I did a few hands a few years ago (and actually should be working on one now...). What I ended up learning from the experience was that standard robot design approaches, and technologies, are quite limiting. For example hinges with axles through them take up too much space, servos being really awkwardly shaped etc. I think the way to crack it is to combine soft robotic actuators, bones which slide past each other, and elastic and tensile elements. Have you looked at tensegrity inspired models of the body? It helps to see how the body can be broken into different elements beyond the typical rigid elements+hinges approach. A hand that is underactuated but held together with elastic and tensile elements can wrap itself around an object it wants to grasp. Have a look at the hand design by Zhe Xu.
Thanks for your comment, I'm actually really interested in Zhe Xu's designs, and I definitely think the next step for this project is what you're describing. What were you designing hands for if you don't mind me asking?
I made them mostly for robotic purposes as I want to build a whole humanoid robot at some point. There was some interest in prosthetics and I chatted with Openbionics back when they started. I also contributed to what became the Enable hand. Since then life got in the way, but also I've been looking at learning how biology approaches these problems and the pros/cons of biology vs standard engineering. There's a lot of cool research going on in soft robotics and programmable matter which could help us approach the design in a more organic way. There's a lot of ways that biology has made its mechanics optimised for it's role. I mentioned the gripping in the previous message, but there's stuff like passive dynamic walking, which helps explain why humans are much more efficient at walking than a robot like Asimo.
I've had a quick look at your videos, there's some fascinating stuff there particularly with your tensegrity spine, is that something you're still working on? I did this hand as a university project so there was a lot to cram into the year, but after seeing that project by Zhe Xu with the little crocheted joint capsules I've been really interested in that sort of thing. I'd like to be working on a new design now but as you say life is getting in the way haha. Let me know if there's somewhere I can follow your progress on your humanoid robot.
I have no experience in robotics apart from watching other people's work, my experience is mostly in electromagnetics, but i've often thought small electromagnets would be good for muscle/tendon replacements, do you both know of any work done in that area ?
This is so impressive. As someome who tried to design a hand in cad software, i extremely underestimated how difficult it was. I don't have the skill to make the complex, organic shapes of a hand like you did with the coverings
This makes me value my hand as the incredible piece of engineering that it is. And gives me hope because maybe we can improve it to further complexity and accuracy of motion
Mechanical tendons. Seems that is the only real way to accomplish full dexterity. The human hands are quite distinct. Great graduate project. I could see an entire robotic cyborg controlled this way and a central hub in the chest that self calibrates and controls the body functions.
I think a flywheel with electromagnetic clutches and capstan amplifiers as the central hub would equal or exceed human muscles unique response times, power and speed to superhuman levels. Maybe carbon nanofiber ropes could take the strain. When available.
When you start counting how many degrees of freedom we actually have, it gets pretty ridiculous. It's really awesome to see something as complex as a bio-mimetic a hand being tackled.
I watched this video like 25 times, it's so soft. the background music and your voice explainign something so interesting for me makes the 20 minutes feels like 5. I can't wait to see the finished proyect. Congratulations
Will, thank you for posting this. Several hours of researching robotic arms for a costume next year and this is one of the best bits of explanatory material I've found. The fact that you took time to break down your process, successes and failures all, has really helped confirm a lot of the concepts for me (and, ironically, only up the road from myself!) I wish you luck with the rest of your degree!
It's nice to see some one finally see the issue with hand prosthetics and design a hand that more human. You have a bright future. I wish I had the education to work with you dude.
It makes me feel good as a fellow human, that someone else is also still studying the robotic hand on their own. There is so much capability in one extra hand. You could have a mounted camera hand that could also record sound and pick up stuff.
Great work. I have seen bionic hands that work well with only two fingers and a thumb. Simplifies the design, reduces weight, complexity and failure points, simplifies software, reduces cost.
I've been thinking of this exact thing for years! It's amazing to see someone actually making it real. One idea I had was to (if you had the equipment, anyway) basically make a scan of the bones in your hand and then 3D-print exact replicas in order to better understand just how everything moves together, as I think that would do a lot to help figure it out.
This is really impressive, i especially liked how you used springs in the joints to make them open back up. I’ve been working on a hand project off and on for a couple years, and there are things in this design that make me want to change some things on mine. Sadly I’m limited since I’m recreating a design from a movie. I’ll be interested to see how you improve your designs in the future
I try a lot of things to build my industrial arm. Most of them don't work. I say this to let you know that I appreciate how hard your work is, and I admire your determination. Stay awesome!
You should consider using something like a flexible filament for the palm area with maybe strips of spring steel strategically placed to replicate the human palms movement. You also need to be careful to keep the lengths of the cable runs the same length, even if you actuate other things like the wrist. The cables should always be pretty taut. When designing linkages don't think about how the human body does it, you can't really do it that way because muscles are different than motors. Focus on the end resultant movemen. Overall the goal should be for design as simple and stiff as possible while still meeting your goals for realistic movement.
Good evening Will, I wanted to point out that it is precisely the discriminant between convergence and divergence of the metacarpal bones that is the biggest problem in designing and creating an artificial hand. The fact that you have correctly observed what is the natural mobility of the bones of the metacarpal, their semi-independence of movement and the different ability to articulate the last two fingers of the hand. All factors that must be calculated not only for size but also for exact structuring. The more they are similar to human bones, the greater the possibility that the limb responds naturally to its use. Congratulations, a greeting from a cybernetic lover.
On the spring in the thumb, try putting the spring inside a tube instead of around a post. The aspect ratio of the post to the hole that keeps it straight is not right. A tube can slide around a post that is the diameter of the spring. Then the diameter is big enough compared to the tolerance etc. that it will be much less likely to bind when moving. This is a great design so far! I want to see how it develops. I think you have one of the better designs among the bunch I have seen recently. I am thinking about making a robotic arm and I'm trying to figure out how complicated of an arm I want to make. Thanks for making this video.
Dude! You just rekindle my passion to design artificial hands for the people that don't have one. I want to create something similar to yours, but a little simpler so it is easy to use. I am frustrated on the fact that every hand on the market is limited to a little number of movements so i want to make one that does not feel fake. Unfortunately all of my ideas are only on paper, but i hope that soon i will have the time and resources to create one. Thanks!
7:58 I don't know if you still working on your hand, but simplest way to twist fingers is place, as first, joint to move finger left-right and second to move open-close. On the video you have that joints in other configuration and this is a reason why your finger not twist. I had the same problem on my project
I made something similar for my coursework in college and I used a similar approach of using servos and pulleys. But a big difference is my project had minimal set up and was controlled a different way. First of all I utilised the 3D printers ability to build supports, so in between my hinges the 3D printer had built supports and these could be "snapped" so that the hinge was now free to move. So I printed the whole hand except for the palm piece as a single print. The palm piece was printed separately so that I could easily thread the wires through the fingers and then later snap the palm in place. I printed the forearm like this also, so I could snap all of the parts together with no use for screws. I had only a basic open and close motion set up using a Myoware sensor so that when the user tensed their arm it would close the fingers. There were some pretty big issues with the design which I would change if I ever redesigned it along with the aesthetics, like yours looks a lot better than mine. Amazing work by the way, I really like the approach that you have taken. Keep up the good work!!!
