Did 1X Just Win the Robot Hand Race? [w/Scott Walter]

[00:00:00]It's rather remarkable how much information you can get from that, right?

[00:00:03]>> Uh you you you can do a lot. It's nice to have the additional tactile sensors and I think there's also kind of a debate there on how much resolution do you really need? Everyone feels like it's nice to have some sort of tactile, but do we need to have it at the human resolution or is low resolution just kind of enough to kind of guide us along?

[00:00:24]>> Hey y'all, it's Dr. Knowit All. I am once again with Dr. Scott Walter who is a hard man to catch these days. In fact, you have tried to escape from me. You've gone all the way over to Eastern Europe to Budapest. Correct.

[00:00:35]>> Correct.

[00:00:35]>> All right.

[00:00:36]>> Correct.

[00:00:36]>> So, uh I I managed to catch you and today we are going to talk about Neo's hands which have created quite the stir on ye oldie internet and on X. But we have to start at a very very specific spot and that is Barrett Bernish who is the CEO of 1X saying Scott is correct.

[00:00:53]That is where we are going to start the entire conversation. So, >> wow. Wow. Okay.

[00:00:58]>> So, congratulations on that one.

[00:01:01]>> Yeah. So, we ended up getting a teaser before the official release as usual with a teaser.

[00:01:08]We are trying to confer as much as possible with as little data as possible and I was making a couple of guesses >> and one of them I wasn't quite you know sure about and someone had given um the impression that I that information was correct and then burnt um recorrected.

[00:01:25]So, uh, yeah, and we'll talk a little bit about and I think is it might have been that my wording was a bit vague and exactly what I meant.

[00:01:32]>> Right. Right. So, anyway, so yes, this is the teaser and of course we have the full uh video and all that kind of stuff. And uh, let me stop this. There we go. I couldn't make it stop. Um, so, so let's just really really quickly back things up before we actually talk about Neo's hand because I think we could probably agree that uh there are like two basic ideas for how to do a humanoid robot hand at this point and there's a hybrid of course between them but one is to put the motors the actuators inside the hand itself and the other one is to do tendons. So obviously this is a tendon- driven solution. Yeah. Yeah.

[00:02:10]Very definite about that.

[00:02:12]>> The whole tree gets a bit more basically you can almost say tendon driven versus non-tendon driven.

[00:02:16]>> Yeah. Yeah. I guess that's true.

[00:02:17]>> When you go into tendon driven, you're getting into more direct drives. And the question is when you do direct drive, do you actually put them in the fingers? Do you do through a linkage mechanism in the palm of the hand? Again, do you do a hybrid where there's a little bit of tendons? There's a little bit of direct drive going on there. And then there's also hydraulics. So I mean and and artificial muscle and all these other things. But for the most part, it's like we are seeing a lot of people, you know, maybe about half the hand companies are going the tendon based route and the other half are kind of a direct either directly at the the joint itself or somehow in the finger or with the linkage mechanism.

[00:02:51]>> Yeah. And I I have to say I'm kind of cheating on on 1X here because I'm wearing a team Frank shirt today. That was unintentional. I put it on this morning and then I realized so they are the opposite of course. They are put the the motors the actuator >> they're using. Yeah. They're using linkage mechanisms which with the actuators actually in the in the palm of the hand not in the fingers.

[00:03:10]>> Yeah.

[00:03:10]>> Yeah. So, so um so 1X has famously been very very tendon focused and everything.

[00:03:17]Uh and obviously we're going to get into the details of all of this, but what um like what are the problems? Why would why is this not the obvious solution?

[00:03:26]You know, why wouldn't you just go with tender driven?

[00:03:30]>> Okay, tendons are just difficult to model uh for for the most part. Everyone wants to do SIM to real. That's why you're seeing a um a lot of companies like like Sharpa and um and Wuji going to direct drive because the main selling point of that is it is um seemed a real gap you know almost goes away. It's very easy to model. Tendon is notoriously difficult to model. They are pretty finicky. You know they can stretch. Uh you can end up having uh other issues of wear and tear and everything else.

[00:03:56]Maintenance everyone's like not sure about it. But they have one really huge advantage and that is you can do remote actuation with them. And you know the the problem is you can only make your motors so small >> to get the torque that you need to get there. But if you do something remotely either with like a linkage mechanism or with a tendon then you can start having much bigger kind of actuation that means you can get bigger grip strength. So if you move the motors to the palm you can only make them so big. So you can still use a tendon based strategy there but you know still it's going to be small.

[00:04:28]But if you move it all the way down here and like a lot of the reasons why tendon based hands tend to be in the forearm is we get this thing called the wrist.

[00:04:35]Right. Right.

[00:04:36]>> And it's like almost the only way you can get the actuation to kind of go through there is to use something like a cable or a tendon to be able to do it.

[00:04:44]Trying to come up with a mechanical mechanism is is very very difficult. So that's part of the reason and a lot of the reason everyone wants to do that is it all comes down to distal mass. So the lighter you make everything at your extremities the better it is overall for control for reflected inertia for like impact making things don't you know don't break or you don't break you know hitting something yourself.

[00:05:06]>> So if you start putting the actu actuators actually in your fingers you're going to already have a lot of inertia there that you're you're trying to stop.

[00:05:13]>> And this way by putting them down here you can make them a lot lighter. You get a lot of natural compliance. So there's many many reasons you want to have a tendon based hand and go with that approach and then I can come up with another list of reasons on why you don't exactly and this is and this is why it's like split down the middle that it's not like everyone is like going over one way and doing it. So 1x is not alone on the tenant based strategy where they are kind of alone is like the whole >> uh tenant based strategy for the entire humanoid itself. Everyone else is doing direct drive. Now the thing to look at is to keep this in mind is that um Barren has always been focusing on um getting rid as much distal mass as possible like always pulling everything in the center. And what happens is that when you look at their skeleton and and go out where you what you think is the actuator for that joint is not actually the actuator for the joint. It's the actuator for the next one.

[00:06:02]>> So >> it turns out when you look at the elbow and and again the strategy that's going to go here is that the elbow is actually going to have two degrees of freedom.

