat 24/7. The purpose of the next generation of robots is to learn generalized behaviors, adapt to circumstances, and carry out circumstance-specific actions with active sensor feedback. Which means completely different requirements for effectors.

Which means: repeatability can go get fucked, for one.

Without repeatability, good luck tuning your robot to do anything reliable.

Real world isn't "reliable". If a robot can't correct for errors, it's not going to survive out there.

> ...that's how it was in year 2020

Humanoid robotics research was pretty popular in the early 2000s already, with remarkable, reproducible results not only in videos. It’s definitively more present in the media now.

The bottleneck for general purpose robots is, and always was, AI.

We've been in a continuous AI breakthrough since 2022. In 2000, this kind of thing was nonexistent.

> Lots of very generic positive comments here but not much thinking about what this is actually useful for.

How can you say that when the person you are responding to is talking about what this is actually useful for?

It took me a a distressingly long time to understand this comment. I'm kind of concerned, and have vowed to get out more.

> upwards of $100 per 30 min

That is why I'm self employed.

Search for ‘teledildonics’ on YouTube for one of the most entertaining and comprehensive Emacs courses you’ll find on the internet.

I can't wait for this to be put on a tall roomba base so it can clean my kitchen.

Feetech is selling actuators which are mechanically R/C type servos, but have a bidirectional computer interface allowing the control computer to find out what's happening at the servo.[1] This isn't new; Dynamixel has been doing it for over a decade. But not at this price point. This Feetech servo is $17, while Dynamixel units start around $70 and go much higher.[2]

The parts list has "need to be strong" for many of the small parts, but they are 3D printed PLA plastic. That's the low end of 3D printing. None of the videos show the hand handling anything.

So this is really the proof of concept model. If there's enough interest, someone could make the parts by injection-molding of something better, such as polycarbonate or glass-filled nylon. The total plastic volume here is so tiny that the plastic cost is negligible, and there's no reason not to use a high-quality engineering plastic.

Nobody seems to do hobbyist injection molding much. TechShop had a desktop injection molding machine, the CNC milling machines to make molds, and even Autodesk Moldflow to design them. But nobody used those tools. A few university maker spaces have similar machines. Because most of the world's plastic stuff is made by injection molding.

[1] https://www.feetechrc.com/

[2] https://www.robotis.us/dynamixel/

[3] https://makerspace.engineering.nyu.edu/machines/pim/

Mold design is still difficult when the parts aren't dead simple. The software I've seen is okay with the simple stuff, but once you get even a little more complex you have to understand simultaneously how to design good parts for molding and how to design good molds, both of which are heavily dependent on the type of plastic you're using and the size of the press you have access to. Not to mention how to machine good molds from metal, which is a challenge all on its own due to surface finish and tolerance requirements (and weird geometry that makes the CAM choke...)

In other words, we're not really there yet to bring that activity into the hobby realm. But I hope that we're not too far away.

They're little linkage parts, mostly flat.[1] Some of those holes are bearings. None of those parts are hard to make, but they need to be strong. They could be made by CNC machining, or in quantity by injection molding, or stamping. But tiny working parts in 3D printed PLA will be too flimsy for that hand to do much work.

Totally fixable problem. Then this hand can go to work.

If this thing catches on, someone might sell an upgrade kit with stronger parts. The designer is already considering a servo upgrade.

[1] https://github.com/pollen-robotics/AmazingHand/blob/main/ass...

I haven't looked too carefully, but I suspect those "needs to be strong" parts could mostly be cnc machined with an inexpensive hobby CNC like a 3020, from aluminiumn sheet/bar stock.

That's what I was thinking, too. All but one of those parts can be easily milled from flat plate. The yoke end is a bit harder, but yoke ends are cheap standard parts and there's probably an off the shelf solution. The ball joints they use are standard R/C car parts. Those are available in in a stronger form, nylon with a brass ball, for EUR 0.53.[1] It won't be hard to strengthen this design so it can do useful work.

[1] https://toni-clark-shop.com/Ball-Joint-for-M2-and-M3

Smiles from Detroit. I would look to press those parts, since most contact surfaces are potentially convex and variant-depth parts appear superfluous to requirements. With a shared punch and die you could pull the nest off a laser, press them all in parallel, and break them after the parallel press-forming. Should be cheaper than injection because the cycles can be faster, the material can be stronger, there's only one die, and the unload will be easier?

Yes. Here are very similar parts produced with a punch press.[1] In Silicon Valley. Or in China.[2]

This sort of thing is probably about $40,000 for the first part, $0.05 for each additional part. Designing and making the custom dies is expensive. Banging out the parts is cheap. Mass production works.

