Designing a robotic hand

[Kaarel Lumi]

Hi

This is a bit related to InMoov but mostly general design / robotics topic.

I have learned a lot from printing and assembling InMoov and now feel that I'd like to design a robotic hand of my own.
There are things I like about the InMoov hand, ideas I'd like to borrow and things that I think could be improved.
Also I would like to design a hand with a more scifi / industrial feel to it and I don't necessarily need to constrain myself with thinking about how it would be easy for others to reproduce.

Also I'd like to keep the overall basic dimensions (and number of digits etc) of a human hand but maybe make it thicker and stronger and more angular, both visually and functionally.

I'd like to incorporate more bolts and bearings into the design. Maybe metal springs.

I'd like to make it serviceable and modular. For example it should be possible to take apart the elbow, the shoulder and there should be hatches for maintenance. Maybe expose more of the mechanics.
I will probably stay modest with the degrees of movement and won't try to emulate everything the human hand can do.

I can draw decently, however I have only just started learning CAD with Fusion 360. It's fun though, I have already designed and printed small widgets that I couldn't have manufactured otherwise or at least not as quickly. 
Also I have been looking at a lot of pictures of robotic hands and thinking about the anatomy of a hand.


If you would design a robotic hand what would you do differently?

P.S. With the spare time I have, don't hold your breath. I will probably be done in 3+ years.

[Kaarel Lumi]

Just a small update. More pictures: https://www.instagram.com/lumikaarel/ 

[gael langevin]

Thanks for the update!

Very robotic hand indeed, I like the idea of using printer belt for the fingers. It's strong.

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

2018-08-08 23:17 GMT+02:00 Kaarel Lumi <terr...@gmail.com>:

Just a small update. More pictures: https://www.instagram.com/lumikaarel/ 

--
You received this message because you are subscribed to the Google Groups "InMoov" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+unsubscribe@googlegroups.com.
To post to this group, send email to inm...@googlegroups.com.
Visit this group at https://groups.google.com/group/inmoov.
For more options, visit https://groups.google.com/d/optout.

[Kaarel Lumi]

[Kaarel Lumi]

So yeah ... it's weird, big, heavy, bulky and completely impractical but I'm kind of starting to get a good feeling about it again. I'm printing the big palm plate now and then I'm planning to assemble all the fingers and the palm. After that probably on to designing the wrist and forearm and a decorative cover for the palm.

[wilco van toorn]

Nice design Kaarel, really like it, its nice strong and industrial, just the way i like it!

Op maandag 27 augustus 2018 20:25:59 UTC+2 schreef Kaarel Lumi:

[Kaarel Lumi]

Thank you!

Learning 3D modelling in CAD and working on this project has been a lot of fun so far. 

[Kaarel Lumi]

Assembly in progress. A few more small bits to print.

[wilco van toorn]

Like the photo style, but could you do some regular color? like to see how the really look, i do really like them.

Keep it up! And maybe someday you thinking of releasing the files? i want to build 1 as a show model (with your name as designer offcourse!!)

Op dinsdag 28 augustus 2018 21:59:28 UTC+2 schreef Kaarel Lumi:

[Kaarel Lumi]

Thanks! 

I've printed it in silver / grey filament so there's not a whole lot of difference compared to the pictures but sure I'll take a few color shots once it's assembled.

About files. Maybe once I have the wrist and cover plate done. But I haven't really designed this with other people in mind.

For example 4mm vertical holes are designed as 5mm (for the bolts to go trough easily) and the 7mm holes for bolt heads and nuts are designed as 8mm to account for PLA shrinking but horizontal holes are designed smaller. So a few test prints would need to be done to see if these designs will work on other people's printers.

[Kaarel Lumi]

assembly in progress

[Morten Enholm]

"I will probably be done in 3+ years."

You seem to be making a lot more speed than that Kaarel :)

Really like the design, it's going to be strong, do you have a Inmoov hand you could place next to it, for size comparison ?

really Steampunk style :D

Keep up the good work..

[Kaarel Lumi]

Thanks!

Just for you guys:

On Tuesday, August 28, 2018 at 11:17:28 PM UTC+3, wilco van toorn wrote:

Like the photo style, but could you do some regular color? like to see how the really look, i do really like them.

On Friday, August 31, 2018 at 3:21:41 AM UTC+3, Morten Enholm wrote:

"I will probably be done in 3+ years."

