WEBVTT 00:00.000 --> 00:01.787 go for it hello everyone can everyone hear me okay 00:01.787 --> 00:04.827 Okay. Yeah. Okay. It's my honor today to introduce 00:04.827 --> 00:07.427 Professor Elliot Hawks from UC Santa Barbara. 00:07.867 --> 00:11.787 And Elliot leads the Hawks Lab, and his research 00:11.787 --> 00:15.167 focuses on the intersection of design, mechanics, and 00:15.167 --> 00:18.807 materials. He designs novel mechanisms and applies 00:18.807 --> 00:22.847 non-traditional materials to solve really challenging 00:22.847 --> 00:25.227 problems in robotics, medicine, biomechanics. 00:25.367 --> 00:28.627 And he's really prolific. His research has been 00:28.627 --> 00:31.547 published in many, many high-impact journals, such 00:31.547 --> 00:33.767 as Science, Robotics, and Nature, and won 00:33.767 --> 00:37.367 numerous awards, such as the NSF Early Career Awards, 00:37.367 --> 00:39.147 NASA Early Career Awards, and the Presidential 00:39.147 --> 00:42.867 Early Career Award for scientists and engineers. 00:42.867 --> 00:45.387 But more on a more personal note, 00:45.527 --> 00:47.847 Ali and I share the same PhD advisor. 00:47.847 --> 00:49.487 He's my academic older brother. 00:49.527 --> 00:52.427 Even very early on in my PhD, when I was still 00:52.427 --> 00:55.087 learning everything, His name keeps getting brought up 00:55.087 --> 01:01.007 in research meetings and brainstorm discussions. 01:01.007 --> 01:04.147 So Elliot quickly became someone I really look up 01:04.147 --> 01:06.847 to and admire. So I'm really happy to have him here 01:06.847 --> 01:09.227 today so we can all learn together what awesome 01:09.227 --> 01:11.527 research he's working on these days. So without 01:11.527 --> 01:14.087 further ado, let's welcome Professor Elliot Hawks. 01:14.467 --> 01:15.367 Thank you, Tony. 01:18.007 --> 01:20.267 Awesome. Thank you for the introduction, Tony. I'm 01:20.267 --> 01:23.827 very glad to be here. And today I'm going to be 01:23.827 --> 01:28.547 talking about this project on a robotic collective. 01:28.547 --> 01:31.917 And we have a lot of different projects in my lab. 01:32.387 --> 01:35.907 I just decided I'd list a few in case some of these 01:35.907 --> 01:38.267 are more up your alley. You can talk to me later. 01:38.267 --> 01:41.567 But a bunch of different robots, climb, jump, 01:41.727 --> 01:43.127 burrow. This one we do with 01:43.127 --> 01:45.227 Professor Goldman over here. 01:45.987 --> 01:49.407 Light robots. We have an intubation robot that we're 01:49.407 --> 01:53.637 doing a startup with to make a device that can autonomously 01:54.027 --> 01:57.247 go into your trachea for emergency intubations. 01:57.247 --> 02:01.387 Working on a brain surgery robot, which is a challenge. 02:02.037 --> 02:04.347 You have to go in through the leg and then go 02:04.347 --> 02:06.427 through the blood vessels and get up to the brain. 02:07.167 --> 02:09.627 A new project on space junk capture. And 02:09.627 --> 02:11.867 then we also have a project on running 02:11.867 --> 02:14.247 shoes. I was chatting with Greg about that. 02:14.567 --> 02:16.807 So anyway, a bunch of different things. But today 02:16.807 --> 02:19.327 I wanted to talk about these collective robots, 02:19.347 --> 02:22.607 partly also because Professor Goldman is a big 02:22.607 --> 02:24.867 collective robot guy. So I figured I would focus 02:25.007 --> 02:28.827 on this one. So these are the little robots. This 02:28.827 --> 02:31.937 is a couple of them. We made about 20 and outline 02:32.847 --> 02:34.397 of what we're going to be talking about today. 02:36.887 --> 02:41.767 The kind of key thing that I'll mention is this T1 02:41.767 --> 02:44.827 transition, which is how these cells move with respect 02:44.827 --> 02:49.487 to each other as they're changing shape and strength. 02:49.647 --> 02:52.287 A little introduction to the group that I worked 02:52.287 --> 02:54.567 with. So this is my grad student who did the 02:54.567 --> 02:58.387 work. This is a professor who studies more on the 02:58.387 --> 03:01.367 biology side, the biomechanics, and his postdoc. 03:02.107 --> 03:08.807 and this work was published last year so with that uh 03:08.807 --> 03:13.327 let's start with the the background so so the goal of 03:13.327 --> 03:15.827 this project was to create a robotic material which is 03:15.827 --> 03:17.607 kind of this nebulous term that people have thrown 03:17.607 --> 03:21.547 around for 30 years or so of it's basically a material 03:21.547 --> 03:24.027 you can imagine that could just take any shape and any 03:24.027 --> 03:26.327 any properties and you can have a blob of it and it 03:26.327 --> 03:30.267 could do anything um you know we've seen this in in movies 03:30.267 --> 03:34.447 uh terminator is kind of a version of this where and 03:34.447 --> 03:36.707 i think this is terminator 2 maybe he's kind of like 03:36.707 --> 03:40.867 this liquid metal thing you know so imagine imagine you 03:40.867 --> 03:44.717 could just uh make this material and and it could 03:44.717 --> 03:46.887 yeah i think this is a good one too he just his hands turn 03:46.887 --> 03:49.567 into these metal things and he turns into this little 03:49.567 --> 03:52.807 blob and and flows and i think that that particular 03:52.807 --> 03:55.967 moment there where he's kind of this flowing blob is 03:55.967 --> 03:58.367 kind of where our mind went with what we wanted to do. 03:58.847 --> 04:00.487 Similarly, more recently, 04:00.857 --> 04:04.207 Big Hero 6. These ones you at least kind of see a 04:04.207 --> 04:07.527 little more of the mechanisms going on, but same kind of 04:07.527 --> 04:10.267 idea. These things have structure and shape, but can 04:10.267 --> 04:15.907 also flow and move past one another at the same time. 04:16.007 --> 04:19.447 So these are a couple of Hollywood's interpretation 04:19.447 --> 04:22.507 of what we were thinking of doing. And then 04:22.657 --> 04:27.467 the question that we were trying to answer or kind 04:27.467 --> 04:30.307 of the goal is, can you make something that can 04:30.307 --> 04:34.567 both be strong, have some shape and a structure, 04:34.567 --> 04:37.937 but also at the same time flow into new shapes? 04:38.747 --> 04:41.477 And there's been a lot of collective robots, and 04:41.557 --> 04:43.767 I'll go through a few of them in a minute. But 04:43.767 --> 04:45.967 this, we thought, was still a challenge in the 04:45.967 --> 04:47.807 field, like something that could really support large 04:47.807 --> 04:50.787 loads, like even the weight of a person or 04:50.787 --> 04:52.947 something like that. but then also be able to flow. 04:54.267 --> 04:58.267 And so we look at current systems. 04:58.267 --> 04:59.647 I'll go through a few of them. 05:00.967 --> 05:06.047 And this is one where the robots essentially are 05:06.047 --> 05:09.207 able to drive around and connect and disconnect. 05:10.087 --> 05:13.007 And so they're able to change shape. 05:15.187 --> 05:19.207 But by being a driving robot, you can imagine if you 05:19.207 --> 05:21.207 had many of these, it doesn't really work anymore 05:21.207 --> 05:23.427 to flow because you have to basically disconnect to 05:23.427 --> 05:25.207 move. So you can only move around the perimeters. 05:25.527 --> 05:29.747 Similarly here, these are rolling, so they can 05:29.747 --> 05:31.887 roll over one another. But if you have too many of 05:31.887 --> 05:33.527 them in the center, you basically get trapped. 05:34.407 --> 05:38.007 And then there's kilobots, which is another nice one. 05:38.007 --> 05:42.307 And these ones, again, only move around the outside 05:42.307 --> 05:44.007 because they're basically driving on the surface. 05:44.947 --> 05:49.447 And they don't actually connect. So they're 05:49.447 --> 05:52.707 more of a shape forming without structure. 05:53.107 --> 05:56.747 So that's what's been done, and also many 05:56.747 --> 05:59.467 other things, but I just chose a few to show. 05:59.607 --> 06:02.067 And at the same time, there's some nice examples 06:02.067 --> 06:05.187 in nature, too. So this is morphogenesis, and 06:05.187 --> 06:07.227 this is what my collaborator, Uche, studies. 06:08.057 --> 06:10.507 And so if you zoom out and kind of think of 06:10.507 --> 06:12.487 this as a material, you're seeing there's these 06:12.487 --> 06:15.227 same kind of flows that we saw in Terminator, 06:15.227 --> 06:18.307 where this is a solid structure, but it's 06:18.307 --> 06:20.367 flowing and changing shape at the same time. 06:20.847 --> 06:23.167 And here's just another kind of cool 06:23.167 --> 06:26.027 3D view of the same kind of thing. 06:26.347 --> 06:29.607 So we're seeing this kind of amorphous shape. You 06:29.607 --> 06:31.627 can kind of start seeing some of the cell levels 06:31.627 --> 06:33.227 in there. And you're seeing these cells are 06:33.227 --> 06:35.427 shifting around with respect to each other, and you're 06:35.427 --> 06:37.787 getting a change in shape of the whole system. 06:38.507 --> 06:42.787 so underlying this behavior in nature are a couple 06:42.787 --> 06:48.687 of different things the first one is termed convergent 06:48.687 --> 06:52.377 extension and this is when you basically have 06:52.377 --> 06:55.327 this is like a block of the tissue and it goes 06:55.327 --> 06:58.287 from this shape and it basically shrinks this way and 06:58.627 --> 07:02.067 elongates this way and it's due to anisotropic 07:02.067 --> 07:05.667 stresses inside and if you zoom into the at the 07:05.667 --> 07:09.457 cellular level you'll see this intercalation, where 07:09.457 --> 07:13.407 these cells, 2 and 8, are moving in between 4 and 5. 