[1] https://www.petersenprecision.com/fine-blanking

[2] https://yonglihaomachinery.com/

You haven’t accounted for the 3D printer yet.

It’s one of best designs I have seen, I admit. But for that price you cannot get absolute encoders outside the motor, reliable force/torque sensors (think picking up a strawberry), tendons (thread below). It might be too limited for research and real-world projects unfortunately.

Would adding AnySkin to the finger tips work? ( https://any-skin.github.io ) ?

Depends entirely on the application I guess, but you'd get further for sure. Either way you are now confronted with more complex cable management and maintenance, and additional weight. You'd at least get around gravity compensation with that approach. Personally, I think you need a lot more sensors and complex sensor fusion to achieve acceptable dexterity to be useful outside of an industrial context and this comes at high costs

Maybe this could be an add-on option for the K-Scale robot that is coming out soon? They want $1k for a 5 finger end effector hand.

https://www.kscale.dev/

Here is what I’m more keen on, rather than the human-like robots that we are all expecting. For instance, I would like a wall-mounted or floor-standing multi-arm robot that serves as a kitchen assistant. One can add or reduce arms as needed/desired. It is custom-equipped with a fire extinguisher, thermometer, and the usual must-haves for a kitchen. It will hold the cutlery, plates, and other items as needed. It will also advise on the likes of, “No, salts usually go in a pinch, would you like me to add in just about 5 grams?”

Thus, similarly for the garage, the DIY table, etc. Just Arms would be good.

Tbh I would rather not have computer-guided knife-flinging arms in my home, be them on wheels or fixed to the wall.

I have a parrot (whose beak is sharp enough, thank you very much) that loves to grab a knife out of the knife block and spin around with it.

I have to put a towel over it though today he pulled the towel off and still grabbed the knife and was holding it up when I turned around.

It's a well known fact that birds aren't real and you must be a target now.

Like a mischievous toddler with wings...

I always imagined robot hands hanging and sliding on rails under top kitchen cabinets.

perhaps sir would prefer… tentacles?

It slices, it dices...

Seriously, though. Vassar Robotics (YC company) has an arm kit available for order now. The original ship date for my order just got pushed back due to an upgrade in the camera spec.

It won't be able to hold knives (I don't think) but there are companies working to bring about your hoped-for wall arm right now.

I don't want arms in the kitchen just yet though I'm happy for someone else to beta test that.

But arms to move laundry from dirty clothes -> washer -> dryer -> clean clothes basket? Relatively defined, repeatable interactions with a worst case of wrecking some clothes seems a safer place to start.

I think the washer-dryer integrated unit handles the first bit without moving the clothes. Add the ability to dump dry clothes into a basket and voila!

Pollen Robotics and HuggingFace are doing a lot for robotics right now!

Wonder if it will get adopted (huggingface robot) I noticed the eyes/cameras go behind the neck thing for sleep mode

Looks like this product is called AmazingHand (and there are billions of "hand"s in the world), so the title might have some room for improvement as far as searchability goes.

Would it be possible to have tendons running through the arms, so the weight of the hand is reduced?

Most tendons materials are elastic. That lead to create calibration problems and require proprioceptive sensors in the hand

Huh, never thought of that

I wonder if companies are experimenting with materials like UHMWPE for non-elastic, high strength-to-diameter tendons.

I dont know if you'd have to weld the dyneema to the anchor points, though

I think you're misunderstanding the essence of the problem. If you use tendons, you'll need a neural network in your control loop that can learn continuously so that it compensates rope stretch and changes in friction through wear and tear.

A lot of problems in robotics reduce down to continual learning. Essentially all system identification tasks become obsolete the moment you have a self learning system and yet we have an AI industry preaching that AGI is around the corner without this "crutch".

Not to mention that they easily break and are annoyingly tedious to be replaced (I had to assist often enough on such "surgeries" in our lab)

You could add encoder patterns to the tendons, like stripes. And then use something like a mouse-sensor to track them. This is still much lighter than servos.

Another idea is to use an external camera or two and to track the fingers with a deep learning model. But this can become messy if other objects are in view. And it might also introduce more control delay in the feedback loop than a simple sensor.

If space permits you should add absolute encoders in the joints. In a finger that’s likely a challenge. In an arm, that’s easy

You don't necessarily need neural networks for that, there are more specialized function approximations that learn faster and enable life-long learning for kinematics/dynamics and extrapolate better than NN which are more general purpose.

I would think pressure and angle sensors tied to a PID loop would suffice, but I'm not a roboticist.

If you can get a feedback loop at the respective joint, then yes, this might be enough. In case of an arm (not this hand), you can often only observe the end of a chain, and then, that’s more complex

Roboy used dyneema tendons if I recall correctly. Fluidic actuators is another option. IMO, additional sensors and sensor fusion are necessary but this will raise the costs and demand to control software significantly. We are researching humanoid robots for quite some decades now and these problems are easily underestimated (similarly to autonomous driving). I doubt we'll see them in our houses very soon.