What a beautiful project, beautiful work, a true artist. I think that the major problem in getting closer to the human hand is the motors, the lack of development of this segment, the development of muscles as muscles and not as engines, I believe that the development of muscles based on magnetic traction based on grouping of mutiltimagnetic cells would be the the development of the most advanced prototypes. but as you said, you are working with the technologies available today, congratulations again.
I think small electromagnets would make good muscle replacements but it's just a theory, i've never attempted any robotic development, i'm just in awe of channels like Will's.
Excellent. The thing is that magnetic force is fast but relatively weak and some kind of gearing is always necessary. If this could be done at the microscopic level it would be great, to have some kind of solid-state ultra strong and efficient solenoid type of actuators.
Hi! Good job on the hand! I did a similar project back in 2014 that was eventually scrapped before motorization, sadly. Instead of bolting everything together, we tried to hold everything together with an elastic mesh. We managed to build a condyloid joint, but I think that the proximalmetacarpal joints (including the thumb) are just hyperboloid joints. I still have a few kilometers of flexinol in my basement if you're (or anyone's) interested haha.
I love the thought process you shared. I drew up an idea like this, but changed to a hydraulic system before I even got enough done with the build to share it. I wish I had 3d printers back then. Oh, use this with a rail guide to move the metacarples laterally.
I am a korean student.I wanted to artificial leg design a from old times.And when these images come out, my dreams are more motivated.I am very excited now.I hope it will be a day when I can work for disabled people.Good video thanks.I'm sorry if I have the wrong sentence.
I appreciate your work, Your way of thinking of a realistic hand is almost same as mine. What my plan is to make the joints and skeleton structure same as human bone and joints are like ball and socket with same degree of freedom. Instead of human muscle and tendon cables will be used. Another thing I think that force /stress and torque of a human hand & fingers can generate Should be same for the robotic hand. Another thing I want to mention is that few years back I saw a video where all the cables at middle of the hand they were using a wooden special shape object to control all types of grips. So in that case my plan is to use lesser number of actuator to complete the project. Definitely what you have done so far is unbelievable work. I am very happy to see your this video. Keep it up Sir. 👍👍
2 cables per joint held under the same amount of tension, as the motor is actuated the infinity cable feeds one way as it pulls the other. Mimick the tendions that are in the actual human hand.
The metacarpal movements are dependant on the tendons or wires that move the fingers and the muscle on the outside of the hand, correct movements can be achieved by placing springs or elastic bands between the metacarpal in a position that matches those cross metacarpal tendons on the human hand
I'm not certain that this will be helpful, however for the finger curling I imagine you could use something along the lines of the turning mechanism in a longboard. I've seen some that use rather wonky looking angled plates, so what I'm imaging here is an angled slide channel in which the finger mech mounts. Additionally I would recommend adding a bit of elasticity to the 'tendons' with either actual elastic cores or spring and slack configs, I've seen this provide torque control and safety functionality with limited complexity on some hand models. Alternatively you could monitor the motor current and attempt to match it to the torque curves of the servos, but that's a touch complex for early prototypes imo. This very well may not be particularly important with the current mock-up, however down the line if you transition to more powerful motors/stronger materials you run the risk of either breaking the hand, the tendon's, or whatever the subject may be holding without proper limiting of some sort. Almost forgot: personally I f***ng hate traditional hobby servos, so I'd also suggest in any final versions upgrading to brushless motors with planetary gear sets and optical encoders, the hobby servos have very little power and typically use potentiometer as encoders which hurt reliability. Planetary gear sets are actually easier to design WITHOUT back-drivability than with, which means you don't need to continuously pump current into your motors to keep the fingers taught. I'd suggest using low profile bldcs, which may even make the lower arm bit more compact. Switching from the brushed motors should also boost efficiency a touch, not bad for something of this sort. Very well done tho, I for one am impressed by the level of artistry you put into it, beyond the engineering it is truly a work of art.
I made a model 35 years ago while I was 15 using simular methods. I had the palm, but the scale was much smaller (around 2 inches). I finished the tendors (with difference methods) . However, since we didn't have small servos to pull so many fingers. Eventually I gave up, as I needed to continue my study. I haven't started again as I know we still don't have such small motors nowadays. In addition, the tendor may lossen and need adjustment
Just to point out actual hands, if this was to be used for a utility function, Hand's "resting position" is mostly closed. For specific reasons, having a hand that is spring closed, would benefit the mechanism when holding things for a long period of time.
Feeling and being high function is freedom from physical Disability's and limitations not many people really understand that. Extraordinary great job 👋😯
A really cool way to do this would be with hydraulics. Have a positive and negative pressure line running into the forearm and then you can have a 4 way reversing valve with a servo to change pressure depending on what you need. Could also have a clamping type thing to shut off pressure once you reach the desired position.
Очень круто, огромная работа и по анализу задачи и по проектированию, просто супер! Это лучший проект по возможностям кисти, который я видел. Подписался на канал.
out of all the designs I have seen. even though you do not like having all those parts. this is one of the most in depth and wonderfully designed models for a robotic hand I have come across. great work. remember incorporating 3d printed parts is key not being completely 3d printed unless you are going for non-functional. use them to proof and mold from for a final design but not as the final. even 3d printed metal (sintered) is not as strong as forged/cast, it is still just a rough draft. your design and work though is great, keep it up. if you fiddle a little more with your slicer setting you might also close up some of the small gaps in the parts between the walls (play with extrusion with a little and overlap) should help with model strength even more.
Will, one option you may want to explore is sand casting using your 3d printed parts to create aluminum versions of them. This is a very impressive design and you could reinforce those areas by 3d printing then pattern casting them. Lost form casting I believe is the correct term. here is a wiki link en.wikipedia.org/wiki/Pattern_(casting) if you're interested. Excellent job on this gives me hope in the future that there are still bright minds coming up nowadays I have been a mechanical designer going on 20 plus years now and you have a rare talent from my experience. Keep it up and the sky's the limit.
Hi Will. I also was trying to solve the problem of a curved ball grip. My solution was to drive each finger with three cables / channels. The centre to close the finger and two opposing channels to control the splaying and sideways contraction.
Maybe the actual anatomical layout of muscles, tendons, bone structures a real hand has may help you achieve the motion you want at around the 7:30 mark? idk if you've looked into that already but I think it'd be worth a shot. looks amazing by the way.