07:13.767 --> 07:17.497 And this is kind of the essential movement 07:18.307 --> 07:21.327 rearrangement that kind of drives all of this motion. 07:21.887 --> 07:25.207 At the same time, there is this thing called tissue 07:25.207 --> 07:29.767 fluidization. And this is basically where the tissue can 07:29.767 --> 07:32.867 remain solid in some areas and fluid-like in other 07:32.867 --> 07:37.347 areas. And it has to do with there's fluctuations in the 07:37.347 --> 07:42.377 the forces between the the cells so you get more 07:42.377 --> 07:45.467 rearrangements over in the fluid -like part but both of these 07:45.467 --> 07:49.247 mechanisms are driven by this same t1 transition which 07:49.247 --> 07:51.947 I kind of showed and when we're looking at the cells 07:52.437 --> 07:55.167 and if you just in the simplest version there's just 07:55.167 --> 07:57.987 four cells and you have kind of a tall diamond and then 07:57.987 --> 08:03.307 you get to a squat diamond and if you if you look 08:03.307 --> 08:07.427 through you know in those videos of the the morphogenesis 08:07.427 --> 08:10.607 you'll be seeing hundreds and thousands of of these 08:10.607 --> 08:13.747 these t1 transitions happening all over the place and 08:13.747 --> 08:16.127 what this allows this is kind of the key that we were 08:16.127 --> 08:19.347 looking for is it allows you to change shape while 08:19.347 --> 08:23.607 maintaining structure and then you can do it deep inside 08:23.607 --> 08:25.867 the collective no longer are you trying to roll around 08:25.867 --> 08:28.867 the outside like the previous robots but you're able to 08:28.867 --> 08:32.807 change shape while you're surrounded around the outside 08:32.807 --> 08:35.027 because you're basically you know you're just sliding 08:35.027 --> 08:37.557 with respect to each other as opposed to rolling so 08:38.187 --> 08:40.987 that was kind of the key is is how do we make this t1 08:40.987 --> 08:44.987 transition so that's what we'll talk about next it's 08:44.987 --> 08:49.007 how to recreate it in a robot and that was the hardest 08:49.007 --> 08:53.027 part so we zoom in into the cell again and we abstract 08:53.027 --> 08:57.267 a little bit we we found three key mechanisms that are 08:57.267 --> 09:04.027 going on in here so first we have these forces so 09:04.027 --> 09:08.367 between two cells there are this inner unit force that 09:08.367 --> 09:10.407 they're pulling basically the cells are pulling on each 09:10.407 --> 09:14.447 other there's also adhesion so this is what gives us 09:14.447 --> 09:17.317 the structure everything's connected so those two cells 09:17.317 --> 09:20.787 are connected together and then finally there's a 09:20.787 --> 09:24.007 polarization which just means that all the cells know which 09:24.007 --> 09:27.567 direction there's some global directionality which 09:27.567 --> 09:30.387 allows them when they're doing these t1 transitions to 09:30.387 --> 09:33.007 have them all occur in the same direction so we can get 09:33.007 --> 09:38.527 macroscopic elongation so then in the robot we wanted 09:38.527 --> 09:41.587 to replicate or recreate in some way each of these 09:41.587 --> 09:46.887 three and so I'll go through these one at a time I'll 09:46.887 --> 09:48.767 introduce them and then I'll go through in more detail. 09:48.907 --> 09:53.807 But the interunit force we did by putting 09:53.807 --> 09:57.307 these small gears, motor-driven 09:57.307 --> 09:59.727 gears, around the periphery of our cell. 10:00.407 --> 10:03.447 And then adhesion is done with these magnets. 10:03.767 --> 10:06.367 They're free to rotate, so any two magnets can stick 10:06.367 --> 10:09.527 to any other magnet. And then polarization we did with 10:09.527 --> 10:12.007 photodiodes, and I'll explain that and polarize late. 10:12.167 --> 10:17.927 Okay, so we start with the T1 transition sorry with 10:17.927 --> 10:25.127 the motorized interunit force so we had to go from this 10:25.557 --> 10:29.127 what we knew from biology which is you know the goal 10:29.127 --> 10:32.057 is to get these top cells to come in and intercalate 10:32.057 --> 10:36.727 between the other two cells or as we saw it in this view 10:36.727 --> 10:40.397 the two and the eight coming in between the five and 10:41.047 --> 10:44.707 the one thing that my collaborator Uche does is he 10:44.707 --> 10:48.707 puts, he'll take out one cell and put a drop of oil 10:48.707 --> 10:51.487 where the cell was, and then he sees how, and actually 10:51.487 --> 10:55.137 this is similar to what Professor Goldman is doing in 10:55.137 --> 10:58.007 his lab with the water and the ants, he basically puts 10:58.007 --> 11:01.387 a drop of oil in there and then sees how that oil 11:01.387 --> 11:05.267 deforms, so then he can back out what are the forces the 11:05.267 --> 11:08.747 surrounding cells are applying to that drop of oil. And 11:08.747 --> 11:12.467 so it was the tiniest force sensor you can imagine. 11:12.877 --> 11:14.947 And with that, we can see the 11:14.947 --> 11:17.927 pattern of forces that are on a cell. 11:19.497 --> 11:22.047 And this would be like for the cell number 11:22.047 --> 11:24.087 five right there. It would be that pattern. 11:24.387 --> 11:28.867 And what we noticed is if we take that and 11:28.867 --> 11:31.987 we split it into quadrants, you can start 11:31.987 --> 11:34.567 seeing there's some patterns emerging. 11:36.367 --> 11:38.347 and in each of these quadrants 11:38.587 --> 11:39.467 there's 11:41.227 --> 11:42.587 basically going 11:42.667 --> 11:44.107 clockwise and counterclockwise 11:44.107 --> 11:45.567 and then opposite in the bottom 11:47.077 --> 11:50.067 and so we took that and we said well we tried 11:50.067 --> 11:51.767 many things and I'm skipping like a year's 11:51.767 --> 11:54.687 worth of prototypes and ideas because it's not 11:54.687 --> 11:57.567 obvious how you get because also in a cell there's 11:57.567 --> 12:00.727 actual flows happening and so there's like 12:02.007 --> 12:05.047 there's particles moving past one another to create 12:05.047 --> 12:08.187 these forces and in a robot that's rather hard to do and 12:08.187 --> 12:10.927 you know we thought about ways of doing crazy soft 12:10.927 --> 12:13.807 robots that that were flowing and things but in the end 12:13.807 --> 12:17.997 we we realized well you could replicate this with 12:18.607 --> 12:22.847 four years that are turning like this that you could get 12:22.847 --> 12:27.147 a good approximate this flow and then do it with a 12:27.147 --> 12:29.127 motorized wheel so one of our earlier prototypes look 12:29.127 --> 12:32.807 like this where it was four gears all driven off a 12:32.807 --> 12:35.027 single motor in the middle so you could just reverse the 12:35.027 --> 12:37.147 direction on some of them compared to the other ones 12:39.447 --> 12:44.427 later versions we ended up driving it with direct 12:44.427 --> 12:47.167 driving each of the motors turns out this and when 12:47.167 --> 12:48.927 I talk about polarization I'll say why that's 12:48.927 --> 12:51.127 important but it was nice to have full control of the 12:51.127 --> 12:55.767 system and so you could basically do that same pattern 12:55.767 --> 13:00.467 like this but by doing it you know selectively 13:00.687 --> 13:03.267 having those two robots or those two wheels turning 13:03.267 --> 13:06.507 one way and the other two trying the other way in a 13:06.507 --> 13:09.247 little video of how that guy looks when it's when it's 13:09.247 --> 13:11.957 turning and so so we're able to control each of those 13:11.957 --> 13:17.187 and then how you can so what's nice about this is as 13:17.187 --> 13:22.207 we vary the polarization we're able to change which 13:22.207 --> 13:23.947 direction each of the wheels is turning so we can 13:23.947 --> 13:27.467 basically change the which way the head of the robot in 13:27.467 --> 13:30.807 real time just by changing which way the the motors are 13:30.807 --> 13:32.687 turning and so as opposed to when it's all coupled 13:32.687 --> 13:35.027 there's always kind of one head to the robot when 13:35.027 --> 13:37.927 they're actually controlled I can I can turn anyone any 13:37.927 --> 13:40.647 direction and change which way is the head in real time 13:42.797 --> 13:47.067 okay so that's the inner unit force probably the the 13:47.067 --> 13:49.527 most complex and challenging part of the project was 13:49.527 --> 13:52.647 figuring this out this one's relatively easy we need 13:52.647 --> 13:55.787 them to stick together we had we added some magnets 13:56.267 --> 13:59.767 I think I have a little video of that just showing 13:59.767 --> 14:02.977 how these magnets can stick together what's nice is that 14:02.977 --> 14:05.777 any two magnets can stick to one another because 14:05.807 --> 14:09.467 they're they're captured but free to rotate so if you 14:09.467 --> 14:13.187 bring them close they can snap together so the robots 14:13.187 --> 14:15.667 don't need to have any particular alignment when they're 14:15.667 --> 14:19.427 sticking together, they can just find each other. 