[00:06:10]It's going to have the normal degree of freedom we have here and it's going to have the wrist degree of freedom. So you can think of the elbow as having two dos there. And you will see something that looks like some actuators right there.

[00:06:19]Those actuators aren't actually for the elbow, but they then go through a cable that will then go down here and do that.

[00:06:25]Where the actuators for the elbow is is actually up here. And they're using this differential drum approach, which means you always have things paired together.

[00:06:33]So you'll have two actuators actually up in the shoulder that are doing two things down here. And again this is like a point of view is everyone thinks of the human arm as being three one three >> right >> and and you can think of that as with solving the math of kinematics but as as a mechanical engineer you should mix it up and say well wait a minute can I do like two two three right or maybe could I even do like two three two there's so many different ways of kind of looking at it and when you look at that way then suddenly you realize hey I can build my mechanics a little bit different so it's still like a human seven off arm we haven't seen enough except from some patents, we got some sort of ideas. And so when we go and examine what we're going to be seeing, let's kind of keep that in mind on where the actuators may be.

[00:07:14]>> Well, and speaking of this, I don't know if you want to start with the still frame image or if you want to start with the videos, but uh this one has a lot of >> Well, I guess go ahead and start with the reveal video because >> All right. So, there we go. So, this is the original original close-up um that we saw prior to the full reveal of the hand, >> the full release. And and from here, we do our normal slow thing. Everyone's like, "Well, how many degrees of freedom is it going to be?" Everything else. So, uh, >> I have to say before we continue, I just found this memerizing to watch just because of how fast the little guys are like all moving in there. It's very cool to watch. It's kind of organic.

[00:07:44]>> Yeah. Now, there's a few things. So, I mean, the first thing is like, well, obviously it's tendon driven >> and you see that they're using what are either called Bowden or Bowden tubes, which are typically the sheets that everyone sees on on bicycles. And we can talk about that a little bit.

[00:07:56]>> Well, and interestingly enough, 3D printers, too, which always care about the endeector mass as well. And so just like what you were talking about with Barrett, not wanting to have distal like you have masses out at the edges, >> uh 3D printers want to have the print head be very light so it can move quickly. So same same uh effect with that. So yeah.

[00:08:13]>> Yeah. Yeah. Yeah. Exactly. So you have some sort of tube that allows you to route um your your wiring or your cabling in this case. So um you know the first thing we know is like well how many degrees of freedom is it going to be? So we know it's going to be a five finger hand because they're not going to make like a three or four and probably not a six. It's going to be five. uh and and trying to count all of these tendons that are there in the actuators. And so I I went through it and tried counting and like I think the best I could come up with is I could see like 20. There might be 22 or a little bit more, but it seemed like right around 20 because it's pretty hard. And then you can start to count the number of actuators there. We don't see the final stack of actuators down at the bottom, but it appears that there's ranks. So we're seeing like at the top we got 1 2 3 4 5 6 7 8 nine looks like 10 that we can kind of guess on the front side. The question is it does it duplicate itself on the back side. So there's somewhere between 20 and maybe a double that number of 40 or something like that and maybe 20 actuators. That's what was not really clear on this image and then we had had to speculate. Now the the one thing was is that you can notice two things on here and that is um the tendons are paired with each actuator. Right?

[00:09:24]>> So the basically using the tendons more um or basically the actuation strategy is more like a pulley.

[00:09:30]>> Yes.

[00:09:30]>> So that means for each tendon you actually have one actuator. You don't have one actuator per tendon. So that means the number of actuators is going to be half whatever the total number of tendons are. And we can see the pairing coming in there. If you look right in the center you can see how the threads come out and you can see kind of how they're clamped down. And then you can also see >> and they are clipped, right? These are like two separate they're not it's not so it can't it can't slip there. There.

[00:09:52]Yeah, >> they can't slip. Exactly. And and that's a way that will allow you to do manual retension. So you would kind of loosen it, pull on the thread until you get the tension that you want. And then you would clamp it down again and do the same thing with the other side.

[00:10:03]>> And then you can see where the Bowden tubes come in. And when you do a close-up of the Bowden tubes, you can see like the spiral wire that goes around there. And then you can see the the plastic sheathing that's around there. So, um, what I had speculated on there was what, you know, the cross-section because if anyone's had like a bicycle cable and had it like explode and then look at it, it looks like they're made out of a spring.

[00:10:22]>> I think this is what you're talking about. Kind of more of a close-up.

[00:10:25]>> Yes. Yes. Yes. Yes. Yes. And so sometimes, you know, that that wire that they're using is usually a round profile. You know, there's some cases that maybe you can buy some that would be a square profile, right? And the square profile will will give you more stability, >> right? uh and be more solid. And it wouldn't be a surprise that 1X has done that because they've already done it in their wire windings, which is unusual for like everyone else when they they do the windings for the wires. The copper wire is always circular cross-section.

[00:10:52]>> Right. Right.

[00:10:52]>> And so they're doing that now. I think >> and the advantage of this is it won't it doesn't have a chance of bunching up in because a lot of this is pushing through. It's not just pulling, but it's pushing through the the bow tube.

[00:11:02]>> Yeah. Yeah. Yeah. Yeah. Yeah. And and of course the thing is you're always going to be getting a lot of compression. So, so when that tension's under when that when you have a lot of tension on the tendon, then that bowden tube is going to be going into kind of a compression.

[00:11:14]And if you got circles on circles, you know, there's like slippage and but if you have two of them coming down here now, that means the load is being carried a lot better, which is why you'd want to have that spiral made out of it.

[00:11:25]Now, I think the confusion is someone thought I might have meant that the actual cross-section of the Bowden tube was square. I mean, that would be wild.

[00:11:31]So, the Bowden tube is circular. It's just the spiral that's in there. And I think that was the record that that Brentton was correcting when he said, "I was correct that the the wire that they use is actually a square profile." What may or may not be clear is whether they're actually building these things themselves or whether is something they've been able to get a supplier for.