Thanks, I printed the shell parts in ABS on an FDM printer intending to smooth them with acetone but they were too thin, so it's actually hand-polished.
@@WillCogley I recently watched a Tested video with a lass that built armour, she said that diamond dust Dremel bits are really good for polishing 3D printed parts quickly. Might be a timesaver.
I must have missed that one, interesting. I used Dremel sanding drums to start with but they just sort of melted the plastic. Thanks for another tip haha
@@WillCogley dremels can often do that, if you want to use them on plastic you should look for model making rotary tools, its exactly the same but you can set them much more slower than a dremel (or at least the dremels I own). I had the same exact issue when I started polishing my model cars, I ruined a very beautyful corvette like that
When do you get the two axis points for the base of the thum, that iwill be awesome. Use the hyperextended thumb and then wag it back and forth and press it forward and back. The 3-D printing is very exciting. And amazingly the wrist only moves a few degrees side to side it mostly bends forward and backward a little further up the alarm.
simple fix to that you can use a flexible ballistic material to account for the bone flexibility. now however you could just straight up print everything in the bone shape like the hand itself. you can then greatly reduce the size of the joints then you should consider creating like a palette for each bone finger tip and then using the smallest tubes possible to disguise them as veins you could run your wires to your actuators. using a medical grade silicone you could then create artificial tendons and muscles now sure they won't serve really any purpose except for cosmetic. I did a hand myself back in highschool for robotics class and i had to address this issue way before i had the luxury of 3d printing. But however i will be working on a prototype under my already patient name but i will post how it is built and made and hopefully this will give you some ideas so you can decrease on parts and size.
Nice work! Have you looked into twisted string actuators for your tendons? Regarding your joints, I printed a hand a few years ago where the joints were ninja flex filament. That worked fairly good.
very good iteration of your design, i can almost already see the next one with fewer parts and more optimized due to 3D printing. Keep up the amazing project! (edit) your mechanical interpretation of the human had is fascinating
Great work man!! This is the first time i see soo much work focused on thumb movement and the look is just amazing i will keep going with the movement of the knuckles, because for complex things like writing or grabing small things are really usefull
I hope you've made progress, I'd love to see the final product in action. this could potentially replace other bionic limbs by being stronger with more mechanical parts rather than the electronic ones.
This makes me curious about how much more compact and quieter this would be with something like the new brushless motors and drive systems that have been introduced in the 3D printing world to replace stepper motors while still retaining positional accuracy.
Pow! Mind blown! I have been using Fusion 360 and have been 3d printing and designing for almost one month now with zero education in either and watching what your doing is very inspiring to me. If I keep on the path I'm taking, seeing what can be accomplished I can only imagine what can be done with this technology. I can't wait to see how your project evolves going forward, it's already so amazing! Have you looked at tenticals and thought maybe a finger tentical might be something to look into for a simpler design? Thanks for sharing and as always Keep Building👍
Thanks Todd, I'm really interested in some of those soft robotics designs like the tentacle you mentioned, I'm looking into artificial muscles too. Good luck with your projects :)
impressive and inspirational ,I ll try your finger joints in my project but still lacking of having ecentric control of flexion movement in a more efficent way ....
There is a certain beautiful crazyness when you are able to witness the fluide movement all through ones hand ,joints, muscle, ligaments, tendons, carteledge, and so on but in the long and short of things you said that you wanted to make it as real as possible or to skaile then what you need to do is make all of the ligaments muscle,bone ,carteledge, vanes for formulation joints and so on and stop focusing all of your attention on each individual joint and take a nice long and in-depth look at the tendons and ligaments as your reference of motion and I bet that all the rest will just kinda fall in place that is what I did when I made my sparton glouve prosthesis
What would help with the flexibility is using less Stiff parts, especially for the exo-skeleton and more flexible parts, hands are dexterous, so they need dexterous materials and components, like rubber.
Instead of using springs I would suggest using both ends of the cable for the push-pull force. I think you should need two servos per finger, three for the thumb and two for the wrist. Also I would dump the idea of feedpack pots - just to complicated, also I would design the joints to not use bolts, just as in a real hand. If you need any design guides... just look down on your hands xD
are you kidding me feedback pots are CRUCIAL if you want a functional hand that can perform tasks. The main brains of the machine need to know what its limbs are doing somehow!
@@artbyrobot1 not really, he's using servos (which have a built in encoder) so the feedback pots are a touch redundant, if you were using a complex kinematic model or a neural net you may make use of the pots, but his simple setup can't reasonably be expected to gain anything from having twice the number of encoders. ~Fyi I'm an ee
Well, this is really inspiring, i love the science, the electricity, robotics, all this kind of things, and besides, i just have 3 fingers on my left hand, i this really inspire me to learn first at all about the 3D printing, and this biometric science
My gosh! This is very very beautyfull and thought through design. Easily has place in top 5 prosthetic arm design ive seen! U designed it all by urself? Right? IF SO THEN I CANT BUT CONGRATULATE YOU. BRILIANT MIND. I LOVE YOUR SOLUTIONS. I can see few invonveniences, but all will be perfected in time. Great work. Keep it up! Prosthetic designs and 3d printing are very close to heart. Seeing someone accomplishing both this well is great! Good luck. I hope u will accomplish to refine this work to its peak! 👍👍
Hat off for the incredible amount of effort done. It would be nice to have a small package 1-wire 2D magnetic sensor like the 14bit ams, to make it easier to built it in each joint and routing the wires. When each joint can be measured, the servo can be replaced by smaller motors and the "twisting fishing line" force conversion + transmission, which reduce the size of the actuator and making the forearm smaller..
PLA is actually one of the strongest materials, although not tough, it has the highest strength of regular materials, unless you go up to some really expensive stuff like ultem or peek, those are insane but expensive af I've learned to love pla, it can be so strong with the correct geometry! btw really liked the video, that hand looks amazing!
I am very impressed with your work! You inspire me to work at a much higher level. I wish you had access to a metal printer and smaller stronger motors. Well done!!!!
our skin has lot to due with gripping ability too, you need to put a soft rubber surface over the entire hand. Now try getting the same gripping characteristics when the hand is wet! You'll need to start creating features in the rubber at the micro/nano level (think lizard hands). Soft robotics is hard
We are so far from a decent mechanical hand, it's heartbreaking. Having lost a hand, I know what a real marvel it is. I think we need to go to a bone and tendon model and work on adapting it to a prosthetic.