14:20.687 --> 14:24.247 And then the other thing we wanted to do with 14:24.247 --> 14:30.507 tuning the magnets is we wanted the collective, 14:30.507 --> 14:33.427 like I said, to be strong and have a structure. 14:33.427 --> 14:36.357 But we also wanted to flow past each other. So when 14:36.937 --> 14:40.707 we disconnect, when they're not in contact, we want 14:40.707 --> 14:43.207 them to be able to move easily. So we wanted a 14:43.207 --> 14:46.067 steep drop off, a high force, and then a steep drop 14:46.067 --> 14:51.387 -off. And so if we plot the magnet force as a function 14:51.387 --> 14:54.767 of distance, you get these nice drop-offs in kind 14:54.767 --> 15:00.227 of the nominal shape that we started with, with 15:00.227 --> 15:02.967 something like this, the red line, where you have two 15:02.967 --> 15:05.007 large magnets that are embedded in your structure. 15:05.007 --> 15:07.507 And they can give you some nominal force, and 15:07.507 --> 15:09.347 they have some nominal drop-off. But a large 15:09.347 --> 15:11.847 magnet has a slower drop -off than a small magnet. 15:11.927 --> 15:15.157 So what we found is instead, if we went to relatively 15:15.287 --> 15:17.327 small magnets but made sure they were as close 15:17.327 --> 15:21.027 as possible, we could match the force but get a quicker 15:21.027 --> 15:23.247 drop-off as we pull them apart. And so that turned 15:23.247 --> 15:26.307 out to be superior to a bigger embedded magnet. 15:30.327 --> 15:33.147 Okay, so that's our adhesion. And then the last one, 15:33.147 --> 15:35.507 which is fun, is how we did polarization. So we need 15:35.507 --> 15:38.307 to tell the robots, you know, which way is north. 15:38.727 --> 15:40.907 because they all need to kind of know which 15:40.907 --> 15:43.147 way, if we want to do this elongation, 15:43.867 --> 15:46.687 each of the cells needs to know which way that is 15:46.687 --> 15:50.127 in order to intercalate in the correct direction. 15:50.987 --> 15:54.027 So this is my student's idea. We tried a bunch of 15:54.027 --> 15:56.647 things. Again, I'll skip all the failed things, 15:56.647 --> 16:00.817 which there was one that was like a strobe light with 16:01.187 --> 16:05.387 stripes that was very seizure-inducing, and so I'm 16:05.387 --> 16:07.307 glad we went away from that because all the videos 16:07.307 --> 16:10.167 were kind of like, oh, Matt, I can't watch that. 16:10.287 --> 16:12.927 So anyway, this solution was quite clever. 16:12.927 --> 16:16.457 So what he did is he put linear polarizing 16:16.647 --> 16:19.607 film on eight photodiodes that are 16:19.607 --> 16:22.067 arranged around the perimeter of the robot. 16:22.307 --> 16:24.927 And then he shines the light. 16:24.947 --> 16:26.387 So in all the videos you'll 16:26.387 --> 16:28.527 see, the light is polarized. 16:29.187 --> 16:32.027 And what happens is you get cross-polarization 16:32.027 --> 16:35.657 on two of them. And so that photodiode 16:35.657 --> 16:37.987 all of a sudden gets very dark. And so it's 16:37.987 --> 16:41.887 a very robust way to tell which way is up. 16:43.227 --> 16:46.087 And then he can control it just by rotating 16:46.087 --> 16:48.967 the source light and then tell which 16:48.967 --> 16:51.747 direction he wants the robots to align. 16:52.117 --> 16:55.127 So it's kind of a cool video he made. So this 16:55.127 --> 16:57.767 is, again, being lit with polarized light. And 16:57.767 --> 17:00.507 then you can see the filters as they cross-polarize, 17:00.507 --> 17:02.787 turn dark, and at the top they turn light. 17:05.887 --> 17:09.367 and then just a little picture of them, 17:09.547 --> 17:11.687 assemble, and you can see on that 17:11.687 --> 17:16.487 top PCB the polarizing filters. 17:18.767 --> 17:23.167 Okay, so those are kind of the three main attributes 17:23.167 --> 17:26.167 that we put into our robot, and then the question was, 17:26.167 --> 17:28.947 well, after all that, did we get the T1 transition? 17:28.947 --> 17:31.427 So here's our robot. 17:31.487 --> 17:35.107 So this is with the light shining this way. 17:35.797 --> 17:37.497 It looks a little bit like Terminator. 17:38.967 --> 17:40.047 There they go. 17:40.327 --> 17:42.867 So they can do this T1 transition. 17:44.747 --> 17:49.547 And quite robust to orientation and alignment. If 17:49.547 --> 17:52.147 you notice, none of the robots are in any particular 17:52.147 --> 17:53.487 alignment. You can kind of see where their boards 17:53.487 --> 17:55.947 are. You just throw them in there. You shine the 17:55.947 --> 17:59.907 light. They all know which way to rotate their motors. 17:59.907 --> 18:03.207 and then because of because they all have their 18:03.207 --> 18:05.427 motors rotating in a certain pattern they create 18:05.427 --> 18:10.967 this flow okay so we show we can make it and then the 18:10.967 --> 18:14.447 next step was controlling it so we we started with 18:14.447 --> 18:17.567 controlling shape we're into both shape and strength 18:17.687 --> 18:21.307 and so the key question is how do we if we have 18:21.307 --> 18:24.767 a a set of these, starting just with that T1 transition, 18:25.937 --> 18:30.947 can we create the T1 in a controlled manner? 18:31.547 --> 18:37.087 And so the main thing we were varying, there's 18:37.087 --> 18:38.407 many things you could vary in this robot, 18:38.407 --> 18:40.987 including, for instance, the adhesion strength, 18:40.987 --> 18:42.967 and that would change how they change shape. 18:43.207 --> 18:47.547 But we focused on the force that the motors are 18:47.547 --> 18:51.327 applying to the neighboring robot through the gears all 18:51.327 --> 18:53.987 right so we've kind of drawn it out here so there's 18:53.987 --> 18:56.267 gonna be a lot of F's and plots with F's in them in 18:56.267 --> 18:58.707 the next coming slides this is the force we're talking 18:58.707 --> 19:01.547 about it's it's the it's the interunit force applied 19:01.547 --> 19:05.687 through the gears and we have two we can do with 19:05.687 --> 19:07.367 this force or that we did with this force is lots 19:07.367 --> 19:10.087 of things you could you could vary but we varied the 19:10.287 --> 19:15.627 the average force so if you look at that force across 19:15.627 --> 19:20.287 time being applied to the to the robots there's the 19:20.287 --> 19:26.787 average force and then now upon how they look 19:27.867 --> 19:31.367 something like this so they're not we're not applying a 19:31.367 --> 19:34.247 steady force or applying this fluctuating force like 19:34.247 --> 19:38.987 they do in the cells and so there's an average and 19:38.987 --> 19:43.647 then we also define this Delta F which is having to 19:43.647 --> 19:49.267 do with the the amplitude of the oscillation and you 19:49.267 --> 19:52.007 know we could have changed the frequency we did not 19:52.207 --> 19:54.687 there was a limited number of years my 19:54.687 --> 19:56.907 grad student would do experiments so 19:57.187 --> 20:00.447 these are the two parameters we stuck with okay so 20:00.447 --> 20:04.447 but I'll show you how the these parameters affect shape 20:04.447 --> 20:08.487 change and so in this experiment is up in the corner 20:08.487 --> 20:11.807 this is again this t1 transition we would start 20:11.807 --> 20:16.287 them in one position we would apply some set of conditions 20:16.287 --> 20:18.617 and we see would they what was the chance of them 20:18.617 --> 20:21.187 intercalating you know in over 10 experiments 20:21.187 --> 20:24.007 something like that and so this is what happens as we vary 20:24.007 --> 20:26.547 the average force you get this nonlinear response 20:26.547 --> 20:28.987 where at first they don't move at all and they just 20:28.987 --> 20:31.927 sit there and they never would move and then there's 20:31.927 --> 20:34.087 this region where there's a chance they'll move and 20:34.087 --> 20:37.027 then you get get enough average force and then they 20:37.027 --> 20:41.747 always move so we see this similarly with the force 20:41.747 --> 20:44.987 fluctuation so in this case we keep the same average 20:44.987 --> 20:47.847 force so the yellow line remains the same we're changing 20:47.847 --> 20:50.267 the magnitude of the fluctuation we see almost a 20:50.267 --> 20:53.427 very similar pattern where you get no chance of moving 20:53.427 --> 20:55.927 a chance of moving in the middle and then every 20:55.927 --> 20:59.837 time moves through at the top so you can now imagine 21:00.647 --> 21:03.107 you could ask your grad student not just to make two 21:03.107 --> 21:05.807 plots but sweep through all the possible parameters of 21:05.807 --> 21:11.607 delta F and average force and make this plot, which 21:11.607 --> 21:15.747 represents many, many weeks of data collection, but 21:15.747 --> 21:19.427 turned out beautifully. So we see basically down here, 21:19.427 --> 21:23.647 as you have low normalized force and low delta F, 21:24.267 --> 21:26.807 then you get no chance of moving through. 21:26.807 --> 21:29.027 And then there's this transition region and 21:29.027 --> 21:31.207 the chance of always going through there. 