[00:11:48]Now, I would not be surprised if they're building themselves because >> um they probably want to have certain specs on exactly how this I mean they're they're probably optimizing it and and the only way you can kind of control it and would probably be able to do it yourself. And if they already have like winding machines, stuff like that, maybe it's not a big deal because you you probably want to get the diameter and everything right. What I don't know is where they also have an inner lining, which would be interesting because I think I've seen on some bicycle cables, don't they have like an inner lining inside of that as well to keep it from rubbing against the spiral that's on the out. So you have like an outer piece of plastic, then you've got a what looks like a spring, and then you have like an inner piece of plastic, and then your cables go through. I think uh for the high-end uh bicycles, they tend to be teflon and not just plastic on the inside to really reduce friction. So, >> yeah. Yeah. Exactly. Exactly. So, so you you're the bicycle expert. You you've had to deal with the maintenance on your bikes all the time. You know, kind of like the stability and the rigidity of them and what the high-end ones would be like versus the ones that I >> but also the really big thing is obviously they are not they're using some sort of daea or something like that, not steel cable. Uh because that would be very stretchy.

[00:12:55]>> Yeah, it definitely definitely is daea.

[00:12:57]>> Yeah. Uh in fact, actually this right hand image right here probably shows that right. This is uh this is the actual cable.

[00:13:04]>> Yeah. Yeah. This that's daea.

[00:13:06]>> Yeah.

[00:13:06]>> Yep.

[00:13:07]>> Yep. So you you're you're seeing uh the pairs of Yeah. the pairs of tendons that are going through there. It's some sort of dynamic blend and um as you can see they are not connected together but in a sense they are mechanically connected together just that you can separately uh tension them as you need it. So what that means you're going to have a pair so you're probably going to have like a flexure >> a flexion and extension pair and an ab and an ad pair and a bunch of others as we kind of go through this. So that's that's the first thing we kind of see.

[00:13:35]The real question is what was the total number and then what we found out on Friday when it was revealed um it's a 25 do hand using their terminology right >> and now I I think we have to qualify that because >> no one's quite figured out how they have got to the number 25 >> and they're using some interesting math here >> I think we can start with the wrist three makes sense for the wrist three dot for the wrist >> right right and now it's not normal normal to do that. Now, when usually when people say >> the hand has 27 degrees of freedom, again, I I say, "Oh, you can tell they're a gamer because they're looking at like the avatar hand that they have."

[00:14:14]And the first thing they have to do is they have to position the hand in space.

[00:14:17]>> Yes.

[00:14:18]>> And that's usually six degrees of freedom. And that's what we call the arm.

[00:14:21]>> Yeah.

[00:14:21]>> And then then the digits, that's where you have the degrees of freedom. And so if you take 27 minus 6, usually people say there's about 21 degrees of freedom in the human head. There's arguments over that, you know, depending upon, you know, the different metacarpals and how they move in your palm and everything else, but generally it's accepted. It's like a little bit north of of 20 uh degrees of freedom that you have in the hand. So, here we can see, and we've even seen in some of the posts, they talk about it being basically 22 plus three, right?

[00:14:45]>> Where three is the wrist, and that's usually not included in there. I can I can cut them a little bit of slack. Um, no, no intention of uh there. Yeah, because they are cables after all.

[00:14:56]>> Because they, you know, the the wrist is kind of integrated in here and part of the strategy and it's not like you have a wrist and the hand above it, right?

[00:15:05]It's like, well, a lot of the the stuff of it is down there. So, there is some kind of integration and the third doth is actually down here by the elbow. So the whole forearm rotates and and you have two that are up there using that same kind of cabling or rope strategy that when you look up around the wrist mechanism in there when you you kind of zoom in you you'll notice that there's like a thicker kind of cabling goes through there and it's differential and it's it's the same mechanism I think they've used in all the other versions of Neo uh with maybe some optimizations a little bit improvements basically the same idea that you get what it looks like they do the yaw first and the pitch second which is an interesting Yes. Well, well, I you don't need I I mean the yaw a limited amount of motion is probably fine because you got to remember you're cabling all of this stuff in through there. They don't want it to yaw too crazy and then the pitch can go.

[00:15:56]>> And so sometimes people do that the opposite order. They might do the pitch first and the yaw second and thinking that that and not realizing that wait a minute what happens if we turn it around and you might realize that oh guess what it makes everything easier. So again, when you're designing stuff, don't get locked in the mindset that it has to be pitch yaw, right? There's nothing wrong with going yaw pitch or anything like that. And and so I've seen some, okay, as we look at this, we have a much better chance of being able to count the cables. Yeah.

[00:16:21]>> And I've gone through there and tried to and I think depending on how many times I freeze it and how many times I look at it, I can always count at least 20.

[00:16:30]Sometimes I see 21, sometimes I see 22.

[00:16:32]So I think there really are 22 cables going through there. That would make sense, right? The thing is, I can only count 10 actuators. And if you look at the pattern, it's like a, you know, two, three, two, three. And there's like, well, wait a minute. That gets us to 10.

[00:16:46]Where's the other one? Well, there's still room for another one maybe around the back. And if if there is another one around the side, that might be why it's hiding two of the tendons that we can't see very easily in there. So, it's not like 40 tendons or anything like that.

[00:17:00]It really appears to be >> 22 tendons.

[00:17:03]>> Right. Now, let's talk about joints, right?

[00:17:07]So, so >> there's a lot >> because usually when people talk about degrees of freedom, that's what they're assuming is like the joints you have.

[00:17:14]>> Each finger has um an MCP down here, the knuckle, which has two degrees of freedom, >> the ab and adduction and the flexion, the extension, and then you have your pip and your dip. So, basically, you've got four joints per finger, >> right?

[00:17:28]>> Including the thumb.

[00:17:29]>> Uhhuh.

[00:17:30]>> So, looking down at the thumb, the thumb also seems to be a four off thumb. is not a five off thumb. It's it's it's a four off thumb. And what they have then is like what the fifth metacarpal is able to do position or or opposition and reposition on that which normally is what everyone calls the pinky. We've seen that in I think the Sharpa hand has that. Don't think the Wuji does, but uh the Tesla hand kind of showed that famously also. So pretty much that means that one is able to come around and you'd have to be able to control that in the hand. So, if you count all that up, you get 21, >> right?

[00:18:03]>> And and I've looked really closely at the thumb and there's like it does not appear that they have if if they had five, they would have three down here at the U the CMC joint.

[00:18:13]>> Yeah. And again, there's kind of a raging ar if you actually pin biomechanists down talking about that.