This is some cool stuff..... N nice use of 3d printing.... To get more sensitive output from end effector... You can use graphene glass.. or sensor.... It will give us more reliable signal... With low power consumption
Have you thought of using spring steel shim stock? It's cheap and you can play around with the various thicknesses. Also, why not use something like a heavy test fishing line for the tendons? Something like 60 lb braided line, then you wouldn't need cable housing and the slop with that mechanism. But you'd have to hook up the flexor and extensor sides, which work if you are using servos. Try printing with more shells and top/bottom layers, it might hide those gaps that let you see into the parts. One last thing, why not print in PETG? It's only slightly harder to print than PLA, but I've always had the best experience using it for mechanical parts.
I suspect that current nerves impulse sensing technology isn't as big a limitation as people think, if it's used right. I think that training an AI to take whatever nerve impulses that _can_ be read, as a cue for what type of motion the user is trying to perform, and matching it to the expected series of motor events needed to recreate that action with the biomimetic hand, would yield good results. Especially if the hand has sensors like multiple low resolution cameras/sonar/radar/lidar, and pressure and heat sensors-whatever is needed-to feed into the AI and give it more cues to better adapt any motion it's tasked to emulate, to the objects it supposed to manipulate. It might even be useful to imbed cheap contact microphones into the tips of the fingers where the grips are, to get an reading of the surface texture of an object being handled, by using the sound of the finger as it brushes against it. It could be combined with a contact pressure switch on each fingertip, to signal more clearly to the AI if a certain pressure threshold has been exceeded. In fact… You could even stack two or three contact pressure switches on top of eachother, each with a higher, specifically calibrated activation pressure, to allow sensing of multiple pressure thresholds accurately. You'd want to choose the exact activation thresholds for each carefully, to make the most of them. And it might simply be better to use a single, more expensive pressure sensor that can read any pressure (within reason) with decent accuracy, depending on how well the stacked pressure switches work in an actual design instead of my imagination. Where was I?… Contact microphones! Have you ever worn a stethoscope, and brushed the sound sensitive membrane across something? Try it, if you can! It produces a sound that's a bit like what you hear if you accidentally brush your eardrum with the tip of a cotton swab. Different textures should produce slightly different sounds, and if the tip of the finger can also tell that pressure is being applied to it, then it can act as a two different clues to the AI, that the fingertip has encountered a certain type of texture signature, instead of, say, picking up vibrations that are traveling through the structure of the hand from somewhere else and giving it a false positive. The more clues an AI is given, the more it can deduce. Same with humans. And like humans, an AI also needs quality training to do quality work. I suppose an alternative to using contact microphones to detect texture, might be to place a near sighted digital camera, with a preference for higher resolution instead of color or light value differentiation. Then two or three LEDs can be imbedded in the tip of the finger as well, surrounding the camera lens equidistantly, and programmed to blink in a pattern that syncs up with the frame rate of the camera, with only one LED on at a time, so that each LED is turned on in a cycle. This should illuminate the surface of an object that is extremely close, in such a way that the light will highlight the parts that are higher, more, from the direction of the LED light source. Then, when the next LED is turned on, and another frame of video is taken, the surface texture will be illuminated from a different direction. The deeper bits in the surface of the object, should be darker, and provide contrast. Using a group of frames taken from the camera, each with a differently angled light source, the AI should be able to puzzle out the surface texture. There are definitely a bunch of engineering problems that would need to be solved to make this work, and digital camera sensors aren't even close to as cheap as a contact microphone sensor and pressure sensor combined. But it's still an option, and you could probably still get five camera sensors of serviceable quality for under 300 USD, if the price of the Raspberry Pi camera modules are any indication. It'll be the scratch resistant lens cover that will set your budget back. Hope you liked my spit-balling of ideas. Keep being awesome!
The pressed in brass threads I think would have been better if epoxy was used vs. Melting them inplace. I also think since joints are using torsion springs to return to neutral you might be able to use kevlar threads to replace the cables.
You need a pallet 2 or a pallet 2 pro a well tuned printer some flexible filament some conductive filament and couple different pla+ and you my friend would be well away looking forward to see the next iteration
The metacarpal bone movement in the little and ring finger seem to allow the tips of the fingers to line up since they are shorter. Without this movement things like playing the guitar would probably not work so well.
As an idea, use not steel cables, but plastic ones, for example from fishing materials. They are more plastic, and in terms of strength they are not inferior
When we want to learn about robotics, we take a good look at ourselves and our biology. Then we realize how nature solved a problem that took thousands and thousands of years in the making. Then we realize how powerful hands are. Nerves, tendons, bones, chemicals that makes tensile structure stiff and so on. The lord settled us a challenge in nature. It's fascinating when we rediscover how complex the structure of the hand is. Robotics is a careful observation of nature. Advanced robotics is deeply rooted in our comprehension of biology.
Hey, I don't know if you know, but i like the idea of having string pulley's for the moving finger prototype. so you inspired me to make my own version of that :)
Cool project! BTW, I see you messing with those inserts in small parts... hold the iron upside down instead, and put the insert on its tip. Then press the part down on it. I use my old iron for this and chopped the tip shorter, so it won't stick out.
All my friends: I'm bored let's watch rap. Me: glued to your every word and rewinding if I don't understand. I was going to ask for STL files but about 4 minutes in I realized that even if I had every single piece in front of me, it would take me about a year to successfully complete it. Well done sir. You're a better robotics engineer than I.
the meta carple.. just ise a hinge between middle fnger and outer wrist. spring load. limited range.. thumb should be several angles..that work together.. find the radius of side motion.. wonderful work.
I did a few hands a few years ago (and actually should be working on one now...). What I ended up learning from the experience was that standard robot design approaches, and technologies, are quite limiting. For example hinges with axles through them take up too much space, servos being really awkwardly shaped etc. I think the way to crack it is to combine soft robotic actuators, bones which slide past each other, and elastic and tensile elements. Have you looked at tensegrity inspired models of the body? It helps to see how the body can be broken into different elements beyond the typical rigid elements+hinges approach. A hand that is underactuated but held together with elastic and tensile elements can wrap itself around an object it wants to grasp. Have a look at the hand design by Zhe Xu.
Thanks for your comment, I'm actually really interested in Zhe Xu's designs, and I definitely think the next step for this project is what you're describing. What were you designing hands for if you don't mind me asking?
I made them mostly for robotic purposes as I want to build a whole humanoid robot at some point. There was some interest in prosthetics and I chatted with Openbionics back when they started. I also contributed to what became the Enable hand. Since then life got in the way, but also I've been looking at learning how biology approaches these problems and the pros/cons of biology vs standard engineering. There's a lot of cool research going on in soft robotics and programmable matter which could help us approach the design in a more organic way. There's a lot of ways that biology has made its mechanics optimised for it's role. I mentioned the gripping in the previous message, but there's stuff like passive dynamic walking, which helps explain why humans are much more efficient at walking than a robot like Asimo.