The issue is not that these problems are easily underestimated, but that the researchers are very proficient at repeating the same mistakes over and over again.

Or both …

In practice force sensing is more useful than proprioception in most cases, at least for grasping. Generally you won't the dimensions of an object you're attempting to grasp anywhere near as well as you would know the shape of the hand you're using so a certain amount of underactuated compliance makes the job a lot easier.

I agree. But even then elasticity had to be taken into account (and maybe even gravity), if the object is very delicate.

Will Cogley has had a few different designs with tendoned hands.

https://www.youtube.com/@WillCogley

https://willcogley.notion.site/

So, there are designs out there for that too!

I wonder if I have time to make one of these and then decorate it to look like Thing for Halloween!

The world is designed with humans in mind, it's great to see robotics evolve in this direction to take advantage of that!

Why was this downvoted?

Literally the reason for all publicly traded robotics companies going up recently.

Anybody knows of similar projects for exoskeleton or support devices?

https://theopenexo.nau.edu/

thanks a lot

Yes, it is a technology originally intended to make spacesuits less brutal on astronauts hand/wrist/forearm fatigue.

Last I checked, project was shelved in 2020 for various reasons. =3

The bigger or biggest question is - how much weight it can lift? If we assume it can lift half a pound, then what changes it’d require to make it lift 10/20/30 pounds and so on?

hands hold, arms lift — a hand without an arm isn't going to have much strength

Yes, but this is "bring your own arm" so the person above you can easily build the arm out to whatever specs they'd like.

They probably want to know relatively important information like

- Breaking force (how much force will break a finger)

- grip force (How much force can the fingers exert to hold an object once closed)

- holding force (combination of grip force and material properties [ex - friction] that gives you an idea how much force can be applied to prevent slipping)

- closing force (How much force is exerted during closing [similar but distinct from grip/holding])

Or, with a lot less specific detail but still generally useful as a starting point...

- payload capacity (approximately how much can this safely manipulate)

This looks like a nice, approachable robotic model of a human hand that can be printed and experimented with.

But... is a human hand the best design for a robot to grip things with? Or could we surmise that the human hand evolved as a pretty good hand given the materials and senses that were available to evolve humans from, while in theory a totally different design might be optimal for gripping when constructed from metal, plastic, motors, etc.?

It’s easier for humans to train a human hand

Depends on what you mean by best ofc :P

If anything, the human world is built for humans, so a lot of existing things are naturally compatible with human hands. Also, take into account flexibility. It might not be the best for one job, but it's really okay at a lot of jobs.

Is there a better design? I ask this genuinely, I really don't know but I suspect that the human hand is versatile enough to allow it to be programmable for a variety of tasks.

You can make it switchable! Think of it more like a dye, if that helps.

Our modern world was built for hand-havers, so we build hand-havers for general interaction with the modern world. This vicious cycle will end, but a certain amount of inertial clunkiness is inevitable in nascent technological disruptions. Especially in moments as grand as the second industrial revolution.

Would you prefer necrobotics!?!

https://www.smithsonianmag.com/smart-news/scientists-use-dea...

How long until human hands are put on robotics??

This is the craziest thing I’ve seen in awhile

A human hand is probably the most appropriate design for a robot to grip a variety of things that were designed to be gripped by human hands.

For any one specific thing, be it a doorknob, a rope, a sheet of paper or fabric, or a pair of scissors, there's probably a different design that's several orders of magnitude simpler and cheaper, and also much stronger and more reliable. Single-axis parallel grippers, circumferential chucks, vacuum cups/vacuum pads, electromagnets, cam lock and release mechanisms, and so on are common in industrial robotics.

Assume your robot's only task is to grab a spool with a 35 +/-0.5 mm ID core from from an infeed rack and place it on a spindle, you're not going to try to build a five-finger human sized servo-operated hand and tuck two of those fingers away to awkwardly pinch outwards from the inside, you're going to grab a Schunk JGZ concentric gripper off the shelf and plumb a pair of air lines to it. If it also needs to grab a tab from some tape on the spool and pull it into the machine, you're just going to add an asymmetric pincer like an angular tumor on two of the jaws - or graft on an entire separate parallel gripper like some polydactyl appendage, or tool-change, amputating and reattaching hands at will.

I have also observed that humans are quite good at anthropomorphizing robot arms: a small, well-tuned motion can be universally recognized as a nod of agreement, shrug of confusion, wave of acknowledgement, or sigh of disappointment, even if the equipment is a bright yellow 6-axis piece of cast iron with menacing claws where the hand (or face? they're often the same) should be. Googley eyes and a "Hi my name is" sticker make this even more convincing.