[Kaarel Lumi]

[Kaarel Lumi]

2304g so far ... and it's not even it's final form :) ... feels really heavy
Also I'm going to trim the belts at some point but I just didn't know until now how long I was going to need them exactly.

Now that I'm this far I'll have to make a video of it crashing or trying to crash a beer can or something I guess.

It would be cool to make a control board for it with knobs for each servo. I don't think it would make sense to try to go the whole gesture and MRL route.

Those extra holes in the hand ... this is where I plan to attach stuff that I haven't designed yet. The two rows of 5 holes are for wrist and the rest are for a decorative cover plate including the ones on the knuckles.
I have a very special design in mind for the wrist. I have no idea whether it's going to work so I'll have to experiment in miniature first.

[Kaarel Lumi]

Did a quick count for future reference:

Parts list for 1 finger:

18 ball bearings 

28 bolts, 36 for thumb

26 nuts, 34 for thumb

5 pull springs

1 10-turn pot

1 servo, 2 for thumb

1 M3x10 bolt

4 M2x8 bolt

4 M2x12 bolt for thumb

15 printed plastic parts, 16 for thumb

+ 1 big printed plastic palm plate

40cm of 9mm HTD3M belt 

TOTAL:

90 ball bearings (4x11x4mm)

148 M4 bolts in various lengths (20, 30, 40, 50)

138 M4 nuts

5 M3x10 bolts

20 M2x8 bolts

4 M2x12 bolts

25 pull springs in various sizes 

5 mexican 10-turn pots

6 servo motors (PDI-6221MG-360 continuous rotation servos + the disk shaped attachments)

77 printed plastic parts
200cm of 9mm HTD3M belt (NB! not regular printer belt. This one has bigger teeth.)

[Kaarel Lumi]

Ugh correction ... one of the servos is PDI-6221MG-180 :)

[Kaarel Lumi]

Playing around with the idea:

[Kaarel Lumi]

My arch nemesis. Soldering. Someone would probably say that this is not how you do it but I'm just going to hide all of this really carefully and try to never think of it again.

[Morten Enholm]

Very nice, going to be fun seeing this in action Kaarel :)

[Kaarel Lumi]

On Tuesday, September 4, 2018 at 10:03:12 PM UTC+3, Morten Enholm wrote:

Very nice, going to be fun seeing this in action Kaarel :)

Thanks! I'll make some vids once I get this thing moving properly.

 

[Kaarel Lumi]

Starting to come together.

There are 3 main components:

1. Hand
2. Hand controller
3. Remote (on a long cable)
   

There are 2 arduinos: 1 in the hand controller and 1 inside the remote. The one inside the remote will send digital data to the hand controller which will use this + finger position from finger pot to move the servos in the hand.

[Kaarel Lumi]

Just finished programming the arduinos and testing for the first time.

Everything works fine except for one pot that seems to rotate freely and get out of alignment almost immediatelly. So ring finger doesn't work at the moment.
First video on my instagram: https://www.instagram.com/lumikaarel/

[Morten Enholm]

Incredible work Kaarel :) no go and crush a beer / soda can for us :D

[Kaarel Lumi]

Thanks! I'll do that but first I'll fix the ring finger and tune the algorithm a bit to get maximum strength out of it.
Currently it does a very simple subtraction between the value that it reads from the finger and the value it gets from the remote. But this means that when it's almost completely clenched like when crushing a can it doesn't send the maximum 180 signal to the continuous rotation servo and sends like 110 or 120 which is barely more than the neutral 90. This means that it uses only a fraction of it's full strength right now. That's my theory. I tried a can already but it got crushed only half way maybe. Actually I have a few options how to do it. One way is to simply tune the algorithm but another way would be to use some of the extra knobs I have to "overdrive" the servos. I prefer the first option because it's more of a long term solution because I plan to need the other knobs for other things eventually. 

[gael langevin]

Very good progress on your hand!

Nice video!

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--
You received this message because you are subscribed to the Google Groups "InMoov" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inmoov+un...@googlegroups.com.

[Kaarel Lumi]

Thanks! I really appreciate all the support.

Also I have a instructional video up on instagram now on how to gently massage a beer can into submission. I've decided to refrain from such strong words as crush from now on.

Actually I think I have an amps problem. Strange stuff starts to happen when I try to crank all the servos at once. Strange but somewhat entertaining. Have a look. https://www.instagram.com/lumikaarel/

I did fix the broken finger though and also from software side it's trying it's best now.