I've had a quick look at your videos, there's some fascinating stuff there particularly with your tensegrity spine, is that something you're still working on? I did this hand as a university project so there was a lot to cram into the year, but after seeing that project by Zhe Xu with the little crocheted joint capsules I've been really interested in that sort of thing. I'd like to be working on a new design now but as you say life is getting in the way haha. Let me know if there's somewhere I can follow your progress on your humanoid robot.
I have no experience in robotics apart from watching other people's work, my experience is mostly in electromagnetics, but i've often thought small electromagnets would be good for muscle/tendon replacements, do you both know of any work done in that area ?
awesome guys
This is so impressive. As someome who tried to design a hand in cad software, i extremely underestimated how difficult it was. I don't have the skill to make the complex, organic shapes of a hand like you did with the coverings
This makes me value my hand as the incredible piece of engineering that it is. And gives me hope because maybe we can improve it to further complexity and accuracy of motion
The accuracy of the motion of your hand is mostly down to your brain.
Mechanical tendons. Seems that is the only real way to accomplish full dexterity. The human hands are quite distinct. Great graduate project. I could see an entire robotic cyborg controlled this way and a central hub in the chest that self calibrates and controls the body functions.
I think a flywheel with electromagnetic clutches and capstan amplifiers as the central hub would equal or exceed human muscles unique response times, power and speed to superhuman levels. Maybe carbon nanofiber ropes could take the strain. When available.
When you start counting how many degrees of freedom we actually have, it gets pretty ridiculous.
It's really awesome to see something as complex as a bio-mimetic a hand being tackled.
I watched this video like 25 times, it's so soft.
the background music and your voice explainign something so interesting for me makes the 20 minutes feels like 5.
I can't wait to see the finished proyect. Congratulations
Well designed, i like the cables in place of tendons.
Will, thank you for posting this. Several hours of researching robotic arms for a costume next year and this is one of the best bits of explanatory material I've found. The fact that you took time to break down your process, successes and failures all, has really helped confirm a lot of the concepts for me (and, ironically, only up the road from myself!)
I wish you luck with the rest of your degree!
It's nice to see some one finally see the issue with hand prosthetics and design a hand that more human. You have a bright future. I wish I had the education to work with you dude.
It makes me feel good as a fellow human, that someone else is also still studying the robotic hand on their own. There is so much capability in one extra hand. You could have a mounted camera hand that could also record sound and pick up stuff.
Great work. I have seen bionic hands that work well with only two fingers and a thumb. Simplifies the design, reduces weight, complexity and failure points, simplifies software, reduces cost.
I've been thinking of this exact thing for years! It's amazing to see someone actually making it real. One idea I had was to (if you had the equipment, anyway) basically make a scan of the bones in your hand and then 3D-print exact replicas in order to better understand just how everything moves together, as I think that would do a lot to help figure it out.
there are already free models online of scanned bones
This is really impressive, i especially liked how you used springs in the joints to make them open back up. I’ve been working on a hand project off and on for a couple years, and there are things in this design that make me want to change some things on mine. Sadly I’m limited since I’m recreating a design from a movie. I’ll be interested to see how you improve your designs in the future
I try a lot of things to build my industrial arm. Most of them don't work. I say this to let you know that I appreciate how hard your work is, and I admire your determination. Stay awesome!
You should consider using something like a flexible filament for the palm area with maybe strips of spring steel strategically placed to replicate the human palms movement. You also need to be careful to keep the lengths of the cable runs the same length, even if you actuate other things like the wrist. The cables should always be pretty taut. When designing linkages don't think about how the human body does it, you can't really do it that way because muscles are different than motors. Focus on the end resultant movemen. Overall the goal should be for design as simple and stiff as possible while still meeting your goals for realistic movement.
This is fantastic. Great job!
I'll be eager to see the next update.
Good evening Will, I wanted to point out that it is precisely the discriminant between convergence and divergence of the metacarpal bones that is the biggest problem in designing and creating an artificial hand.
The fact that you have correctly observed what is the natural mobility of the bones of the metacarpal, their semi-independence of movement and the different ability to articulate the last two fingers of the hand. All factors that must be calculated not only for size but also for exact structuring.
The more they are similar to human bones, the greater the possibility that the limb responds naturally to its use.
Congratulations, a greeting from a cybernetic lover.
On the spring in the thumb, try putting the spring inside a tube instead of around a post. The aspect ratio of the post to the hole that keeps it straight is not right. A tube can slide around a post that is the diameter of the spring. Then the diameter is big enough compared to the tolerance etc. that it will be much less likely to bind when moving.
This is a great design so far! I want to see how it develops. I think you have one of the better designs among the bunch I have seen recently. I am thinking about making a robotic arm and I'm trying to figure out how complicated of an arm I want to make. Thanks for making this video.
Dude! You just rekindle my passion to design artificial hands for the people that don't have one. I want to create something similar to yours, but a little simpler so it is easy to use. I am frustrated on the fact that every hand on the market is limited to a little number of movements so i want to make one that does not feel fake. Unfortunately all of my ideas are only on paper, but i hope that soon i will have the time and resources to create one. Thanks!
Thanks Vlad, I'm glad you like it. Best of luck with your own designs :)
@@WillCogley Thank you
7:58 I don't know if you still working on your hand, but simplest way to twist fingers is place, as first, joint to move finger left-right and second to move open-close. On the video you have that joints in other configuration and this is a reason why your finger not twist. I had the same problem on my project
I made something similar for my coursework in college and I used a similar approach of using servos and pulleys. But a big difference is my project had minimal set up and was controlled a different way. First of all I utilised the 3D printers ability to build supports, so in between my hinges the 3D printer had built supports and these could be "snapped" so that the hinge was now free to move. So I printed the whole hand except for the palm piece as a single print. The palm piece was printed separately so that I could easily thread the wires through the fingers and then later snap the palm in place. I printed the forearm like this also, so I could snap all of the parts together with no use for screws. I had only a basic open and close motion set up using a Myoware sensor so that when the user tensed their arm it would close the fingers. There were some pretty big issues with the design which I would change if I ever redesigned it along with the aesthetics, like yours looks a lot better than mine. Amazing work by the way, I really like the approach that you have taken. Keep up the good work!!!
7:55 impressive level of thought process and detail man.
Nice video and great to see someone try and progress technology that could help a whole lot of people.
Excelent work Mr. Cogley! Beautiful design, and the improvement of lateral rotation on fingers is terrific! Congratulations!!!
What a beautiful project, beautiful work, a true artist. I think that the major problem in getting closer to the human hand is the motors, the lack of development of this segment, the development of muscles as muscles and not as engines, I believe that the development of muscles based on magnetic traction based on grouping of mutiltimagnetic cells would be the the development of the most advanced prototypes. but as you said, you are working with the technologies available today, congratulations again.