[00:18:20]They will say the CMC is really a twod dooff system where you have a coupled joint that cause your thumb to kind of rotate, but it's not really three do.

[00:18:30]So, if you're going to three do, you're giving yourself a little bit of overkill on there, right?

[00:18:34]>> Uh but there may be some way that you arrange it that you get a very similar kind of orbit of >> of the thumb. So there's no reason and and again when you add that extra degree of freedom your CMC joint becomes really sloppy. So you're almost like I would recommend don't do that just stick with a two and that sort of sees seems to be what what we're seeing on there. So it's like well wait a minute if that's 21 where do we get the 22 >> right? I think the 22 really comes down to that they're counting the tendons because if you think about it, the tendon is kind of a control, you know, it's a control lever. Whoop! I can pull on this one and I can pull on that one.

[00:19:10]>> But it becomes a bit disingenuous because they're tied together.

[00:19:16]>> Yeah.

[00:19:18]>> So, if you look at the hand, if you look at the human finger, you could say I have abduction and adduction.

[00:19:24]>> We would usually say that's the same degree of freedom. But you might say ab add. That's two degrees of freedom.

[00:19:30]>> Yeah.

[00:19:30]>> Flexion extension. That's another two degrees of freedom. It's like it doesn't really come. It doesn't really work out that way. But you know, and um and and again, if I had a linkage, a hard linkage in there for the abduction, there would be one actuator there and I would say that's one degree of freedom.

[00:19:47]I would not say that's two degrees of freedom, >> but because I have two tendons there.

[00:19:50]Yeah.

[00:19:51]>> Right. Right. Now, this is the interesting thing is you get the two tendons that come down that are not tied together.

[00:19:56]and and and you can see they're kind of snipped and you're able to tension them differently. But if you look at actually the actuation of the rest of of Neo, they have like these ropes that go around and you can see all these capsins and everything else they have.

[00:20:08]>> They're actually using a pulley mech system where you have them wrapped around. It's actually a single cable.

[00:20:14]>> They're like potted at different ends.

[00:20:16]They come down the actuator, wrap around and go off the other way. And of course, what it does, it pushes and pulls on it.

[00:20:21]But because it's not cut in the middle and then tied down, >> right, >> that's considered one degree of freedom.

[00:20:26]>> Yeah.

[00:20:27]>> But evidently you take it, you cut it and then put a couple of screws in there. Now it's considered two degrees of freedom, right?

[00:20:33]>> So this is I'm calling this out. It's like this is not a 25 degree freedom hand. Okay? Because they really only have 11 degrees of freedom of control of the digits, >> right?

[00:20:43]>> And then you want to throw in the others there, you can't. Now, um, they are doing some things to give them very close to full control of of what's going on and we can we'll be able to see a little bit what's going on. So, the first thing is let's look at the cable bundles.

[00:20:57]>> Yeah. Well, also I do notice that there is that orbiting that that very large >> um >> I don't know if it's a cap stand or if that's an actual uh >> that's a weird one. I'm not it's kind of it's gonna be interesting to see how they are moving that but I think in some of the videos we have seen it kind of rotate around there. But that creates an orbital motion rather than just a specific separate degree of Yes. Yes.

[00:21:19]Yes. Yes. Now, if you kind of look at the pinky or if if we're actually able to close up uh do a close-up of the palm anywhere.

[00:21:26]>> See if I can get >> We're going to start to see the bundles going through there. So, we can see there's a bundle of cables.

[00:21:33]>> That's as big as I can get it.

[00:21:34]>> That go all the way to the left side.

[00:21:36]>> Mhm.

[00:21:37]>> Um just to the pinky. And if you when you zoom in enough and you try counting like do this at home because you'll probably get better resolution, >> right?

[00:21:45]>> You can probably convince yourself there's about five cables going on up there, >> four of them are for the pinky itself and one of them is probably for the repositioning of the fifth metacarpal.

[00:21:56]>> And then if you go to the other what would be the ring finger, you'll see like another bundle that's kind of racing up in that area. can also count about five there, which means you've got the one that's doing the um the opposition of the fifth >> metacarpal.

[00:22:10]>> So that's why there's five is because each one of them steals one more to get that pinky finger, you know, to be able to swing over like that.

[00:22:17]>> But normally it's four going up to each of the digits and then you'll see another and you'll see all all those bundles. And eventually you get in there and you'll probably see that you can make out 22 distinct tendons, >> right?

[00:22:29]>> Uh it's it's a bit hard on that, but it's there. But when you really look close at the CMC joint or if you even if you look in the thumb, if you look at the back of the thumb, >> you'll notice that on what is and I had an interesting discussion last night is is I think the the middle joint, what we normally would call the pip joint on all the others is actually the MCP of the of the thumb. And then there's like an either an IP or PIP joint or dip joint.

[00:22:56]I think it's a dip at the end. But if you look at that, >> you will see that there's two tendons going through there. And then you go up to the next one, you'll see there's only one that pokes out.

[00:23:04]>> Right?

[00:23:04]>> So what I am seeing is that on the the distal failings, there's one tendon that's going out there, right?

[00:23:10]>> And then if you go down one, you see another one. And if you get down to what would be sort of the MCP, especially on the others, you can see three, right?

[00:23:16]>> So when you zoom in really close, you'll see three going up and then two and then one.

[00:23:20]>> Now with the Tesla hand, we were really able to see just two going up there.

[00:23:24]There was a flexure going up. the flexure was kind of hidden. It was buried inside.

[00:23:29]>> These seem to be on the outside. And because they're on the outside, that probably means all three of those, you have three tendons that are that can act as a flexure at that joint.

[00:23:38]>> So, when they're all rowing together, then you really are able to to close that and get a lot of flexion >> with, by the way, up to 45 newtons of force, which is pretty reasonable. I checked a human uh an adult male that's relatively strong can do about 58 to 100 under very intense uh focus, but 45 is on the order of the same amount of force. So, yeah.