21:33.207 --> 21:35.507 And then you can think of these are basically slices 21:35.507 --> 21:41.787 of this plot too and then we ran a simulation so 21:42.027 --> 21:46.927 Uche's postdoc is great at simulations I won't go 21:46.927 --> 21:49.727 in the details here but he basically built it I mean 21:49.727 --> 21:52.507 these are relatively simple things to simulate and 21:52.957 --> 21:57.967 he got a very similar result on on his and his 21:57.967 --> 22:01.907 little space space plot of course he can sample a lot 22:01.907 --> 22:05.127 more data points than we did but a similar trend 22:06.027 --> 22:09.827 so that was shape change so now look at a strength 22:09.827 --> 22:14.187 change and so in this case is about you take three 22:14.187 --> 22:17.787 robots and you put a weight on top of them and then 22:17.787 --> 22:20.667 you see does the weight crush the robots or not and 22:23.047 --> 22:25.577 it looks kind of like this so you have the robots 22:26.967 --> 22:29.907 you turn them on and they essentially soften 22:30.267 --> 22:33.247 and it falls I will also say this is 22:33.647 --> 22:35.807 credit to my grad student who made robots 22:35.807 --> 22:37.927 robust enough that he could drop large 22:37.927 --> 22:40.247 weights on them and they would keep ticking 22:41.347 --> 22:43.147 so I put that up in the corner just so you 22:43.147 --> 22:44.907 remember we're talking about what we're calling 22:44.907 --> 22:47.267 the critical force is what's the force needed 22:47.267 --> 22:49.787 to go from this triangle to flattened out 22:52.047 --> 22:54.787 this is basically the force of every range of robots 22:54.867 --> 22:58.287 And it looks like this. So this is with 22:58.287 --> 23:00.227 everything off, just basically the experiment 23:00.227 --> 23:03.987 when they're passive and you load it up to 23:03.987 --> 23:07.887 140 newtons and eventually they yield. There's 23:07.887 --> 23:09.487 kind of this second peak, which is weird. 23:09.647 --> 23:12.207 There's kind of a semi-stable or another 23:12.207 --> 23:13.947 stable state they get to. It kind of looks 23:13.947 --> 23:16.227 like that in the middle there when the 23:16.227 --> 23:19.027 magnets engage again and then it gives out. 23:19.307 --> 23:22.667 But the thing we're going to be plotting is that 23:22.667 --> 23:26.707 first peak of when they go from solid into failed. 23:28.527 --> 23:30.887 And again, just to remind you, we're 23:31.257 --> 23:32.737 playing with these two parameters. 23:33.047 --> 23:35.187 So I'll just throw those back up there. 23:35.567 --> 23:40.427 Average force, and then this delta F, 23:40.427 --> 23:42.197 which is the magnitude of the fluctuations. 23:43.927 --> 23:44.847 OK. 23:46.887 --> 23:52.807 OK, so then for all of the shape change ones, we were 23:52.807 --> 23:54.527 telling them the polarization direction, because 23:54.527 --> 23:56.607 they needed to know. All of the four needed to know. 23:56.607 --> 24:01.847 For strength change, we do not. So this is a random 24:01.847 --> 24:04.787 orientation. So when we turn them on, they all 24:04.787 --> 24:07.067 are facing in different directions when they go. 24:07.227 --> 24:08.027 OK, 24:09.027 --> 24:11.807 just to set up the experiment. And then this is what 24:11.807 --> 24:14.607 the plots look like. So this is now the critical force 24:14.607 --> 24:19.427 as a function of the normalized inter-unit force. 24:19.707 --> 24:22.107 And we see it drops off. So that one 24:22.107 --> 24:24.447 is normalized. That's the 140 Newton 24:24.887 --> 24:27.507 case. And then it drops off from there. 24:28.247 --> 24:31.487 Similarly, with the delta F, we see a drop off. 24:31.707 --> 24:38.147 This is at some normalized force here, 24:38.147 --> 24:39.707 so something around here. So it doesn't 24:39.707 --> 24:41.207 go as high because they're already 24:41.307 --> 24:44.807 moving even with no vibration and so you don't go all 24:44.807 --> 24:47.867 the way up but you see a similar trend of dropping off 24:47.867 --> 24:54.267 and once again you did a nice sweep and this one's a 24:54.267 --> 24:56.687 little less clear the other one kind of had this nice 24:56.687 --> 25:00.387 diagonal line through it but you do you still see the 25:00.387 --> 25:02.947 same trend as you go from the bottom corner up to the 25:02.947 --> 25:08.307 top corner of the critical force dropping off and again 25:08.307 --> 25:12.167 we did a simulation which shows a similar trend and 25:12.167 --> 25:15.747 again like the experiments we don't see as extreme of 25:15.747 --> 25:21.047 a diagonal where it's more of a the trend is more just 25:21.047 --> 25:24.087 across the vibrations than the magnitude of the force 25:26.047 --> 25:28.587 and we did one last thing I'll just mention briefly 25:28.587 --> 25:32.667 because I find it interesting as we sweep through 25:32.667 --> 25:34.867 these how does it affect power consumption so 25:36.367 --> 25:40.147 So for shape change, as we go from here to there, I 25:40.147 --> 25:41.927 mean, you can imagine you can do this kind of the brute 25:41.927 --> 25:44.527 force way, which is you just turn on your motor 25:44.527 --> 25:47.747 strong enough to force through, or you do the case where 25:47.747 --> 25:51.807 you do a lower total force and you vibrate and fluctuate. 25:51.807 --> 25:55.227 And do you save, can you do it with less power? 25:55.847 --> 25:57.547 And it turns out, yes, 25:57.867 --> 25:59.207 significantly less power. 25:59.817 --> 26:01.637 So this is for the shape change. 26:03.167 --> 26:07.247 as we vibrate we can drop the power by a factor 26:07.247 --> 26:11.267 of two or so two or three similarly with the 26:11.267 --> 26:14.387 strength change experiment when we're we we want to 26:14.387 --> 26:17.367 soften the material or lower the critical force 26:17.647 --> 26:21.257 you get even a larger power now also 26:21.407 --> 26:25.367 mentioned that this is power not energy turns 26:25.367 --> 26:28.327 out energy is roughly constant in the two 26:28.327 --> 26:30.667 cases because this one takes much longer. 26:30.747 --> 26:33.727 So it's kind of an interesting finding. 26:33.727 --> 26:37.967 It still might be interesting if you are, for 26:37.967 --> 26:41.447 instance, a power-limited system and you only have a 26:41.447 --> 26:44.607 certain amount of power you can get from either your 26:44.677 --> 26:48.567 batteries or your motors, or in biology you have 26:48.567 --> 26:52.827 some flow ATP and you basically can't produce a power 26:52.827 --> 26:54.847 more than this, then fluctuating is an interesting 26:54.847 --> 26:56.767 way to do things you couldn't have done before. 26:57.447 --> 26:59.547 So anyway, that's power stuff. 27:00.527 --> 27:03.227 And then this talk was supposed to be about a 27:03.227 --> 27:05.947 collective, so I should show more than four robots. 27:07.807 --> 27:09.087 So again, 27:09.327 --> 27:12.827 limited by years of grad student years on this project, 27:12.827 --> 27:16.447 we settled on 20 robots as a number that was like, 27:16.447 --> 27:19.487 yeah, we can say that's a collective, but you're not 27:19.487 --> 27:21.807 going to be here forever. So this is what it looks like. 27:22.787 --> 27:25.987 and he should turn the light on and then 27:25.987 --> 27:29.207 he's polarizing across the page this 27:29.207 --> 27:31.667 way so you can see them do their thing 27:32.707 --> 27:35.367 and again I'll note they had no particular 27:35.367 --> 27:38.447 alignment to start I'll play that one more time 27:40.747 --> 27:46.067 so when you see he basically just threw these in 27:46.067 --> 27:48.787 here he put them into a hex pattern but none of 27:48.787 --> 27:50.567 the robots are oriented in any particular way and 27:50.567 --> 27:52.087 you can just kind of throw them in and then at 27:52.087 --> 27:53.847 this point you could grab robots or take them out 27:53.847 --> 27:56.647 like they don't know where any of no one knows 27:56.647 --> 27:59.027 where anyone else is they're basically all just 27:59.027 --> 28:02.307 blindly doing the same thing and then this emergent 28:02.767 --> 28:06.307 elongation occurs so in real life is a little messy 28:06.307 --> 28:08.547 you can see there's there's some that turn into a 28:08.547 --> 28:12.447 complete line in some areas they're kind of by twos in 28:12.847 --> 28:17.007 the the simulation you get a little cleaner result 28:17.347 --> 28:20.347 or you get something that It does like this. 28:23.047 --> 28:24.567 And if you keep running it, it 28:24.567 --> 28:26.527 will run out fully into a line. 28:29.647 --> 28:32.627 We looked at how the rearrangements. 28:34.447 --> 28:36.927 First, we looked at T1 transitions. But it turns out 28:36.927 --> 28:42.047 not all the rearrangements are the exact definition of 28:42.047 --> 28:44.867 the T1. Sometimes you have three cells moving to two. 28:44.867 --> 28:46.447 So it gets a little messy. So we just looked at 28:46.447 --> 28:49.957 rearrangements. so changes in topology so this is the initial 28:49.957 --> 28:53.427 topology this is somewhere further and for an example 28:53.427 --> 28:56.857 here this one had 12 rearrangements this one had 25 28:56.857 --> 29:02.187 and you can just plot those as you can correlate those 29:02.187 --> 29:06.367 to the shape essentially and more rearrangements you 29:06.367 --> 29:08.887 are getting a you're changing the aspect ratio which is 29:08.887 --> 29:13.527 showing that your rearrangements are occurring in some 29:13.527 --> 29:15.977 kind of ordered way that you're getting elongation. 