Read your comment after making mine, re: magnetics.
I think small electromagnets would make good muscle replacements but it's just a theory, i've never attempted any robotic development, i'm just in awe of channels like Will's.
Excellent. The thing is that magnetic force is fast but relatively weak and some kind of gearing is always necessary. If this could be done at the microscopic level it would be great, to have some kind of solid-state ultra strong and efficient solenoid type of actuators.
Hi! Good job on the hand! I did a similar project back in 2014 that was eventually scrapped before motorization, sadly. Instead of bolting everything together, we tried to hold everything together with an elastic mesh. We managed to build a condyloid joint, but I think that the proximalmetacarpal joints (including the thumb) are just hyperboloid joints.
I still have a few kilometers of flexinol in my basement if you're (or anyone's) interested haha.
I love the thought process you shared. I drew up an idea like this, but changed to a hydraulic system before I even got enough done with the build to share it. I wish I had 3d printers back then. Oh, use this with a rail guide to move the metacarples laterally.
I am a korean student.I wanted to artificial leg design a from old times.And when these images come out, my dreams are more motivated.I am very excited now.I hope it will be a day when I can work for disabled people.Good video thanks.I'm sorry if I have the wrong sentence.
I love the design! 🤍
EDIT: I just want to add this. If I'll ever need a cybernetic hand/prosthesis, I would choose your design!
I appreciate your work, Your way of thinking of a realistic hand is almost same as mine. What my plan is to make the joints and skeleton structure same as human bone and joints are like ball and socket with same degree of freedom. Instead of human muscle and tendon cables will be used. Another thing I think that force /stress and torque of a human hand & fingers can generate
Should be same for the robotic hand.
Another thing I want to mention is that few years back I saw a video where all the cables at middle of the hand they were using a wooden special shape object to control all types of grips. So in that case my plan is to use lesser number of actuator to complete the project.
Definitely what you have done so far is unbelievable work. I am very happy to see your this video. Keep it up Sir. 👍👍
2 cables per joint held under the same amount of tension, as the motor is actuated the infinity cable feeds one way as it pulls the other. Mimick the tendions that are in the actual human hand.
thanks for helping other people. thanks for the tech. this helps the future. bionics and mechatronics is caring. good work. noble work.
The metacarpal movements are dependant on the tendons or wires that move the fingers and the muscle on the outside of the hand, correct movements can be achieved by placing springs or elastic bands between the metacarpal in a position that matches those cross metacarpal tendons on the human hand
Stellar work Will. Super tricky project to tackle. Love the great exploded animations as well. Good luck with it
I'm not certain that this will be helpful, however for the finger curling I imagine you could use something along the lines of the turning mechanism in a longboard. I've seen some that use rather wonky looking angled plates, so what I'm imaging here is an angled slide channel in which the finger mech mounts.
Additionally I would recommend adding a bit of elasticity to the 'tendons' with either actual elastic cores or spring and slack configs, I've seen this provide torque control and safety functionality with limited complexity on some hand models. Alternatively you could monitor the motor current and attempt to match it to the torque curves of the servos, but that's a touch complex for early prototypes imo. This very well may not be particularly important with the current mock-up, however down the line if you transition to more powerful motors/stronger materials you run the risk of either breaking the hand, the tendon's, or whatever the subject may be holding without proper limiting of some sort.
Almost forgot: personally I f***ng hate traditional hobby servos, so I'd also suggest in any final versions upgrading to brushless motors with planetary gear sets and optical encoders, the hobby servos have very little power and typically use potentiometer as encoders which hurt reliability. Planetary gear sets are actually easier to design WITHOUT back-drivability than with, which means you don't need to continuously pump current into your motors to keep the fingers taught. I'd suggest using low profile bldcs, which may even make the lower arm bit more compact. Switching from the brushed motors should also boost efficiency a touch, not bad for something of this sort.
Very well done tho, I for one am impressed by the level of artistry you put into it, beyond the engineering it is truly a work of art.
I made a model 35 years ago while I was 15 using simular methods. I had the palm, but the scale was much smaller (around 2 inches). I finished the tendors (with difference methods) . However, since we didn't have small servos to pull so many fingers. Eventually I gave up, as I needed to continue my study. I haven't started again as I know we still don't have such small motors nowadays. In addition, the tendor may lossen and need adjustment
Just to point out actual hands, if this was to be used for a utility function, Hand's "resting position" is mostly closed. For specific reasons, having a hand that is spring closed, would benefit the mechanism when holding things for a long period of time.
This is so cool! It has much less useless plastic and more realistic motion than other competitors! :D
I think this is the most Human like Hand I have ever seen. Good Work.
Feeling and being high function is freedom from physical Disability's and limitations not many people really understand that. Extraordinary great job 👋😯
Super cool project man! great job!
A really cool way to do this would be with hydraulics. Have a positive and negative pressure line running into the forearm and then you can have a 4 way reversing valve with a servo to change pressure depending on what you need. Could also have a clamping type thing to shut off pressure once you reach the desired position.
Очень круто, огромная работа и по анализу задачи и по проектированию, просто супер! Это лучший проект по возможностям кисти, который я видел. Подписался на канал.
I really like your work, its inspiring, specially the combination of functionality and good looking parts
out of all the designs I have seen. even though you do not like having all those parts. this is one of the most in depth and wonderfully designed models for a robotic hand I have come across. great work. remember incorporating 3d printed parts is key not being completely 3d printed unless you are going for non-functional. use them to proof and mold from for a final design but not as the final. even 3d printed metal (sintered) is not as strong as forged/cast, it is still just a rough draft. your design and work though is great, keep it up. if you fiddle a little more with your slicer setting you might also close up some of the small gaps in the parts between the walls (play with extrusion with a little and overlap) should help with model strength even more.
You should implement camera for item shape recognition. That hand could pickup different shapes..Look nice.
Will, one option you may want to explore is sand casting using your 3d printed parts to create aluminum versions of them. This is a very impressive design and you could reinforce those areas by 3d printing then pattern casting them. Lost form casting I believe is the correct term. here is a wiki link en.wikipedia.org/wiki/Pattern_(casting) if you're interested. Excellent job on this gives me hope in the future that there are still bright minds coming up nowadays I have been a mechanical designer going on 20 plus years now and you have a rare talent from my experience. Keep it up and the sky's the limit.
Hi Will. I also was trying to solve the problem of a curved ball grip. My solution was to drive each finger with three cables / channels. The centre to close the finger and two opposing channels to control the splaying and sideways contraction.
Maybe the actual anatomical layout of muscles, tendons, bone structures a real hand has may help you achieve the motion you want at around the 7:30 mark? idk if you've looked into that already but I think it'd be worth a shot. looks amazing by the way.