But if you need a single tool to grip a doorknob, a rope, a sheet of paper, a pair of scissors, AND an unknown variety of other arbitrary household objects... it's probably best to start with an approximation of the human hand. Also, while claws may be appropriate for a work environment with the robot inside a fence, in collaborative situations hands are just less intimidating.

Good point, well put.

> For any one specific thing, be it a doorknob, a rope, a sheet of paper or fabric, or a pair of scissors, there's probably a different design that's several orders of magnitude simpler and cheaper, and also much stronger and more reliable.

Also, if you’re designing a robot gripper for any one specific thing, it’s quite possible that you can tweak the design of that specific thing to make the task easier. As an extreme example, screws and screw drivers evolve in parallel.

As an aside, this "robot tentacle" paper was referenced in a recent HN story: "SpiRobs: Logarithmic Spiral-shaped Robots for Versatile Grasping Across Scales"[1]

Seems like a pretty high bang-for-the-buck for versatility and capability with only a few cables controlling it.

[1] https://arxiv.org/pdf/2303.09861

Why should not we build robots that mimic us?

Among other things, the ability to pretrain for a task by just copying human motion is pretty powerful.

I agree that there are some robotic designs that unnecessarily mimic human limbs. I have in mind heads, and feet (instead of wheels).

A hand however, is useful because so many manufactured objects have been constructed for their purpose.

Feet are used for roughly the same reason a human-like hand is preferable - human-designed spaces tend to not be perfectly compatible with wheeled locomotion.

The ability to negotiate stairs is table stakes for a household robot. It's already a pain when one's Roomba-like is defeated by a small ledge...

Well we aren't all wearing mecha-claws to improve upon our feeble human design.

The human hand is arguably the best general purpose gripper of human-scale objects. Only took evolution a couple of hundred million years to figure it out.

If you can limit the scope of things to be gripped, e.g. a sheet of paper, a baby chicken or a 100x100mm square steel girder then no doubt there is a better design out there.

Not exactly. The human hand is really advanced within the constraint of "you can't just arbitrarily replace damaged parts". If you can swap in replacement fingers, 3 of them is fine (and much easier to model and perform grasping calculations.)

The question you are asking opens up a whole other field of questions. The amount of AI and robotics "in the field" is only going to increase. As that increase comes do we want to continue to build for a human capabilities and limitations, or do we want to build for machine capabilities and limitations.

I think the ethical approach is to build for human capabilities and limitations. We have already seen what happens when we allow business to optimize for the lowest common denominator, and that is why we have regulations that emphasize accessibility. If we allow or encourage businesses to build robots that lack human capabilities and limitations that operate in the real world alongside humans, then even if those robots are assistive in nature (either a prosthetic robot hand, or a full blown humanoid robotic assistant), we will displace or redefine what humans are capable of, and diminish the role of and respect for human beings in our society.

This is great, but to make a comparable hand, we also need very sensitive sensors, at minimum pressure and temperature, across the entire surface area.

Absolutely, though I'd be ok with just pressure as a starter.

Does anyone know what design considerations, if any, might've gone into deciding to have 4 fingers instead of 5? And what tradeoffs that entails?

I actually saw this posted a few days back on Twitter and had been wondering if there was any deeper consideration for the number of fingers. It seems like you save around $10 in parts by getting rid of a finger, based on the information in the BOM.

It looks like the width of the servos driving each finger makes a certain finger spacing necessary, and I bet five made for an awkwardly wide hand.

Ignore the fact the hand is as thick as a servo motor, for now, and just consider the palm width.

This 4-finger hand is just below 4 inches wide, across the knuckles. By the standards of gloves for humans, this is already "X-Large"

The SCS0009 servos are 1/2in wide and there are two for each finger. Adding two more servos would make it 5 inches wide, taking it into "XXX-Large" territory.

Beautiful design and I love that it’s in onshape with an aliexpress BOM. I might build this!

I love that it's cartoon style -- 3 fingers and a thumb instead of four

This is what open source hardware should look like: fully accessible, modifiable, and fit for iterative improvement. It’s no longer about closed “prod tech” it’s about community driven evolution.

Cool project. I've heard that robotic hands should have finger lengths that are not equal for more dexterity. The variation in finger length of a human hand is a feature not a bug.

I swear I've seen this before somewhere..

Great stuff. So 2x$135 and I'll finally get my t-shirts folded for me in the laundry room.

Looks cool. Is there any other open source projects similar to this?

what are some automation use-cases for something like this?

feels like a cool project/toy

Reply on news.ycombinator.com