29:17.447 --> 29:20.567 Also, we can look at how the fluctuation has 29:20.567 --> 29:22.707 changed its number of rearrangements. So as we 29:22.707 --> 29:24.827 would expect from our previous results, you 29:24.827 --> 29:30.427 see more rearrangements as you add more fluctuation. 29:30.427 --> 29:33.187 And then as you plot them, the aspect 29:33.187 --> 29:35.727 ratio versus time with different fluctuations, 29:35.727 --> 29:37.967 there's all kind of plots looking at the 29:37.967 --> 29:40.667 same effect but slicing it in different ways. 29:40.847 --> 29:43.727 Basically, you see we get more aspect ratio change as 29:43.727 --> 29:47.147 we fluctuate more so the the the fluctuations are 29:47.147 --> 29:53.727 helping with the shape change and so that was shape 29:53.727 --> 29:57.347 change we can also do this with strength so now we instead 29:57.347 --> 30:00.027 of our three robots we put all 20 of them and stick 30:00.027 --> 30:04.067 a mass on the top and we can turn on the fluctuations 30:05.247 --> 30:10.947 and we can melt the material so we can look at how 30:10.947 --> 30:14.727 that the we're now calling it a yield stress because 30:14.727 --> 30:17.767 there's enough particles to kind of simulate a stress 30:17.767 --> 30:23.227 -like thing in units of newtons per unit and a unit 30:23.227 --> 30:28.507 being the the robot number of robots so this is loading 30:28.507 --> 30:31.747 it up till it breaks this was quite an experiment 30:31.747 --> 30:35.167 because now it's not three robots it's a lot I forget 30:35.167 --> 30:39.637 how many yeah hundreds of newtons of force here and 30:40.067 --> 30:42.307 then And then when you turn on the fluctuations, you 30:42.307 --> 30:46.307 get almost a fluid-like behavior in the extreme case. 30:47.447 --> 30:49.307 And then you can do a similar experiment where 30:49.307 --> 30:52.667 you vary the amount of fluctuation, and you can see 30:52.667 --> 30:56.387 the drop off of yield stress with fluctuation. 31:00.407 --> 31:03.597 And because we're controlling with light, we can control 31:03.947 --> 31:07.467 certain areas and not other areas. So you can imagine 31:07.467 --> 31:11.107 shining a light on just this area, or shining the 31:11.107 --> 31:15.327 polarizing on this area, and we can have this area 31:15.327 --> 31:20.487 remain off in this area fluctuating and essentially get 31:20.487 --> 31:24.147 this elongation, which is, you know, a far cry from 31:24.357 --> 31:29.067 embryo, but you can imagine how this is, in essence, 31:29.067 --> 31:31.147 there's parts of the robot or parts of the collective 31:31.147 --> 31:34.107 that can remain static and part of it that are flowing. 31:34.777 --> 31:36.207 We knew the same thing with strength 31:36.207 --> 31:37.767 change, so we put two weights on. 31:37.887 --> 31:39.407 And the question is, can we 31:39.407 --> 31:41.607 drop one and hold the other one? 31:42.327 --> 31:45.827 So we can, again, fluctuate one side and 31:45.827 --> 31:48.367 drop that weight and keep this one on top. 31:51.267 --> 31:52.107 OK. 31:52.127 --> 31:58.227 So that was all of the experiments. And 31:58.227 --> 32:00.667 of course, any good robot paper wouldn't 32:00.667 --> 32:02.367 be complete without a demonstration. 32:02.467 --> 32:05.147 And so my grad student then 32:05.147 --> 32:06.647 spent a lot of time doing these. 32:07.587 --> 32:09.207 This is his favorite part, though. 32:09.447 --> 32:11.477 As you'll see, they're creative. 32:11.477 --> 32:13.867 So how could you use these things? Again, 32:14.067 --> 32:16.147 I think that the long -term vision is these would 32:16.147 --> 32:19.327 be microscopic and flowing like that metal in 32:19.327 --> 32:21.807 Terminator. Right now we have 20 of them. So 32:22.017 --> 32:23.627 you're going to have to use your imagination a 32:23.627 --> 32:25.527 little bit to kind of see where this could go. 32:26.167 --> 32:30.247 But here's what he came up with. So he wanted to do 32:30.247 --> 32:32.527 kind of a bridge building type thing where you could 32:32.527 --> 32:36.967 form these pillars and kind of shape them and 32:36.967 --> 32:40.347 then bring them together, and then solidify it, and 32:40.347 --> 32:43.767 then apply a large load. So he's kind of had it flowing 32:43.767 --> 32:46.827 and then solid. He's doing this by time, sorry. 32:47.717 --> 32:49.647 This is a great question. How is he doing 32:49.647 --> 32:52.287 this? This is grad student, multiple 32:52.287 --> 32:55.107 lights, trying 100 times. Exactly. Yep. 32:55.527 --> 32:58.747 Not the robot sound like individual, right? Correct. 32:58.747 --> 33:01.067 So they all have the same rule, which is 33:01.067 --> 33:04.447 do this rotation that we programmed in with 33:04.447 --> 33:06.607 a certain fluctuation and a certain force 33:08.007 --> 33:10.847 and do it in the orientation that you sense 33:10.847 --> 33:12.807 the light. So I think the most he got to 33:12.807 --> 33:14.587 was two lights. I think this is two lights. 33:15.047 --> 33:19.147 And so basically, these two halves are getting 33:19.147 --> 33:21.327 different signals, but they have the same brains. 33:24.907 --> 33:27.227 Currently, yeah, this is a great question. 33:27.227 --> 33:30.447 and ripe for future work, I would say. 33:30.567 --> 33:32.217 Yeah, yeah, absolutely. 33:33.967 --> 33:36.487 I'll keep going, but yeah, feel free to interrupt me. 33:36.507 --> 33:39.767 He did an injury healing, so basically took 33:39.767 --> 33:41.967 some of the robots out of the middle and then 33:41.967 --> 33:44.947 just turned them on, let them fluctuate, 33:44.947 --> 33:49.167 and then they come back into a healed state. 33:52.647 --> 33:54.987 This one's a bit of a stretch. He made a gripper. 33:55.427 --> 33:59.387 So this is his robot gripper. And he elongates, 33:59.387 --> 34:03.047 sticks onto this thing, pulls it across, 34:03.607 --> 34:04.807 dumps it here. 34:05.567 --> 34:07.487 And then he changes the shape of the 34:07.487 --> 34:10.207 gripper to make it wider, grabs the 34:10.207 --> 34:11.947 bigger object, and drops it off here. 34:12.967 --> 34:14.727 And this is all done. 34:15.407 --> 34:17.647 You might notice there's kind of a weird texture 34:17.647 --> 34:20.207 going on in the background. This is like on a surface. 34:20.447 --> 34:22.227 So this is not vertical. This 34:22.227 --> 34:24.027 is laying down on a table. 34:24.627 --> 34:27.747 But there was too much friction to actually pull the 34:27.747 --> 34:30.377 things along. So this is covered in glass microbeads. 34:30.377 --> 34:33.107 So this is like the least friction you could get. 34:33.107 --> 34:35.427 And that's where the weird surface texture is from. 34:36.027 --> 34:37.567 This is all magnets. 34:37.567 --> 34:39.107 So, yeah, they have magnets. And 34:39.107 --> 34:40.447 so you put some magnets in there. 34:41.147 --> 34:43.487 Yeah, yeah, yeah. There's a lot going on here. 34:44.047 --> 34:45.407 A lot of cheating. 34:47.087 --> 34:49.167 That one's the biggest stretch. 34:49.677 --> 34:51.107 And I like this one. I think this is my 34:51.107 --> 34:54.027 favorite. This is the last one. um so this 34:54.027 --> 34:56.927 one he termed it manipulation so the team 34:56.927 --> 34:59.867 is trying to get that thing into there 34:59.967 --> 35:01.767 and they work really hard 35:03.567 --> 35:05.347 and then they don't quite get it 35:05.347 --> 35:09.307 but then but then they come back yay 35:10.367 --> 35:12.707 i just think they're like so cute you're like cheering 35:12.707 --> 35:15.907 for them to to to get in there and then this is one 35:15.907 --> 35:18.427 of my favorite parts we we lost the gear over here 35:19.747 --> 35:22.447 so that's them doing some manipulation again 35:22.447 --> 35:25.487 the control of that is like yeah a lot 35:25.487 --> 35:28.067 of Matt knowing how to control these Hank 35:28.067 --> 35:30.327 another lost gear this was a hard experiment 35:31.457 --> 35:33.527 so yeah don't ask him to do that one again 35:34.327 --> 35:36.587 okay so I think I'll just wrap up 35:37.627 --> 35:40.987 so yeah I mean basically just summarize how we 35:40.987 --> 35:44.287 had this collective and the key takeaway is we want 35:44.287 --> 35:45.777 to make something that could form a structure 35:45.927 --> 35:47.607 but could also flow 35:49.357 --> 35:51.607 then as I mentioned a lot of future considerations 35:51.607 --> 35:55.647 this is definitely a first kind of pass at this 35:56.447 --> 35:59.107 we want to scale size so these things I should have 35:59.107 --> 36:00.827 brought some with me they're about you know hockey 36:00.827 --> 36:04.187 puck ish size so smaller would be very interesting 36:04.747 --> 36:07.747 and then with that number I say also 36:07.747 --> 36:10.627 three dimensions is a big next step 36:10.627 --> 36:13.767 how can we do it not just on the plane 36:14.147 --> 36:18.067 a lot of those demos were either on a table or on 36:18.067 --> 36:22.727 a vertical surface or at some tilt angle between. 36:23.687 --> 36:25.267 And then, obviously, I think, 36:25.267 --> 36:27.407 as Professor Goldman mentioned, 36:27.507 --> 36:30.787 some way to control these things other 36:30.787 --> 36:32.767 than my grad student with flashlights. 