This looks amazing, one of the best bionic / robotic hands I've seen. Were the white covers FDM printed? How did you finish them?
Thanks, I printed the shell parts in ABS on an FDM printer intending to smooth them with acetone but they were too thin, so it's actually hand-polished.
@@WillCogley I recently watched a Tested video with a lass that built armour, she said that diamond dust Dremel bits are really good for polishing 3D printed parts quickly. Might be a timesaver.
I must have missed that one, interesting. I used Dremel sanding drums to start with but they just sort of melted the plastic. Thanks for another tip haha
@@WillCogley dremels can often do that, if you want to use them on plastic you should look for model making rotary tools, its exactly the same but you can set them much more slower than a dremel (or at least the dremels I own).
I had the same exact issue when I started polishing my model cars, I ruined a very beautyful corvette like that
When do you get the two axis points for the base of the thum, that iwill be awesome.
Use the hyperextended thumb and then wag it back and forth and press it forward and back. The 3-D printing is very exciting.
And amazingly the wrist only moves a few degrees side to side it mostly bends forward and backward a little further up the alarm.
simple fix to that you can use a flexible ballistic material to account for the bone flexibility. now however you could just straight up print everything in the bone shape like the hand itself. you can then greatly reduce the size of the joints then you should consider creating like a palette for each bone finger tip and then using the smallest tubes possible to disguise them as veins you could run your wires to your actuators. using a medical grade silicone you could then create artificial tendons and muscles now sure they won't serve really any purpose except for cosmetic. I did a hand myself back in highschool for robotics class and i had to address this issue way before i had the luxury of 3d printing. But however i will be working on a prototype under my already patient name but i will post how it is built and made and hopefully this will give you some ideas so you can decrease on parts and size.
Nice work! Have you looked into twisted string actuators for your tendons? Regarding your joints, I printed a hand a few years ago where the joints were ninja flex filament. That worked fairly good.
Love the transitions between CAD and real photos!
very good iteration of your design, i can almost already see the next one with fewer parts and more optimized due to 3D printing. Keep up the amazing project!
(edit) your mechanical interpretation of the human had is fascinating
Amazing design I am now busy with assembly the inmoov hand but this project looks awsome
The best hand prosthetic I've seen.
Great work man!! This is the first time i see soo much work focused on thumb movement and the look is just amazing i will keep going with the movement of the knuckles, because for complex things like writing or grabing small things are really usefull
currently im trying to make a prosthetic finger for a project in school and this video really helped. thx! :)
I hope you've made progress, I'd love to see the final product in action. this could potentially replace other bionic limbs by being stronger with more mechanical parts rather than the electronic ones.
I’ll be back...
Yo... You back?
This makes me curious about how much more compact and quieter this would be with something like the new brushless motors and drive systems that have been introduced in the 3D printing world to replace stepper motors while still retaining positional accuracy.
great job. you should use a brushless outrunner motor for the wrist.
Pow! Mind blown!
I have been using Fusion 360 and have been 3d printing and designing for almost one month now with zero education in either and watching what your doing is very inspiring to me.
If I keep on the path I'm taking, seeing what can be accomplished I can only imagine what can be done with this technology.
I can't wait to see how your project evolves going forward, it's already so amazing!
Have you looked at tenticals and thought maybe a finger tentical might be something to look into for a simpler design?
Thanks for sharing and as always Keep Building👍
Thanks Todd, I'm really interested in some of those soft robotics designs like the tentacle you mentioned, I'm looking into artificial muscles too. Good luck with your projects :)
Thank you for sharing this project, I've got inspired for my own project.
best of luck
impressive and inspirational ,I ll try your finger joints in my project but still lacking of having ecentric control of flexion movement in a more efficent way ....
Love from india, your videos are very inspiring. Keep us filled with your style😄.
Extremely well done
There is a certain beautiful crazyness when you are able to witness the fluide movement all through ones hand ,joints, muscle, ligaments, tendons, carteledge, and so on but in the long and short of things you said that you wanted to make it as real as possible or to skaile then what you need to do is make all of the ligaments muscle,bone ,carteledge, vanes for formulation joints and so on and stop focusing all of your attention on each individual joint and take a nice long and in-depth look at the tendons and ligaments as your reference of motion and I bet that all the rest will just kinda fall in place that is what I did when I made my sparton glouve prosthesis
What would help with the flexibility is using less Stiff parts, especially for the exo-skeleton and more flexible parts, hands are dexterous, so they need dexterous materials and components, like rubber.
Instead of using springs I would suggest using both ends of the cable for the push-pull force. I think you should need two servos per finger, three for the thumb and two for the wrist. Also I would dump the idea of feedpack pots - just to complicated, also I would design the joints to not use bolts, just as in a real hand. If you need any design guides... just look down on your hands xD
Yes I agree, I have played with many robotic arm control configs, and the dual cable is at the top of my list atm
are you kidding me feedback pots are CRUCIAL if you want a functional hand that can perform tasks. The main brains of the machine need to know what its limbs are doing somehow!
@@artbyrobot1 not really, he's using servos (which have a built in encoder) so the feedback pots are a touch redundant, if you were using a complex kinematic model or a neural net you may make use of the pots, but his simple setup can't reasonably be expected to gain anything from having twice the number of encoders.
~Fyi I'm an ee
Well, this is really inspiring, i love the science, the electricity, robotics, all this kind of things, and besides, i just have 3 fingers on my left hand, i this really inspire me to learn first at all about the 3D printing, and this biometric science
My gosh! This is very very beautyfull and thought through design. Easily has place in top 5 prosthetic arm design ive seen! U designed it all by urself? Right? IF SO THEN I CANT BUT CONGRATULATE YOU. BRILIANT MIND. I LOVE YOUR SOLUTIONS. I can see few invonveniences, but all will be perfected in time. Great work. Keep it up! Prosthetic designs and 3d printing are very close to heart. Seeing someone accomplishing both this well is great! Good luck. I hope u will accomplish to refine this work to its peak! 👍👍
Thanks, I appreciate your support :)
Hey mate, found your channel today and you have totally blew me away looking forward to content in the future.
Hat off for the incredible amount of effort done.
It would be nice to have a small package 1-wire 2D magnetic sensor like the 14bit ams, to make it easier to built it in each joint and routing the wires.
When each joint can be measured, the servo can be replaced by smaller motors and the "twisting fishing line" force conversion + transmission, which reduce the size of the actuator and making the forearm smaller..
PLA is actually one of the strongest materials, although not tough, it has the highest strength of regular materials, unless you go up to some really expensive stuff like ultem or peek, those are insane but expensive af
I've learned to love pla, it can be so strong with the correct geometry! btw really liked the video, that hand looks amazing!
I am very impressed with your work! You inspire me to work at a much higher level.