[00:23:59]>> Yeah. Yeah. Now, now they're probably doing the same trick with like the differential on there that if you pull on just one or the other, you get the abduction out of it. So, um, so they're doing that and the one that's going to the middle because it's going, you know, basically straight through the center of that joint isn't really putting anything for abduction, right? So, you can kind of say there's probably like really weak coupling there, but for the most part, it's like it doesn't exist. It's the other two that are the outboard ones that are actually causing the abduction adduction. But when all three of them row in the same direction, boom, you're able to get that down. Uh, but what that means is when you get it is you're going to get that clawing motion that is going to close all the other IP joints, right?

[00:24:37]And you can see that u whatever that tendon is that is going up and it might even be a little bit different here because if I'm counting it right, they've got two that are kind of going up and then then then you've got one that's going up there. So, it's almost like they have one that makes it all the way out to the distal, >> right?

[00:24:52]>> And then they have like two that are kind of making it up there to get you a little bit of control and then the others that are down there. So that seems that boy the the way the routing is there gets rather interesting and that everything is going to close down but we're only seeing three there. We know there's a fourth one and that's going to be the extensor in the back.

[00:25:09]Yeah.

[00:25:10]>> So they they have active extension on there and that means the two of those it has to really kind of struggle and you probably will then be able to get control of that.

[00:25:18]>> Right.

[00:25:19]>> The question is how good of the control is it? Is is it really going to be the true precision control that you're thinking? um because there is some kind of coupling but with the antagonism they can take that out.

[00:25:30]>> The other thing we've seen there's a lot of hyperextension. So >> yeah, actually let's let me um because I wanted to show the zipper. I think let's go to this one.

[00:25:40]>> Uh I I really this intro video is really amazing and I I'll leave a link to this in the description so you can see it's just the intro of the hand. Um but when we get to notice that the hand kind of collapses entirely, right? It's not two fingers. It's like um but we'll see this with the zipper for example that there tends to be kind of like an overco compression u when it's doing that. I mean these are really amazing things by the way. I'm not I'm sitting here going like wow this is really impressive what they're doing. But if you do look at it you can see this right here. Watch the other fingers. Yeah. It's just kind of like there's like a compression, >> right? And and and the thumb does an interesting kind of hyperextension there as well that uh is going to be expected and that's going to be the the issues with using um these kind of strategies where you don't have the same number of tendons and the tendons are not being done exactly the same way they have in the humans. Um but you also notice that yeah there's weird hyperextension on a lot of those IP joints.

[00:26:37]>> Right. Right. And they do actually show some very distinct hyperextension. Uh the other thing that we are not seeing here that we see in all of Figur's videos is this does not say autonomous.

[00:26:46]So I'm guessing this is being done remotely. Um do you have any information on that or not?

[00:26:53]>> No, I I don't have I don't have any anything like that. So it it's it's possible. Um and that you know they they could be doing a tell operation here.

[00:27:01]But >> that's that's the hyperextension which I I can't do.

[00:27:04]>> Yeah. Yeah. And I think and I think you were able to see how the thumb kind of rolled there. So I think we do see that.

[00:27:09]And then and there was also another one with the the gloves off right there like that. Yeah.

[00:27:15]>> Right now you can see also the the the problem with you you have >> uh when you bring these gloves in here how the the material kind of uh puckers in a lot of places >> and and this is like the big challenge is like now that's on there to give them uh like uh IP68 kind of protection.

[00:27:32]>> Uh it does more than that though because they actually say well let's see hold on let me go back here. Uh this is actually a big deal. I read a read up on this because it was fascinating. There is the tactile sensing. So they have force sensing in the tendons themselves. So they can feel the back pressure in the compliance, but then the actual white whatever that is material has sensors built in that allow it to uh do uh shear and slip detection.

[00:27:59]>> Oh, okay. That that's what would be interesting is whether that's covering up the sensors or whether the sensors are in there because >> the other option is it could be in these but I believe it's actually in the material.

[00:28:09]>> Where's the wires? That's that's the thing is I I've been looking for it. Um I uh Roden Mirdan Gustav and I were also asking like where are the potential wires that are going through there uh that would be carrying it? I've had a lot of other people ask me do you see any wires? I'm like I I cannot see any.

[00:28:27]So that would make sense that perhaps it's it's actually built into the glove which I think is is the better way to do it because then you don't have to worry about internal routing or anything.

[00:28:34]>> I I have to go find the original but actually it was either Bar or Dar one of the Anyway, somebody said specifically that the material was not a passive material. It was it was a sensing material.

[00:28:46]>> That would make sense. So then you'd wire from the end of the gloves into some dock where it would be able to then transmit the sensing. And I think the other thing we want to look at is like you see the tendon drive ratio there.

[00:28:57]>> Uh yeah, this is very low for most motors.

[00:29:00]>> Yeah, but it's interesting. You see that there's two there's there's 5 to1 and 15 to1. So you want it to be low this this QDD so you can get the back driveability but more importantly the torque transparency and with the torque transparency that means you have basically torque sensing built into the tendons exactly like you said there.

[00:29:15]Right. But what's kind of curious is like the 5 to1 is something that's very easy to do with a single stage uh planetary reducer. um the 15 to1 you get into that territory that you almost have to go to a single stage you can do it depending upon the size what I'm speculating is since they're talking about like the 25 off there is the 5 to one are all the the digits and that the 15 to1 are probably what they have for the for the wrist >> and that they would probably be able to do that maybe as a single stage because those actuators are a little bit bigger so that that's speculation on my part I don't think they have um different gear ratios for different tendons in the fingers themselves. It's just that it might be distinction between wrist and and digit joints.

[00:29:59]>> Right. Right. Yeah. And they tend to all be the same size here. So, I think you're probably right.

[00:30:03]>> Oh, yeah. Yeah. They're all I mean you what they could be the same size actuator, but you could always put another stage in there, but that means it's a little bit longer. It's a little bit more massive. You know, all the stuff you want to avoid. You just, you know, more complexity. So more than likely, you know, they're trying to do everything single sh and say 15 to1 you can do, but it's just always going to depend upon your your geometries.

[00:30:24]>> But but that is a big because most most of these actuators that would be in the robots would be more in the order of 100 to one, right? So like approximately an order of magnit it's it's hard to get 100 to one with planetary without like a lot of different stages in there or a much bigger diameter. So when you get the 100 to one, that's a strain wave.