36:34.187 --> 36:35.527 Okay, and I think I'll pause 36:35.527 --> 36:37.707 there and take any questions. 36:48.717 --> 36:49.257 Yep. 36:50.157 --> 36:51.077 Great 36:59.097 --> 37:01.677 question. So we did constant. 37:02.077 --> 37:03.977 I think there's a whole other set of 37:03.977 --> 37:06.097 experiments where you could vary it. 37:06.097 --> 37:09.257 And I think you could get control like we were 37:09.257 --> 37:11.517 showing with all those plots of varying forces. You 37:11.517 --> 37:14.477 would see similar plots as you varied adhesion. 37:15.557 --> 37:19.197 We also spent quite a bit of time tuning 37:19.197 --> 37:20.637 that magnet, which I didn't talk about 37:20.637 --> 37:23.057 here. It's a very specific magnet force. 37:24.357 --> 37:26.037 If it's too small, 37:26.217 --> 37:30.117 they just kind of fall apart. You can put them 37:30.117 --> 37:31.757 together, and then they start doing their thing, and they 37:31.757 --> 37:34.397 just fall apart like this. And then if it's too 37:34.397 --> 37:37.147 strong, they basically can't move. All right, so 37:37.147 --> 37:40.317 there's definitely a window of that magnet force that's 37:40.317 --> 37:44.277 useful, and it has to do with basically how strong 37:44.277 --> 37:46.597 your motors are. So these things are all coupled, 37:46.597 --> 37:50.037 like the force and the magnet, and so there's probably 37:50.037 --> 37:51.977 some nice nondimensional terms you'd come up with 37:51.977 --> 37:54.337 to say how strong you want your magnets. But yeah, 37:54.477 --> 37:55.817 not currently, but it's 37:55.817 --> 37:56.757 definitely something you could do. 37:56.937 --> 37:57.737 Yeah, 38:02.157 --> 38:02.837 Greg? 38:09.977 --> 38:10.817 Yeah, 38:12.677 --> 38:15.317 I would say, 38:16.477 --> 38:18.797 yeah, yeah, yeah, yeah. 38:20.057 --> 38:22.997 Yep, yep, yeah, yeah, yeah. No, we have thought 38:22.997 --> 38:25.077 about, could you just make these things like 38:25.077 --> 38:27.597 yay big and make bridges or whatever out of 38:27.597 --> 38:29.977 them. I think that's actually an easier problem. 38:31.277 --> 38:33.457 Yeah, I would say we could go down 38:33.457 --> 38:37.097 a factor of 3 or 4 relatively easy, 38:37.497 --> 38:38.737 like pager motor 38:39.117 --> 38:39.957 size. 38:39.957 --> 38:41.377 I mean, you basically keep it all the same, 38:41.377 --> 38:43.277 just put tiny motors in and you'll be set. 38:44.377 --> 38:46.617 Beyond that, you're probably going to have 38:46.617 --> 38:47.937 to go away from electromagnetic motors. 38:48.137 --> 38:50.057 But there's other motors, I mean, 38:50.057 --> 38:51.637 piezo, solid state things, I could 38:51.637 --> 38:54.237 imagine doing similar things like this. 38:58.757 --> 39:01.297 yeah I think in the long run the light 39:01.297 --> 39:03.077 solution as Dan mentioned is probably 39:03.077 --> 39:06.497 not the way to do this because you need 39:06.707 --> 39:07.837 they all need to be looking 39:07.837 --> 39:09.397 at you they need to see you 39:11.977 --> 39:13.897 it's nice precision yeah 39:14.137 --> 39:15.017 yeah 39:17.477 --> 39:18.797 no it's not 39:19.097 --> 39:19.977 yeah 39:20.417 --> 39:23.097 it gets trickier in 3D but you could imagine 39:23.097 --> 39:25.717 multiple lights coming in and they're clear and 39:25.717 --> 39:27.757 when do they cross? You could do some things. 39:28.037 --> 39:30.617 I think 3D is also a big challenge. 39:31.397 --> 39:33.757 I think another thing is probably you'll need 39:33.757 --> 39:36.317 compliance. I didn't go into this too much, but 39:36.697 --> 39:39.097 at the scale we were at, we could get 39:39.417 --> 39:42.397 away with rigid and we could still get 39:43.097 --> 39:44.377 bulk movements. 39:45.157 --> 39:48.637 I think at the million level scale, you're 39:48.637 --> 39:50.257 just going to entrap your center ones 39:50.257 --> 39:51.457 and they're not going to be able to move. 39:51.717 --> 39:53.877 Then you probably need the cells are basically 39:53.877 --> 39:57.517 deforming when they're doing this and so I think 39:57.517 --> 39:59.097 we'll probably need a compliant version which 39:59.097 --> 40:01.657 we've we've built and tried and we said that's a 40:01.657 --> 40:04.297 future paper because it's it adds complexity yeah 40:07.097 --> 40:07.957 yeah 40:22.537 --> 40:24.117 But how do you physically get 40:24.117 --> 40:25.837 that to work for the robots? 40:26.297 --> 40:31.187 Yeah. So OK. So in terms of where this came from is, 40:33.397 --> 40:37.317 like I said, it's from the data that Uche took with his 40:37.317 --> 40:40.417 little droplets. He actually sees these fluctuations 40:40.637 --> 40:44.217 in the forces, the interunit forces, intercellular 40:44.217 --> 40:48.117 forces. And so they've hypothesized why those exist. 40:48.117 --> 40:50.717 And so I think part of this was, let's build a 40:50.717 --> 40:54.087 system. I mean, this is pure robophysics from Dan. 40:54.977 --> 40:56.777 Let's build a system and just play with 40:56.777 --> 40:58.117 it and see what happens. And I think the 40:58.117 --> 40:59.237 one other thing that fell out of it was 40:59.237 --> 41:02.697 the power result, which we didn't predict. 41:03.597 --> 41:09.157 And maybe that's – just now they're starting to be 41:09.157 --> 41:13.857 able to do cellular, cell -level metabolics. And so 41:13.857 --> 41:15.857 maybe they're going to be able to test some of this. 41:15.937 --> 41:20.517 Are we seeing reductions in power when you fluctuate 41:20.517 --> 41:22.397 it or if you knock that out and don't let it fluctuate, 41:22.397 --> 41:24.897 does it require more power or can it just not move? 41:26.577 --> 41:29.517 How do we do this? I mean, in a robot 41:29.517 --> 41:31.157 site, it's pretty easy. We just send 41:31.157 --> 41:38.027 a signal, PWM, to modulate the torque. 41:38.057 --> 41:39.777 They basically modulate the voltage, 41:39.777 --> 41:43.097 and then that modulates the output from the motor. 41:43.997 --> 41:45.717 and we have direct control over 41:45.717 --> 41:47.317 each motor so we can just basically 41:47.497 --> 41:48.897 fluctuate it they're always 41:48.897 --> 41:50.477 turning in the same direction 41:50.557 --> 41:55.597 but it's just the torque as it's going in that direction 41:55.597 --> 41:57.437 so we're not modulating again there's a million things 41:57.437 --> 42:00.517 we could modulate we could change the speed you could 42:00.517 --> 42:02.817 change the direction we don't do that you could 42:02.817 --> 42:05.997 change the spacing here this is kind of just a schematic 42:05.997 --> 42:08.677 but in reality this spacing is always constant so we 42:08.677 --> 42:12.607 keep a constant frequency you could vary that in time 42:12.607 --> 42:15.457 There's a lot of things that you could continue to test. 42:17.417 --> 42:18.457 Correct. 42:20.137 --> 42:23.197 No, okay, so this spacing was constant, 42:23.197 --> 42:26.117 yep, but this was not constant. 42:26.437 --> 42:33.177 We have Gaussian distributions around two points 42:33.647 --> 42:36.557 the same distance above and below the mean. 42:36.617 --> 42:38.277 Does that make sense? 42:39.577 --> 42:43.577 Exactly. And so you basically, if you look at all the 42:43.577 --> 42:48.417 things below the mean, they will fall around some level 42:48.417 --> 42:50.857 here, the same on the other side. And then we can 42:50.857 --> 42:55.097 move where those means are with respect to the center. 43:07.217 --> 43:10.137 With the strength change or shape change? 43:11.297 --> 43:13.457 Was it this one you're thinking or this one? 43:14.567 --> 43:16.097 Blue one? Yeah, yeah, yeah. Perfect. Yeah. 43:19.297 --> 43:21.297 Could be, yeah, yeah, yeah, yeah, yeah. 43:21.797 --> 43:25.177 Like averaged across some noisiness. I think it has 43:25.177 --> 43:27.877 to do with that, which is why the magnets question is 43:27.877 --> 43:30.537 interesting, because you could explore that if you 43:30.537 --> 43:33.357 change your magnet force, do you just shift that line 43:33.357 --> 43:35.737 around? Is it just that you need, and it could be 43:35.737 --> 43:39.157 you need a certain number of peaks getting above that 43:39.157 --> 43:41.077 max, but I think it definitely has something to do 43:41.077 --> 43:43.517 with that, of like how often are you getting above 43:43.517 --> 43:47.257 some critical force or critical torque to overcome your 43:47.257 --> 43:49.697 magnets and create movement. Yeah, question here. 43:53.057 --> 43:53.897 You're 44:35.937 --> 44:37.537 saying as we get smaller and smaller. 44:37.557 --> 44:41.057 But would you imagine if we scale everything 44:41.057 --> 44:44.797 together, so I would change the gears as well. So let's 44:44.797 --> 44:46.957 say I keep the same number of teeth, 12 teeth. 