I wish you had access to a metal printer and smaller stronger motors.
Well done!!!!
Wow this is great. The company Ottobock(prosthetics) has to give you a call! This is great!
Not to be a downer, but OttoBock already offers a hand very similar to this (but much more compact).
Great quality content with well research I ever found in Bionics.
Great work!
The hand is a wonderful work of art!
our skin has lot to due with gripping ability too, you need to put a soft rubber surface over the entire hand. Now try getting the same gripping characteristics when the hand is wet! You'll need to start creating features in the rubber at the micro/nano level (think lizard hands). Soft robotics is hard
We are so far from a decent mechanical hand, it's heartbreaking. Having lost a hand, I know what a real marvel it is. I think we need to go to a bone and tendon model and work on adapting it to a prosthetic.
This is some cool stuff..... N nice use of 3d printing.... To get more sensitive output from end effector... You can use graphene glass.. or sensor.... It will give us more reliable signal... With low power consumption
Saw one that a senior student made in my school and been amazed since
This is an excellent video! You've done some really high quality work here. I'm thoroughly impressed.
Nice model! The bike brake cables are also known as "Bowden Cables."
Have you thought of using spring steel shim stock? It's cheap and you can play around with the various thicknesses. Also, why not use something like a heavy test fishing line for the tendons? Something like 60 lb braided line, then you wouldn't need cable housing and the slop with that mechanism. But you'd have to hook up the flexor and extensor sides, which work if you are using servos. Try printing with more shells and top/bottom layers, it might hide those gaps that let you see into the parts. One last thing, why not print in PETG? It's only slightly harder to print than PLA, but I've always had the best experience using it for mechanical parts.
You are very good at explaining things !
Well done !
Good attempt 👌👍👏
I suspect that current nerves impulse sensing technology isn't as big a limitation as people think, if it's used right. I think that training an AI to take whatever nerve impulses that _can_ be read, as a cue for what type of motion the user is trying to perform, and matching it to the expected series of motor events needed to recreate that action with the biomimetic hand, would yield good results. Especially if the hand has sensors like multiple low resolution cameras/sonar/radar/lidar, and pressure and heat sensors-whatever is needed-to feed into the AI and give it more cues to better adapt any motion it's tasked to emulate, to the objects it supposed to manipulate.
It might even be useful to imbed cheap contact microphones into the tips of the fingers where the grips are, to get an reading of the surface texture of an object being handled, by using the sound of the finger as it brushes against it. It could be combined with a contact pressure switch on each fingertip, to signal more clearly to the AI if a certain pressure threshold has been exceeded. In fact… You could even stack two or three contact pressure switches on top of eachother, each with a higher, specifically calibrated activation pressure, to allow sensing of multiple pressure thresholds accurately. You'd want to choose the exact activation thresholds for each carefully, to make the most of them. And it might simply be better to use a single, more expensive pressure sensor that can read any pressure (within reason) with decent accuracy, depending on how well the stacked pressure switches work in an actual design instead of my imagination. Where was I?… Contact microphones! Have you ever worn a stethoscope, and brushed the sound sensitive membrane across something? Try it, if you can! It produces a sound that's a bit like what you hear if you accidentally brush your eardrum with the tip of a cotton swab. Different textures should produce slightly different sounds, and if the tip of the finger can also tell that pressure is being applied to it, then it can act as a two different clues to the AI, that the fingertip has encountered a certain type of texture signature, instead of, say, picking up vibrations that are traveling through the structure of the hand from somewhere else and giving it a false positive. The more clues an AI is given, the more it can deduce. Same with humans. And like humans, an AI also needs quality training to do quality work.
I suppose an alternative to using contact microphones to detect texture, might be to place a near sighted digital camera, with a preference for higher resolution instead of color or light value differentiation. Then two or three LEDs can be imbedded in the tip of the finger as well, surrounding the camera lens equidistantly, and programmed to blink in a pattern that syncs up with the frame rate of the camera, with only one LED on at a time, so that each LED is turned on in a cycle. This should illuminate the surface of an object that is extremely close, in such a way that the light will highlight the parts that are higher, more, from the direction of the LED light source. Then, when the next LED is turned on, and another frame of video is taken, the surface texture will be illuminated from a different direction. The deeper bits in the surface of the object, should be darker, and provide contrast. Using a group of frames taken from the camera, each with a differently angled light source, the AI should be able to puzzle out the surface texture. There are definitely a bunch of engineering problems that would need to be solved to make this work, and digital camera sensors aren't even close to as cheap as a contact microphone sensor and pressure sensor combined. But it's still an option, and you could probably still get five camera sensors of serviceable quality for under 300 USD, if the price of the Raspberry Pi camera modules are any indication. It'll be the scratch resistant lens cover that will set your budget back.
Hope you liked my spit-balling of ideas. Keep being awesome!
The pressed in brass threads I think would have been better if epoxy was used vs. Melting them inplace. I also think since joints are using torsion springs to return to neutral you might be able to use kevlar threads to replace the cables.
You need a pallet 2 or a pallet 2 pro a well tuned printer some flexible filament some conductive filament and couple different pla+ and you my friend would be well away looking forward to see the next iteration
The metacarpal bone movement in the little and ring finger seem to allow the tips of the fingers to line up since they are shorter. Without this movement things like playing the guitar would probably not work so well.
As an idea, use not steel cables, but plastic ones, for example from fishing materials. They are more plastic, and in terms of strength they are not inferior
When we want to learn about robotics, we take a good look at ourselves and our biology. Then we realize how nature solved a problem that took thousands and thousands of years in the making.
Then we realize how powerful hands are. Nerves, tendons, bones, chemicals that makes tensile structure stiff and so on. The lord settled us a challenge in nature. It's fascinating when we rediscover how complex the structure of the hand is.
Robotics is a careful observation of nature. Advanced robotics is deeply rooted in our comprehension of biology.
"I was looking at this fuckoff big stepper motor". I enjoy the language used in the video. Feels very real :)
Hey, I don't know if you know, but i like the idea of having string pulley's for the moving finger prototype. so you inspired me to make my own version of that :)
Cool project! BTW, I see you messing with those inserts in small parts... hold the iron upside down instead, and put the insert on its tip. Then press the part down on it. I use my old iron for this and chopped the tip shorter, so it won't stick out.
All my friends: I'm bored let's watch rap.
Me: glued to your every word and rewinding if I don't understand.
I was going to ask for STL files but about 4 minutes in I realized that even if I had every single piece in front of me, it would take me about a year to successfully complete it. Well done sir. You're a better robotics engineer than I.
the meta carple.. just ise a hinge between middle fnger and outer wrist. spring load. limited range.. thumb should be several angles..that work together.. find the radius of side motion..
wonderful work.