[Kaarel Lumi]

I did manage to pump the amps up a bit. So the hand now consumes up to ~150w (I measured). Also the anomalous movement seems to have stopped now.

Cans still prove to be worthy adversaries though. In a struggle with a metal beverage container one of the brave servos let its smoke out for the final time.

I still have a few so just going to replace it but I'm unsure how to safeguard against it in the future. The servos get pretty hot very fast. I turned voltage down a bit but I guess temperature sensors would be the real solution. 

[Kaarel Lumi]

So I'm trying to come up with a design for the wrist and I found a cool design some time ago that I wanted to try.

So far haven't found a person online who could tell me anything about it. The author doesn't reply etc...

Decided to design a quick mock in fusion 360 to study how this works.

As soon as I defined all the joints everything went red and all the joins hate each other.

I'm starting to think that it may be impossible with regular bearings and may require ball joins everywhere.

[gael langevin]

Hello Kaarel,

What kind of servo do you use? If a servo cannot reach it's desired position, and keeps being powered, it will burn.

You might also have a Amp issue, when all servo run at once and force on something, they will draw a lot of current, if your power supply is not up to the demand, servos will start to jitter.

Regarding the design for the wrist, you might want to look for the Drupp Neck for InMoov, it uses the same principle.

https://www.thingiverse.com/thing:2059967

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--

[Kaarel Lumi]

Hi

I'm using these JX PDI-6221MG 360 degree continuous rotation servos that I couldn't use for InMoov :) I measure their actual position with a potentiometer that is connected to the arduino. The thing is that they take values from 0 to 180 and the difference between this value and 90 determines how much force the servo applies in that direction. So something like 0-90=-90 CW and 180-90=90 CCW. So when trying to squeeze something I may want it to apply the maximum possible force. Which means it will start to overheat. What I want to do is build some kind of safeguard mechanism that will let me apply maximum force for a while but stop before it gets dangerous for the servo and for this the only foolproof way probably is to use temperature sensors I think. I was able to solve the amps issue by simply turning the pot on my power converter that limits amps so I have enough current now.

As for the mechanism thank you so much! This is very encouraging. So it should work but maybe not in Fusion 360.  

[Kaarel Lumi]

Printed it out so I could play with it a bit. It works! 

[Kaarel Lumi]

First part of the full scale mechanism.

https://www.instagram.com/lumikaarel/

[Perry S]

Do you have any pictures in color? 

[Kaarel Lumi]

Sorry no but also I print in grey filament so there wouldn't be much difference.

[gael langevin]

Very nice progress!

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--

[Kaarel Lumi]

Thanks! Here's another step along the way:

[Kaarel Lumi]

Feels very solid and tight. Practically doesn't bend, twist or rattle in any way. Probably <1mm play if I really try. So far so good. I hope I can build the rest of it to the same tolerances and make it work as well.   

[Kaarel Lumi]

[wilco van toorn]

Could be a nice option for the neck. I know there is the drippy neck which is similar but I love your style.

Where did you buy the pots? Those are the same as in the servos and I would like to order some extra.

But keep the updates coming, really enjoy watching your progress!

[Kaarel Lumi]

Thanks!

The pots I use in the remote:

https://www.aliexpress.com/item/4pcs-Potentiometer-switch-5K-aircraft-model-potentiometer-switch-shaft-length-10MM-rotation-angle-of-200-degrees/32805956109.html

And here's a link Gael found a while back:

https://fr.aliexpress.com/item/Import-diameter-13-type-toy-plane-180-degrees-potentiometer-B5K-long-legged-new-special

About the neck ... yeah maybe but as I've said before I'm really not designing these things with recreatability in mind. For example I have no idea where to get another 30mm ball bearing with a 5M threaded hole. I got this one from a guy who makes doorknobs for boats. :D But lets see in the future. Maybe I'll get to publishing the files eventually.

[Kaarel Lumi]

[Morten Enholm]

Hi,

Your design is getting better and better ;)

Since this seems to be based on "Parloma" same as the Drupp neck for inmoov,

have you looked into the kinetics and how to control it in a simple way with myrobotlab ?

Trying to find some code on the net to see if we can get Drupp / Parloma implemented in myrobotlab, but can't find any :/ 

Except for this old code:  https://github.com/parloma/penguin_wrist

[Kaarel Lumi]

Hi! Thanks!