44:46.957 --> 44:49.717 So if you zoomed in and in and in, to you, like 44:49.717 --> 44:51.887 if I took a picture, it might look the exact same. 44:53.137 --> 44:56.347 assuming I can scale my motor which I can't but I 44:56.347 --> 44:59.337 could I'd probably have the same backlash in my teeth 44:59.857 --> 45:02.617 same alignment issues I think 45:05.237 --> 45:07.617 assuming I can make at that scale of course 45:07.617 --> 45:09.917 now if you say you have some resolution of 45:09.917 --> 45:12.317 your fabrication then absolutely as you get 45:12.317 --> 45:14.237 smaller you're going to get worse and worse 45:15.887 --> 45:17.357 but what you also mentioned about 45:17.357 --> 45:18.857 the teeth is interesting because 45:19.037 --> 45:21.137 we actually those teeth again There's a lot of 45:21.137 --> 45:23.497 things in here that I didn't have time to talk about, 45:23.497 --> 45:26.247 but the teeth are very particularly designed, 45:26.487 --> 45:29.837 the number of teeth and the shape of them, so 45:29.837 --> 45:33.217 that they grip well enough to transmit the torque. 45:33.237 --> 45:37.817 But then when you need them to be able to also break 45:37.817 --> 45:40.297 away, right? So when we're doing the fluctuation, then 45:40.297 --> 45:43.237 we want them to move past one another. So there's a, 45:43.237 --> 45:45.597 just like the magnets where we need, there's a band 45:45.597 --> 45:48.697 where magnets work. There's also a band in tooth design. 45:48.697 --> 45:51.557 For instance, if we had squared off our teeth, it did 45:51.557 --> 45:54.717 not work. Like we have like square teeth like this. 45:54.857 --> 45:56.717 It would just jam up. The whole thing would 45:56.717 --> 45:58.717 just jam up. So actually that triangle shape 45:58.717 --> 46:00.317 is important because it allows it to give 46:00.317 --> 46:02.597 pretty good force but also slip if it needs to. 46:02.807 --> 46:04.157 So you would have to keep all 46:04.157 --> 46:06.277 that stuff as you scale down. 46:06.297 --> 46:10.197 And then things like adhesion might become a problem. 46:10.197 --> 46:12.577 So at very small Skiles, things start sticking 46:12.577 --> 46:15.677 together and adhesion forces become significant compared 46:15.677 --> 46:19.117 to your other forces. And so there's that issue. 46:19.117 --> 46:20.957 Everything might just lock up because of adhesion. 46:20.957 --> 46:22.697 Yeah, so a lot of things to think about with scaling. 46:22.697 --> 46:23.217 Yep. 46:24.107 --> 46:26.637 One more? Was there one back there? Okay, here. 46:36.317 --> 46:43.017 I like that one better. Yeah, that's cool. Yeah, 46:43.017 --> 46:45.377 yeah, the distributed one. So one project we want 46:45.377 --> 46:48.517 to do is, Okay, so imagine they have some sensors. 46:48.517 --> 46:50.177 So right now, their only sensors are to sense the 46:50.177 --> 46:52.677 global thing. But imagine they could sense something 46:52.677 --> 46:54.977 about their state, maybe the force is acting 46:54.977 --> 46:57.417 on them. And so you have a bunch of them in here, 46:57.417 --> 47:00.437 and you apply some load, kind of like when we showed 47:00.437 --> 47:02.237 the ones with the weight on the top. And then 47:02.237 --> 47:07.397 every robot senses the force being applied to it. 47:07.397 --> 47:09.597 And so when we load that thing, and 47:09.597 --> 47:11.257 let's say we have 100 robots now, it's 47:11.257 --> 47:12.857 bigger, and we load it with a plate, 47:13.457 --> 47:15.837 of force chains running through that depending on the 47:15.837 --> 47:18.097 arrangement and there's probably a few kind of hot 47:18.097 --> 47:21.417 spots where you could you could see there's kind of 47:21.417 --> 47:24.257 high force concentrations and then there's going to 47:24.257 --> 47:26.497 be areas where there's very little force and what would 47:26.497 --> 47:30.397 be cool is if they could have some rule that like 47:30.397 --> 47:33.337 as the force gets higher they move more maybe and so 47:33.337 --> 47:36.157 then you can basically soften and flow in these high 47:36.157 --> 47:38.437 stress areas reduce the stress there and then 47:38.437 --> 47:40.997 eventually show you can get perfect stress distribution 47:40.997 --> 47:43.097 across the whole thing and load it to, I don't know, 47:43.097 --> 47:45.477 five times as much as you could have before. So yeah, 47:45.477 --> 47:47.837 but that would be with local rules. Like, be very 47:47.837 --> 47:49.237 cool if we don't have to talk to them at all. We just 47:49.237 --> 47:52.217 say, do your thing. And then they're each, you know, 47:52.417 --> 47:54.457 listening to their local 47:54.457 --> 47:56.597 environment and responding. Yeah. 48:01.567 --> 48:04.187 They need a goal. Yeah. So they'd be coded. They 48:04.187 --> 48:06.787 all have to be coded. But the question is, 48:06.787 --> 48:08.567 yeah, could you do it without them knowing anything 48:08.567 --> 48:10.387 about anyone else? I don't know. Maybe not. 48:10.437 --> 48:12.167 The easy way to do it would be to look at the 48:12.167 --> 48:14.217 stress chains and then tell, hey, you guys move. 48:15.187 --> 48:17.147 I think the more interesting question is, because 48:17.147 --> 48:18.687 when you scale these to millions, it's going to 48:18.687 --> 48:21.367 be hard to communicate to each one, right? They're 48:21.367 --> 48:23.227 probably going to need to know their own rules. 48:23.507 --> 48:25.267 Yeah, but no, cool question. There could 48:25.267 --> 48:27.667 be some combo, like, you know, cells. 48:28.027 --> 48:29.787 Okay, so I'll maybe step back a little bit, because 48:29.787 --> 48:32.867 cells have both going on. They have internal control 48:32.867 --> 48:36.427 loops going on, but they also, there are macroscopic 48:37.537 --> 48:39.967 gradients and chemicals and things going through the 48:39.967 --> 48:42.907 tissue to kind of coordinate motion too. So there could 48:42.907 --> 48:44.567 be some combination maybe where they're doing local 48:44.567 --> 48:48.107 stuff, but there's some general rules that are saying, 48:48.107 --> 48:50.407 hey, everyone, we want to kind of elongate this way. 48:50.447 --> 48:51.027 Yeah. 48:51.087 --> 48:53.407 So maybe a combination I'll say. How about that? 48:53.447 --> 48:53.927 Yeah. 48:55.697 --> 48:57.887 I think I'll speak to you one more time. 49:01.807 --> 49:02.647 Sorry. 49:05.107 --> 49:06.507 In two weeks. 49:17.127 --> 49:17.967 Why? 49:19.467 --> 49:25.237 Why did one need to build very specific brokers 49:25.237 --> 49:28.307 and all that stuff? Why not just simulate this? 49:28.307 --> 49:32.447 Ah, because simulations are wrong. Because every 49:32.447 --> 49:35.397 simulation is wrong. That's an interesting case. 49:38.087 --> 49:39.127 Yeah, 49:40.247 --> 49:42.927 yeah, yeah. For them. For themselves, they're right. 49:42.927 --> 49:45.307 They are, by definition. 49:45.307 --> 49:46.067 Yeah. 49:46.167 --> 49:47.127 Yeah. 49:56.547 --> 49:58.627 Yeah, I mean, this is your 49:58.627 --> 50:00.327 field, so what is the pick? 50:00.327 --> 50:01.087 Yeah, 50:01.247 --> 50:04.207 well, I think for one, the one is that, you 50:04.207 --> 50:05.647 know, the simulation, there's a lot of tuning 50:05.647 --> 50:07.827 in that simulation to get it to match reality, 50:08.727 --> 50:10.427 whereas reality happens. 50:10.647 --> 50:12.247 My reality, yeah, yeah, yeah. 50:13.827 --> 50:18.767 So is my reality representing all possibilities? 50:18.767 --> 50:20.627 Obviously not, right? I mean, like I said, we've 50:20.627 --> 50:23.407 hand-tuned many things in this to make it work. 50:27.727 --> 50:30.127 I think thinking about the, you know, their improved 50:30.127 --> 50:32.707 value. I think you were trying to correlate to 50:32.707 --> 50:37.107 that. You mathed on anything that would have been 50:37.107 --> 50:40.907 inconvenient or possibly were just a big simulation. 50:41.027 --> 50:43.227 I don't know if you found anything wrong. 50:43.247 --> 50:45.087 So that's just what there is. Yeah. 50:45.607 --> 50:48.147 Because you've thought about that. And what I'm, 50:48.147 --> 50:51.257 I mean, the answer is yes. He learned many things. 50:51.907 --> 50:54.687 I mean, I think a lot of the, OK, so the power 50:54.687 --> 50:55.927 stuff, for instance, that was something we 50:55.927 --> 50:57.927 didn't even think about testing. And then he 50:57.927 --> 51:00.647 just noticed on his power, he's like, whoa. 51:01.307 --> 51:04.547 But then maybe you could do that with simulations, 51:04.547 --> 51:07.367 too, if you're kind of curious, and if the simulation 51:07.367 --> 51:10.277 is set up in a way that you can really probe things. 51:12.127 --> 51:14.967 There's lots of readouts. Yeah, there's 51:14.967 --> 51:16.907 all kinds of readouts and things like that. 51:21.567 --> 51:24.187 But that's just a physics irrational when you think 51:24.187 --> 51:26.227 about this stuff. It's just that because you 51:26.227 --> 51:28.167 made these robots with these funny gears, but it's 51:28.167 --> 51:30.847 not dead. It feels like it's not an action or not. 51:30.847 --> 51:34.127 Unless you say, OK, I've decided I'm 51:34.127 --> 51:35.947 going to make them. I have 10,000 for a 51:35.947 --> 51:37.367 month, and I need them to do something. 51:38.287 --> 51:40.087 That's totally better. In the absence 51:40.087 --> 51:41.867 of that, you've just created .. 51:41.867 --> 51:44.267 Yeah. What's your point of view? That's fine. 51:44.267 --> 51:47.227 The question is, are there benefits to it? 51:47.447 --> 51:49.527 Because it's a lot harder than a simulation. 51:49.647 --> 51:51.427 That's your point, right? Yeah. 51:52.987 --> 51:54.007 Yeah, 51:56.967 --> 51:59.287 well, I would also say part of our 51:59.287 --> 52:03.447 goal is an actual functional system, 52:03.707 --> 52:06.417 right? This is a step towards this 52:06.417 --> 52:10.267 vision of creating a real robot material. 