[00:30:43]>> Yeah. Yeah.

[00:30:43]>> Uh yeah. And so the strain wave will usually like be from 50 to 100 to one.

[00:30:48]But everyone agrees that when you get at 50, you don't really have QDD anymore.

[00:30:51]>> Um when you're down in the range of 10 to one, you're definitely QDD. You might still be QDD up to like 20 or 20 five.

[00:31:00]Some might say you could go as far as 40. You know, no one quite is quite sure when you start losing the transparency through these these higher gear ratios, but if you're keeping it under 20, then you've got torque transparency.

[00:31:11]>> Right. Right. Yeah. And and it's it's and also you've got built-in compliance because it's a relatively low uh drive ratio.

[00:31:18]>> Yes. Yes. And and and the other advantage is again all all your motor mass is back in your forearm. It's not there in the digits and so the digits can move fast. Watch why you can do stuff like that. Why? They can hit with a hammer and it's not going to break.

[00:31:29]>> Right. Right. Yeah. And that was I think that comes up really shortly here is the person's whacking him in the hand as he opens. By the way, opening a bag of chips is not uh trivial. Like you have to have >> that's that's Tell me about it.

[00:31:42]>> I've struggled I've struggled with it myself sometimes. Yeah.

[00:31:47]>> So, >> I don't know if the packaging is just getting tougher, if I'm just getting older.

[00:31:52]>> Don't quite have the forearm strength that I used to.

[00:31:54]>> Yeah. Yeah. Here it is where it's like, you know, trying to get the funion out of the out of the bag. So, I guess it's an advertisement for funions.

[00:32:02]>> And there you can definitely distinctly see the way the thumb kind of rotates as it as it moves backwards. So, >> and yeah, and you can see the hyperextension. Now, I think there is one other video that they showed with the gloves off. Yes. Where we get kind of a better idea of the movement.

[00:32:16]>> Yeah. Uh here. I've tried to I've tried to get them. Yeah. So, I I just find this fascinating to watch. I just love watching the little the little things go >> as they're There we go. So, um and and probably if we frame by frame this and break it down, you'll be able to see how the because you can see the tendons moving. That part's the most fascinating one. It's going like this.

[00:32:38]>> Right. Right.

[00:32:38]>> That's that's kind of when it's moving one finger at a time, then you you're able to kind of tell. I was able to see that there was like the abduction that was going on.

[00:32:46]>> So, at the very beginning, you're seeing the ring finger abduct, and you could see exactly which was the motor that was doing that, right?

[00:32:52]>> Uh, one of But one thing that was rather interesting, if you go right to the very beginning, you'll see the ring ring finger abducts >> and then it bumps into the pinky >> and it forces the pinky to kind of move with it.

[00:33:02]>> Yeah.

[00:33:03]>> So, the the the pinky is is like in a neutral position and suddenly it gets pushed away >> right there. Right before that. Yeah.

[00:33:08]and then it doesn't return. And and it's sort of interesting that it didn't resist that that you maybe that's showing a little bit of compliance. I couldn't quite see whether one of the motors moved as a result because you think it would. I mean, it's like you did that, you would see a slight bit of movement, >> but I thought it was interesting that the position of the pinky didn't go back to where it should have been, >> right, >> when when the two interfered with it.

[00:33:32]So, it um it makes me makes me wonder about the position control uh with that movement.

[00:33:39]>> Yeah, that's Wow. I hadn't really gone into that much detail, but that is interesting that it didn't just come back to where it was like originally.

[00:33:47]>> Yeah, it it should should kind of resist or it's like, oh, you're pushing me out of the way, but when it comes off >> right there was Yeah. It pushes it right out of the way and then it stays still when the other finger moves back.

[00:33:56]>> Yeah. Yeah. And that could that could just be a simple control strategy.

[00:33:59]>> Right. Right.

[00:34:01]Uh right there it does it again. Yeah.

[00:34:04]So it does it twice at the beginning right there. So it pushes it out of the way and it definitely does not move back until later, >> right?

[00:34:12]>> And then it moves. So it's also I mean this is this is clearly an automated sequence, right? This is just something that they built in for testing. So it's quite possible that it's just not built into the sequence for it to go back to where it was previously.

[00:34:26]Um but but yeah, I mean there's a lot of control. Clearly it works. Yes. And it it looks >> and when you see when you see the yawing going back and forth, you can see what I think is like one of the drums. I don't think that's actually the actuator there. Yeah.

[00:34:39]>> For the um for the yawing going back and forth. So like, you know, right dead center, we see something moving back and forth.

[00:34:44]>> Right.

[00:34:45]>> I think that is being actuated um by some tendons that are actually up by the elbow.

[00:34:51]>> Okay.

[00:34:52]>> And that that you know they're then routing it right down the center. I could be wrong about that. I don't think that's the actual actuator for it.

[00:34:58]Right.

[00:34:58]>> Um, >> well, it's certainly not Maybe it is.

[00:35:01]>> It's It looks like it looks like a pin around which everything else is moving.

[00:35:05]So, yeah.

[00:35:05]>> Yeah. Yeah. So, so something is going on there. So, it could be that >> if it is the actuator. I'm just thinking it's because we can see the cabling that goes around there, the rope for the wrist. And you can see how it kind of crosses itself there and would go in on there and and that might just be the drum that allows it to rotate. If it is an actual actuator um then they sort of changed their philosophy of trying to keep that that distal mass away but it may be they say hey you know the other complexity was there there was already additional mass there to begin with and it's just easier to go ahead and do it.

[00:35:36]So um it may be that they've changed their their actuation strategy but my suspicion is it's still happening back at the elbow.

[00:35:43]>> Yeah. Yeah. Uh and and again, it's just there's there's a certain beauty to watching this all happening. And it's also in terms of biomimicry, this clearly has an advantage because you don't have a giant thick like space glove hand that with all the actuators built in and everything. You've got a very thin hand. It's very lightweight.

[00:36:03]It it can you can see how quickly the fingers can move without any issues. So there's there's some significant advantages to this sort of design. Um so yeah. Uh let's see what else. Um we talked about I I don't know that it's autonomous. I don't think it is autonomous. The sensing stack is really interesting. Do you have you ever done uh I mean the obviously if we have like the the built-in in the material sensing for shear and slip and things like that, that's cool. But when you do a direct pressure sensing like this, you're pushing back on the tendon. So what is it sensing? It's sensing the the compliance motion of the of the actuator. Is that what it's is causing that?