Actually the idea is from one of Vitaly Bulgarov's designs https://www.instagram.com/p/Bm_1i61Awlr (look at the hip and shoulder joints) which seems to be utilizing a central ball joint which the Parloma or Drupp don't.

@yangzinho actually has one closeup render of the joint: https://www.instagram.com/p/Bm5_SCQggqQ

I haven't yet made the parts that would allow connecting motors to it. I figure things out as I go. So I'm going to use my 12 knob "remote" first and then see later whether I can find a better way.
The kinetics are interesting. For me at least it's difficult to fully understand how it works without holding it in my hand and playing with it.
Also it's highly versatile. It can basically rotate + tilt in 3 directions. Or by fixing the outer ring it could be set up to only tilt in 3 directions.

I can probably come with a model for my joint at some point but it's very complicated I don't know whether it will apply to other variations of this mechanism.
By moving one arm all the joints rotate by varying amounts and nonlinearly. The head itself rotates mostly around the axis of the arm that is to the right of the arm that is being manipulated.

[Kaarel Lumi]

[Kaarel Lumi]

Progress has been kind of slow ... I'm struggling again a bit with the design.

But I do have some pictures to share. 

[Morten Enholm]

Just a heads up for you Kaarel, kinematics and service is now in place for this kind of design in a test version of Myrobotlab Nixi, i am currently doing test on it for Kwatters, so you will be able to use logical controls over you servos :)

btw, you design are just breathtaking :)

[Kaarel Lumi]

That's fantastic! Don't hold your breath for me though ;) I will be modifying my servos for 360 degree rotation and adding external 10 turn potentiometers that will connect directly to arduino like I did with the fingers. I have this ~3:1 belt drive that I need to consider. I will then do the initial testing with my "remote". After that I can start playing with MRL. :)
Also thank you very much for your support! I really appreciate it.

[gael langevin]

Impressive!!

I can't wait to see a video in action!

The weight is conderable, you will have to redesign the complete robot to your needs to hold those hands, but keep it up you are doing very well.

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--

[Kaarel Lumi]

Thanks! Yea ... it's super heavy already. I hope I can get the elbow and shoulder done somehow but I don't really have a strategy yet. I have a few ideas but it's getting quite challenging.

No plans for an entire robot at this point. This one is going to have to end at the shoulder if not at elbow and get bolted to a concrete wall or something. And then it's time for the next project.

I have some 8mm threaded rods I was planning to use for the elbow similarly to your "pistons" but maybe I should stick to the theme of using belts somehow. Maybe I can find a few more interesting mechanisms to try.

[Kaarel Lumi]

Just an update with pictures. Finalized the wrist design and I'm just printing parts now. Broke my printer a few days ago but got it running again.

More stuff on my instagram: https://www.instagram.com/lumikaarel/

[gael langevin]

Very nice engineering!

Thanks for posting.

Gael Langevin

Creator of InMoov

InMoov Robot

@inmoov

--

[Amin Amore]

Hi,

I need your input regarding the following, I'm looking for remote hand in order to give access to remote user to do some dental training:

• Users should be able to control the hand remotely and do work on dental manikin
• Control via gloves is preferred option
• Any other option?

Do you have any idea how to achieve that?

Looking forward for your input

 

On Wednesday, May 23, 2018 at 1:43:16 PM UTC+2, Kaarel Lumi wrote:

Hi

This is a bit related to InMoov but mostly general design / robotics topic.

I have learned a lot from printing and assembling InMoov and now feel that I'd like to design a robotic hand of my own.
There are things I like about the InMoov hand, ideas I'd like to borrow and things that I think could be improved.
Also I would like to design a hand with a more scifi / industrial feel to it and I don't necessarily need to constrain myself with thinking about how it would be easy for others to reproduce.

Also I'd like to keep the overall basic dimensions (and number of digits etc) of a human hand but maybe make it thicker and stronger and more angular, both visually and functionally.

I'd like to incorporate more bolts and bearings into the design. Maybe metal springs.

I'd like to make it serviceable and modular. For example it should be possible to take apart the elbow, the shoulder and there should be hatches for maintenance. Maybe expose more of the mechanics.
I will probably stay modest with the degrees of movement and won't try to emulate everything the human hand can do.

I can draw decently, however I have only just started learning CAD with Fusion 360. It's fun though, I have already designed and printed small widgets that I couldn't have manufactured otherwise or at least not as quickly. 
Also I have been looking at a lot of pictures of robotic hands and thinking about the anatomy of a hand.