52:18.937 --> 52:21.267 I need a lot more than a million dollars. 52:32.447 --> 52:34.807 wait say that say it again so 52:38.567 --> 52:43.807 no no no yeah yeah no i wouldn't say this design is 52:43.807 --> 52:46.617 the design absolutely not but there's no way i'm 52:46.617 --> 52:48.977 going to get to the design without building a design. 52:49.337 --> 52:52.357 I can't simulate my way to the solution, no way. 52:52.797 --> 52:54.927 No. I mean, there's so many 52:54.927 --> 52:56.597 details that he worked out. 53:01.177 --> 53:04.197 Sure, but I would learn that by building the gears. 53:06.877 --> 53:09.037 Yeah, there's so many things. I think, Dan, 53:11.397 --> 53:13.357 no one even had made something that could 53:13.357 --> 53:15.757 support load and flow at the same time. 53:16.037 --> 53:18.597 We needed to take that step with 53:18.597 --> 53:20.657 a system. I don't think this is, 53:20.827 --> 53:23.017 I think we just, for instance, we discovered 53:23.017 --> 53:25.377 this idea that, or I don't know, discovered 53:25.377 --> 53:29.337 is the right word, but built a mechanism that 53:29.337 --> 53:32.597 allowed us to shape change while under load. 53:32.977 --> 53:36.117 The question is, okay, good. And the question is, 53:36.117 --> 53:40.077 does that mechanism, or that principle transfer to 53:40.077 --> 53:42.597 other things? I think, yeah, I think that one, yes. 53:42.597 --> 53:45.117 I think that one, yes, absolutely. You 53:45.117 --> 53:47.097 might not have to use gears or whatever. 53:47.097 --> 53:51.897 But I think there is a fundamental, to change 53:51.897 --> 53:57.617 topology in a tight-packed collective, you have to slide 53:57.617 --> 53:59.697 relative to your neighbor. You have to dial it. 53:59.697 --> 54:03.597 And so there's no way to do that by 54:03.597 --> 54:06.057 driving around the outside. And so you 54:06.057 --> 54:09.027 can't do these wheeled mechanisms. 54:09.377 --> 54:11.277 I can't actually. Well, in a 54:11.277 --> 54:12.557 very, very long time scale. 54:12.557 --> 54:13.737 No, if you have a million. 54:14.697 --> 54:15.537 No, 54:16.217 --> 54:18.617 I'm saying if you only have rolling, you can't do it. 54:19.817 --> 54:22.357 You can't transfer, right? If you're always 54:22.357 --> 54:24.057 bonded together, right? If you're always 54:24.057 --> 54:25.637 in your solid state, which you guys are. I 54:25.637 --> 54:27.637 remember first talking about five years ago. 54:27.637 --> 54:30.057 You're only rolling, you get started. 54:30.057 --> 54:31.417 You're always in the 54:31.837 --> 54:34.477 Which is the Jaeger thing, the granular bots. 54:34.477 --> 54:37.817 They roll, which are nice in a 10-unit thing, 54:37.817 --> 54:40.797 but there's no possibility of scaling that one. 54:42.717 --> 54:45.337 it really is the constraint we talked about this 54:45.337 --> 54:50.097 you have all your guys always bonded to each other 54:50.557 --> 54:52.777 so like in Geom's little guys 54:52.777 --> 54:55.717 they can create free volume 54:55.857 --> 54:59.317 they can vibrate back to each other you open up 54:59.317 --> 55:00.997 the free volume then you can move wherever you want 55:02.017 --> 55:03.737 but if there's no free volume you 55:03.737 --> 55:05.757 have to translate that's a thing 55:06.837 --> 55:08.997 yeah it's not about the gears 55:09.037 --> 55:12.817 it's making structures that are always bonded, 55:12.977 --> 55:16.717 always cohesive, yet can change shape. It can 55:16.717 --> 55:18.577 change topology. It can change topology is 55:18.577 --> 55:20.497 more than changing shape, right? It's not just 55:20.497 --> 55:22.217 stretching something. Oh, that's right. Yeah. 55:23.157 --> 55:23.737 Yeah. 55:24.057 --> 55:25.597 Doing it thin. 55:25.597 --> 55:28.137 It's the T1. I mean, as I basically said, that T1 55:28.137 --> 55:33.717 is kind of the building block of this, is if you 55:33.717 --> 55:35.937 can take four that are like this and then change 55:35.937 --> 55:39.277 the topology by translation, and then you can scale 55:39.277 --> 55:42.537 that. you can make many, many shapes. My point is 55:42.537 --> 55:44.977 that one day you'd be selling people, like you say, 55:44.977 --> 55:47.977 a sheet of this material, not like a bunch of 55:47.977 --> 55:50.717 little independent robot agents that come together. 55:50.837 --> 55:53.517 Your guys sort of are hard to pre-sell. 55:53.517 --> 55:55.677 Yeah, yeah. You can imagine 55:55.777 --> 55:59.057 in 3D it's like a beaker of this stuff, right? 55:59.057 --> 56:00.557 You take it to the space station and it 56:00.557 --> 56:02.977 makes any tool you want or whatever. It's like 56:02.977 --> 56:05.117 the ultimate Swiss army knife eventually. 56:05.357 --> 56:07.417 But to get there you needed to 56:07.417 --> 56:09.517 figure out cohesion and shape change. 56:10.177 --> 56:13.857 This is where I'm struggling now, I see that 56:13.857 --> 56:20.017 you are, your robots can, your robots and your 56:20.017 --> 56:22.337 material are one part of each other, right? 56:22.337 --> 56:25.437 Mm-hmm, mm-hmm, mm-hmm. Oil, mop, gas, robots, 56:25.897 --> 56:27.057 like our smart equipment, 56:27.057 --> 56:27.917 right? Yeah, they're useless. 56:28.017 --> 56:30.397 No, that's actually quite interesting. I didn't 56:30.397 --> 56:32.417 mention this, but it can't move. By itself, it can't 56:32.417 --> 56:34.877 move, which I think is good and bad. It can't do 56:34.877 --> 56:38.497 anything by itself. but it's also a pretty interesting 56:38.597 --> 56:39.917 it no longer 56:40.077 --> 56:42.097 it doesn't matter where it is anymore 56:42.097 --> 56:44.317 like all those other robots I show 56:44.317 --> 56:46.657 needed to crawl across the surface like 56:52.317 --> 56:55.007 yeah that's the same there they just flap right 56:55.007 --> 56:58.937 yep so it's similar in that way and I think that 56:58.937 --> 57:01.457 is the only way to do it if you want to scale 57:01.457 --> 57:04.017 to a material because many of the robots can't 57:04.017 --> 57:05.997 touch the ground anymore I mean if you want to get 57:05.997 --> 57:08.657 away from 2D, you can't have robots that crawl. 57:10.297 --> 57:11.957 Unless you're all crawling on each 57:11.957 --> 57:13.457 other, I guess you could imagine. 57:14.057 --> 57:15.777 You're crawling on each other. That's essentially what 57:15.777 --> 57:17.477 my robots are doing. They're crawling on each other. 57:17.537 --> 57:20.477 A lot of them 57:20.677 --> 57:22.377 come together, they form 57:22.797 --> 57:24.637 slugglers around a bunch of three bodies. 57:25.597 --> 57:28.737 But those guys are moving on one another. 57:28.737 --> 57:29.977 They don't need to walk in the 57:29.977 --> 57:31.357 ground. I guess that's all I was saying. 57:31.597 --> 57:33.907 But they don't need to. They can also... 57:35.347 --> 57:38.057 Yeah, they do both, yeah, you see. What I like about 57:38.057 --> 57:40.437 yours is that if you're playing the game that says, 57:40.437 --> 57:42.437 suppose the material comes pretty simple, right? 57:42.437 --> 57:46.277 I give you this jar of hawks or a bat or something. 57:46.997 --> 57:48.637 Hawks, hawk stuff. 57:49.237 --> 57:50.097 Then, 57:50.517 --> 57:51.237 right, 57:51.457 --> 57:53.617 you're sort of generating shape 57:53.617 --> 57:55.117 change without the volume. 57:55.577 --> 57:56.317 Yeah, yeah. 57:57.437 --> 57:58.297 Isovolumetric. 57:58.297 --> 57:58.957 Yes. 57:59.157 --> 58:01.737 A class of isovolumetric collective robots. 58:01.737 --> 58:02.397 Yeah, yeah. 58:04.037 --> 58:07.477 Which some do. I would say granular bots would fall 58:07.477 --> 58:10.797 in that, but it's because of the rolling constraint, 58:10.797 --> 58:13.137 it can only move particles around the outside. 58:13.597 --> 58:14.077 Right? 58:15.717 --> 58:18.197 That really is a different space than the 58:18.197 --> 58:20.357 granular bot thinks. So you can be more rigid 58:20.357 --> 58:22.617 than they can, and it also can't be as fluid. 58:22.697 --> 58:24.077 Can't be as fluid? 58:24.487 --> 58:26.177 Can't be as fluid, you said? 58:26.217 --> 58:28.277 It already goes to the gas thing. 58:28.277 --> 58:29.977 Yeah, but you can't, like, these 58:29.977 --> 58:32.057 bots can't operate, right? They can. 58:32.097 --> 58:33.617 they can't do anything once they've 58:33.617 --> 58:35.337 separated but they can separate 58:39.857 --> 58:41.857 actually there's some cool videos I didn't show 58:41.857 --> 58:45.077 where it loses parts and then they reconnect 58:45.537 --> 58:48.457 because then something will extend so you can break a 58:48.457 --> 58:50.437 part and then you extend and then you can condense 58:50.437 --> 58:53.497 again I don't think I showed it either where they elongate 58:53.497 --> 58:55.417 one way we can bring it back and elongate the other 58:55.417 --> 58:57.897 way too so you can do some of that but no, exactly 58:58.777 --> 59:01.417 they're not moving The center of mass 59:01.417 --> 59:03.737 always stays fixed is another way to think 59:03.737 --> 59:05.837 about it, right? They can't go anywhere. 59:06.157 --> 59:08.817 Those are, because it's a wall, it's probably 59:08.817 --> 59:10.437 a counter-changing period, right? Yeah.