[00:36:47]>> Yeah. Um >> Okay. So, the back driving >> usually is Yeah. There's something's happening to the current.

[00:36:52]>> So, there's like a little bit of the back driving. Um and that the the motor current gives you an idea of what the resistance torque is, >> right?

[00:37:00]>> And so there is kind of some signal that's coming through there to to let you know so you don't have to measure it directly. So that's that's a pretty good indirect measure, >> right?

[00:37:08]>> Yeah.

[00:37:08]>> Okay. and and you can it's it's rather remarkable how much information you can get from that, right?

[00:37:14]>> Uh you you you can do a lot. It's nice to have the additional tactile sensors.

[00:37:20]And I think there's also kind of a debate there on how much resolution do you really need? Everyone feels like it's nice to have some sort of tactile, but do we need to have it at the human resolution or is low resolution just kind of enough to kind of guide us along? uh you know it comes down to again you know James Dama always talks about how we can count the the quarters in our um in our pocket.

[00:37:42]>> Yeah.

[00:37:42]>> By going through you know and even tell which is which you know oh is that a quarter? Is that a dime? Is that a nickel? You know just just from uh you know the touch but um >> do we need our robot arms to have that level >> of sophistication and to be able to do a majority of the stuff we'd like it to do?

[00:38:00]>> Right.

[00:38:01]So yeah, I I I think there's a balance between trying to have human scale level of touch and what is like sufficient to have the robots be able to do like 95 98% of normal operations.

[00:38:14]>> Right. Right. Yeah. It's unlikely we're ever going to tell Neo to put his hand in his pocket and count the quarters in there. So probably an unnecessary skill.

[00:38:25]>> Yeah. Yeah. Yeah. Exactly. And and uh basically change is going away, right? I haven't carried change in I don't know how many years now.

[00:38:34]>> I know I was saying I wasn't joking with somebody else. I was saying like my kid wouldn't know how to do that because he hasn't carried change in his pocket ever. So >> there are so many places you go to they just they just say credit card only, no cash everywhere.

[00:38:46]>> That's crazy.

[00:38:47]>> And and so you Yeah, you you have to have some other means of paying that is is not cash based anymore. Um- which I I I find kind of interesting now that even when you go to like little outdoor markets and stuff like that, farmers markets and stuff like that, they don't want to take money anymore.

[00:39:00]>> No, no, it's a pain in the butt. Um, so so uh I I do want to get to the household safety elements of this. Is there anything else you want to cover before we get to that point?

[00:39:12]>> Uh, no. It's just that you know obviously the gloves is how you get to that level of IP68 which I think is like the highest um uh ingress protection which I believe is what the standard is called and that uh one of them is for like um >> objects and particulates and things like that and the other is water. So the first number six is like a scale. I think it goes probably from does it start at one? Does it start at zero? Uh from from everywhere it's like something is is so big you can get your hand stuck in there, right? And and that's like maybe one or zero or one and then the next level up is like you get your finger in there and then you keep on going up and then then it's like um other particles, hair or or dust and stuff like that. And so when you get to six it means like you don't have to worry about dust or anything getting in your mechanism. And then then the other are different levels of moisture protection of you know it can handle you know splashing water that's being kind of shot at you know dripping down vertically a jet of water >> versus completely submerged. Yeah.

[00:40:10]>> Versus completely submerged. So if you can get to 68 that that means you're protected against about everything. Uh the only danger of course is like the design of these mechanisms and we've seen it also in the Tesla hand is that when the IP joint kind of opens up there you have this very big pinch point that opens up. which is something that everyone wants to protect. The problem is that when that closes that that area of a pinch point, your outer covering usually gets caught in there. Yeah.

[00:40:35]>> And so we're seeing everyone's gloves just getting chewed up. And we can also see that happened in the Neo Glove that when it was doing the the Lego blocks that the pinky finger actually had one of the IP joints, you could see there was a little tear.

[00:40:48]>> Yeah.

[00:40:49]>> That that that that actually happened.

[00:40:51]And you can see that there's little um uh what would accordion, you know, uh things in there all >> you get the accordions there to to help it stretch. But even even like right there you could see it. Wait, let me back up just a little bit while he's picking the grapes. Um right there there's a there's a tear right in the glove right there at the at the left hand right at the edge.

[00:41:12]>> Okay. Yep. And I think you you'll also see it in the uh in the Lego block that there's also >> the other thing that's important to note is they said that this is food grade material. So, um I'm assuming that's what the So, in other words, it it's safe to be in the home, which is something that they really really >> Oh, yeah. Yeah. Yeah. Yeah.

[00:41:28]>> Which which means it would be able to operate in a hospitality setting as well.

[00:41:32]>> Yes. Exactly. Exactly.

[00:41:33]>> Now, um Okay. If you stop it like right there, >> you can see on the pinky.

[00:41:40]>> Yeah. Little little chunk right there.

[00:41:42]So, yeah.

[00:41:42]>> Yes. The pinky right over the white Lego um at the what would be the dip joint.

[00:41:47]Yeah. Yeah. And those gloves are probably expensive if they have sensors in them, right? So I >> I don't know if that's a consumable item or if it's something that's easy to clean. So if you're going in food service, either you have to be able to do a wash down, >> right, >> or they would just go ahead and and replace the glove, >> right? Yeah. And that is um h that's interesting because if it was if it is a if it was a passive glove which apparently it's not because I saw either anyway I think it was Barrett who said it but they are therefore not completely disposable right they're not like latex gloves where you can just throw them away and put another set on. They cost some money because they have sensors inside them. So u >> now they probably are not that expensive but >> hard to say. I mean, they're probably expensive now, but maybe at scale the price is able to get down. And if it gets down to a reasonable amount, it might be, oh, that's what we're we're used to using for consumables anyway, right?

[00:42:40]>> Uh on the shop floor, >> right? And and there there would come a point where uh like they're going to get better with the material, too, because obviously they're still studying what material works best and over a million compressions and expansions doesn't break. I I think the only other thing was um that we might want to just talk very briefly about the 2 million cycles.