If you would design a robotic hand what would you do differently?

P.S. With the spare time I have, don't hold your breath. I will probably be done in 3+ years.

[Kaarel Lumi]

Hi

First of all I will not pretend to be an expert in this but I'm happy to share what I know and discuss the topic anyway :)

My first recommendation would be to do some thorough research and try to figure out whether what you're trying to do has already been done or exists already. Motion capture gloves and robotic hands have been combined before it's the most natural thing to try with a robotic hand so it's not a particularly new concept. But what you're looking for is probably quite extreme accuracy in both movement and feedback.

In general creating an accurate robotic hand that replicates the human hand in every way is extremely difficult and has not been achieved so far. It's probably easy to sink tens of millions of dollars or more into the development.

The closest are probably some advanced prosthetics and some other prototypes. Which can cost alot. It'd expect $100k - $1m for a good prostethic hand. 

Typically the ones that try to replicate every possible range of motion and also the strength do not fit in the same form factor as the real thing.

In terms of feedback we have a very long way to go still.

So what is currently done in medical fields? Instead of trying to replicate a human hand usually they start with a few industrial style robotic arms for the overall accuracy, replace the gripper at the business end with a new custom made single purpose precision instrument and then more than one such robotic arm is used for the operation. Perhaps even 4 or more. This is a good accurate and practical approach that gets the actual job done in most cases. I can't imagine letting a robot reach something into my mouth to pull my cheek asite to get a better angle at that molar though.

The next major component is the glove or some contraption that again needs to be extremely accurate and needs to provide excellent feedback to the user.

It's important to think about whether the glove should be physically rigid and actually stop the user or whether gentle pressure is enough feedback.

These things defintely already exist but I have no real knowledge to share here.

The third component is actually the connection between the first two components.

You're looking for very low latency probably less than 10ms. Or the feedback will arrive too late and the mannequin will have a bad time :)

Please describe what you're trying to achieve a bit more if you can.

Regards
Kaarel

[Amin Amore]

Hi,

Thanks for your reply. The objective is to provide dental training remotely to dentists on specific procedure. Now the dentist are travelling 1000s km just to get access to the mannequin for few hours.

In your reply you mentioned the 10 ms delay in order to get good feedback that is about 600 km fiber optic delay, plus the addition of computing time. 

I was unable to spot any company or organisation with prototype or working solution.

The other possible solution is the VR usage to provide the training, but I think the cost to develop customized solution is too high.

Any other ideas or suggestions are welcome.

Thanks again for your input

Regards 

[Kaarel Lumi]

Hi

I did a quick search online and the dental (manikins?) phantom heads or simulation units are super cheap. A few hundred dollars each. That's less than 1 implant tooth. Why are the dentist travelling to them?

Or are you using some much more expensive ones? The rest of the equipment is expensive sure but If you already have a working dentist's office then attaching a cheap plastic head to the chair seems like a trivial thing.

I mean just the glove would be at least 10 times as expensive or perhaps much more.

Could you turn it around and telepresence the instructor to the students site?

Send the student a package:
1) cheap dental phantom head. (perhaps modified for the specific purpose)
2) cheap webcam. (optional)

3) some specific material or tool depening on the procedure that's being taught (optional)

The student will:
1) attach the phantom head to a chair in an existing dentist's office.

2) connect the webcam to a laptop and attach the webcam to the dentist chair or lamp in a way that gives the instructor a good overview of what's happening.

3) see the instructor over skype on the laptop screen.

Everyone would save time and money. Realistic experience, not overcomplicated and seriously robots or computer games are really expensive to develop.

[Amin Amore]

Hi,

Yes for sure. But its not about basic training with CAM, that will make life so easy. This is really complex procedure that need be checked by the head doctor. The cam experience is not good enough :)

I agree that the cost is so prohibitive but I'm exploring the market.

Thanks for your feedback 

[Kaarel Lumi]

Don't dismiss the idea too easily. Could something be done to allow checking the work? Could the student take photos of the intermediate and final results or maybe even move the webcam closer for inspection?

The head doctor can't see into the patient's mouth anyway while the student is working right? So student needs to do something then the doctor can check then the student can do something else etc probably.

Think about what else could be done to improve the experience.

I totally understand being biased towards your own idea. I do that a alot and I may be doing it right now. :)