[00:43:03]So I know you have a thought or two about that because they were like, "Hey, we've done two million cycles." I kind of said, "Okay, at two cycles per second, that's not it doesn't last that long." You know, that that's only going to give you somewhere about a year and a half to two years worth of actual >> Yeah. Yeah. at most. And and the other thing is the um >> it's a lot >> a lot of these tendons they they know you can get a lot of cycles out of if you stay >> um within the range.

[00:43:31]>> So if if you never go beyond like what's the the normal rating you're okay and then there's a maximum right and the thing is if you go up to the maximum then your cycles go down. So you quickly will go down to like 200,000 cycles >> if you are pushing it very close to to what would be the maximum. Uh, so there's two ways of doing it is like never go anywhere near that's like like that kettle bell, right? It's like it's not going to do that two million times.

[00:43:53]Yeah.

[00:43:53]>> You know the picking the grapes, it'll probably be able to do that two million times because you're not putting a big load on that, right?

[00:43:59]>> Um, >> and the other thing is like the the test that was on there, I you know, looks like they are using the Bowden tubes sliding through there because that that's going to be very important. But it was like a constant mass. I don't know, you know, the mass may have been one of those that kind of made sure it did not exceed kind of the normal limits, but still sometimes it' be nice to have the loads vary a little bit as you're going along there because that also can cause issu issues with creep and fatigue and everything else is going to go on there if if it's that's more of what >> Yeah.

[00:44:29]>> because sometimes if you always have a constant load on it, it's better than one that's like on and off on and off, right? And the it looks like the the Bowden tubes are just straight.

[00:44:37]>> Yes. Whereas if you actually look in the hand itself, you notice they're kind of coming around and they're doing these S curves all around there and and and you'll see they're also under a lot of attention because they're having to cross into everything else. And I was like one of the things we're wondering is that that bundle you're getting underneath some kind of sheathing or or protective material. So now that's getting kind of bunched in there. And you also know from like your own bicycles that >> that when you do something like that, it's just going to pop out somewhere else. I mean, you cannot, you know, you've got to give it the room to expand. And uh that's so getting the length of those tubes is also critical.

[00:45:13]I mean, you want them kind of to be as long as possible to get a nice arc in there, but at the same time, you probably have to figure out for all the different ranges of motion, what can I do to make those things as short as possible so they all kind of stay confined. And also uh we have to deal with the fact that when you have anything that's of that length which is going to involve probably metals in it in the outer tube at least that you have expansion and contraction. If it's negative - 20° centigrade versus you know 50 50° centigrade which it could be in a very very hot truck or something like that that it's inside working that it could be that's a huge temperature range and these tubes themselves are going to contract. Yeah, the tubes would probably Now, now you probably know Dynamo better because you're a climber.

[00:45:56]>> Um, I think that's pretty much temperature insensitive, isn't it?

[00:45:58]>> It's supposed to be. Yeah, unless you light it on fire.

[00:46:01]>> So, yeah.

[00:46:02]>> Yeah. Yeah. Yes. But, but I think it it doesn't really stretch. I mean, it's one of these it's almost like everything that you you want in a cable.

[00:46:09]>> Um, but you're right that, you know, now you've got this other material and that >> you may not be getting creep from that, but now your your Bowden tubes are going to be expanding or contracting. And there's also fatigue on them as well. So you are going through a lot of cycles and compression and hopefully they don't don't fracture anything else. So there there's yeah there's a lot of complexity there. And again there's probably some sort of interior sheathing in there. Uh it it looks to me that it's also designed to uh get like a manual tuneup every now and then.

[00:46:39]>> Yeah. Yeah.

[00:46:41]>> So that's it's it's probably that you're going to get creep, but it doesn't mean you have to throw the whole thing out or replace the tendons. just have to retension them.

[00:46:50]>> Yeah. And >> it looks like the retentioning might >> I think it's just a standard Allen Allen wrench, you know, interface. So, it's just like >> Yeah. Yeah. Yeah. Yeah. Yeah. So, so it would be I don't know, you know, maybe uh monthly, quarterly, something like that that you would go through and and then maybe after a year or two the you know, a full-blown replacement of >> of the tendons and maybe the tubes and everything else. And that might be something that's like, you know, a half a day thing. But this retentioning worst case like a minute per tendon. So maybe within 22 minutes you'll be able to do it. Probably faster than that if you know what's going on. And I imagine what they do is like you you get you put the actuators at sub neutral position and then you probably stick the hand in something to make sure they're not moving around. So it's like fixtured and then you just undo those things and you've got something that lets you know do you have the right amount of tension.

[00:47:37]>> Yep.

[00:47:37]>> And then tie it off.

[00:47:38]>> Yeah. And and and quite honestly, because they're standardized, they may eventually be able to have a robotic just like a thing that they put on and goes like just kind of does it all like super quickly. So >> yeah. Yeah, they would probably come up with Yeah. special tooling to be able to do it really quick.

[00:47:54]>> Yeah. So anyway, wow. So much to so much to talk about. This is really crazy. We haven't even talked about the rest of his body, which of course is also tendon based, which is really wild. Um but I am uh I'm fascinated to know if anybody from 1X wants to respond to the comments. we'd particularly like to know, but if anybody else wants to respond, that would be awesome, too. Any other information you have, any other questions you have, we'd love to know.

[00:48:16]And thank you, Scott, for being with us on this evening. I don't know, you're rooting for Norway or England, or you just going to remain neutral in the in the game.

[00:48:23]>> By the time this comes out, the game will be over. So, we'll know what the answer is.

[00:48:28]>> The answer is I'm rooting for beep.

[00:48:32]I I I just want it to be a good game with uh with with no one putting their fingers on the on the thumb on the scale.

[00:48:40]>> Yeah, exactly. Yeah, exactly.

[00:48:42]>> Hopefully no controversial calls. Just let him play.

[00:48:44]>> Exactly. All righty. This will be fun everybody. Thanks Scott so much for coming to us from Budapest in the middle of the night. I appreciate that. Make sure you like and subscribe. Follow Scott over on X and we'll see you in the next video. Bye-bye.