WEBVTT - Mama Bots Having Baby Bots 0:00:00.160 --> 0:00:07.160 Brought to you by Toyota. Let's go places. Welcome to 0:00:07.360 --> 0:00:14.760 Forward Thinking. Hey everyone, and welcome to Forward Thinking, the 0:00:14.840 --> 0:00:17.919 podcast that looks at the future. It says, mama's don't 0:00:17.960 --> 0:00:20.840 let your babies grow up to be robots. I'm Jovin 0:00:20.840 --> 0:00:24.880 Strickland and I'm Joe McCormick. And have you two ever 0:00:24.960 --> 0:00:28.600 noticed how much better robots are than organic life forms 0:00:28.640 --> 0:00:32.559 all the time. That's why I only work with various 0:00:32.560 --> 0:00:35.200 things that allow me to interact with machines as opposed 0:00:35.200 --> 0:00:39.520 to actual human beings. Like if it's a prerecorded type 0:00:39.600 --> 0:00:43.879 message and I navigate through a menu using buttons, and 0:00:43.920 --> 0:00:46.159 I don't have to talk to a person, that's awesome. 0:00:46.440 --> 0:00:51.519 They're certainly less offensively smelly. Well that's just because we 0:00:51.560 --> 0:00:56.400 haven't mastered the artificial intelligence sector of creating really lifelike 0:00:56.440 --> 0:01:02.480 bo Yeah, well we have, we have scientists working on that. Yeah, yeah, no, 0:01:02.640 --> 0:01:05.640 I was kidding. But there are, of course situations where 0:01:05.680 --> 0:01:07.520 it's much better to be a robot than to be 0:01:07.560 --> 0:01:10.360 an organic life form, and we've talked about plenty of 0:01:10.360 --> 0:01:13.160 them on the podcast before. Maybe in space exploration where 0:01:13.160 --> 0:01:16.800 you're at risk, or in or in search and rescue 0:01:16.840 --> 0:01:19.400 scenarios where again a human would be at risk or 0:01:19.920 --> 0:01:22.400 you know, any number of hundreds of others. Really, Yeah, 0:01:22.640 --> 0:01:27.360 just just performing the same physical job. Anything that would 0:01:27.400 --> 0:01:31.240 make a human board or or repeated stress injury, that 0:01:31.280 --> 0:01:33.120 kind of stuff. I mean things that where you do 0:01:33.160 --> 0:01:36.000 it once, it doesn't hurt, but you do it, you know, 0:01:36.200 --> 0:01:39.560 once every minute for eight hours a day, and it's 0:01:39.840 --> 0:01:43.679 and eventually, yeah, the button pushing muscles start to wear 0:01:43.720 --> 0:01:46.759 down in the bones and etcetera. Yeah, robots don't get 0:01:47.120 --> 0:01:51.080 don't get physically injured, they don't get emotionally injured. It's 0:01:51.160 --> 0:01:55.600 real hard to traumatize a robot. Yeah, I know. Now, 0:01:55.760 --> 0:01:59.720 we we've talked about the advantages. Of course, the life 0:01:59.800 --> 0:02:04.240 forms have over robots and limited in each individual case. 0:02:04.320 --> 0:02:07.960 So a person like a worker is a lot easier 0:02:08.000 --> 0:02:10.680 to train for a new task than a robot is. Well, 0:02:10.760 --> 0:02:13.320 depending on if if the robot is mechanically suited to 0:02:13.360 --> 0:02:16.000 the new task, you can reprogram it s well. But 0:02:16.000 --> 0:02:19.359 but a worker can learn many new tasks. Yeah, hypothetically, 0:02:19.639 --> 0:02:24.760 most robots are pretty good at doing one thing right, right, Um, Yeah, robots, 0:02:24.919 --> 0:02:27.720 as it turns out, they pretty much can do what 0:02:27.880 --> 0:02:30.280 you built them to do, and they're pretty bad at 0:02:30.320 --> 0:02:34.760 doing anything outside of that. In general, right, Robots don't 0:02:34.800 --> 0:02:39.120 really evolve. Yeah, so yeah, so humans are more adaptable 0:02:39.120 --> 0:02:42.320 on the individual scale, but they're also more adaptable as 0:02:42.360 --> 0:02:45.440 a type of thing. Sure, over time, if the air 0:02:45.480 --> 0:02:49.280 pressures from the environment on a living organism, it will evolve, 0:02:49.320 --> 0:02:51.880 it will adapt, or it might go extinct. But assuming 0:02:51.919 --> 0:02:54.960 it doesn't go extinct, it will adapt and evolve to 0:02:54.960 --> 0:02:57.839 to fit the environment that it needs to make its 0:02:57.840 --> 0:03:01.840 living in. Robots don't do that, no, they remain. Yeah, 0:03:02.720 --> 0:03:04.800 so that's what we're going to talk about today exactly. 0:03:04.840 --> 0:03:07.600 So let's talk about like, like general robots right now. 0:03:07.639 --> 0:03:12.800 They they are too pretty distinct pathways for robot design 0:03:13.520 --> 0:03:15.200 in general, and of course there are a lot of 0:03:15.200 --> 0:03:18.440 branches within these, but we can go one path where 0:03:18.480 --> 0:03:21.400 we design a robot for a specific task and that's 0:03:21.440 --> 0:03:23.520 all it is supposed to do. So rumba is a 0:03:23.520 --> 0:03:26.079 great example. What a rumba is supposed to do is 0:03:26.080 --> 0:03:28.600 clean the floor. It's not supposed to do anything else 0:03:28.600 --> 0:03:33.280 other than it doesn't clean the cat box. In my opinion, 0:03:33.320 --> 0:03:37.240 this is this is probably like of robots and all 0:03:37.320 --> 0:03:40.920 of the good ones. Yeah, the most efficient and the 0:03:40.960 --> 0:03:44.400 most effective robots, I would argue, are of this type 0:03:44.400 --> 0:03:48.280 because we only have to concern ourselves with the design 0:03:48.360 --> 0:03:51.600 elements that will enable the robot to complete its task, 0:03:51.720 --> 0:03:54.360 and we eliminate everything else. Right, you don't need to 0:03:54.400 --> 0:03:58.560 include anything that doesn't involve cleaning the floor, or navigation 0:03:58.840 --> 0:04:02.080 or returning to like a docking station to recharge. When 0:04:02.120 --> 0:04:05.520 you're designing a robot like the Rumba, you just need 0:04:05.560 --> 0:04:10.240 those basic elements in it. Anything else is superfluous. Uh. 0:04:10.480 --> 0:04:13.040 Doesn't stop us from doing things like having crazy Rumba 0:04:13.080 --> 0:04:16.240 fights where we put a balloon and something sharp on 0:04:16.240 --> 0:04:18.320 on a couple of them and have them joused each other. 0:04:19.000 --> 0:04:20.800 But it doesn't mean we don't put our cats on 0:04:20.839 --> 0:04:24.360 top of them and just watch what happen. Maybe put 0:04:24.360 --> 0:04:26.640 our cat in a shark costume first and then put 0:04:26.640 --> 0:04:28.479 the cat on top of a roomba. Yeah, that sort 0:04:28.480 --> 0:04:31.600 of thing. But in general, they just do what they 0:04:31.640 --> 0:04:33.160 were supposed to do, and if you wanted them to 0:04:33.200 --> 0:04:35.400 do anything else, you're kind of out of luck. Or 0:04:35.560 --> 0:04:38.520 we tried to build general purpose robots that are capable 0:04:38.520 --> 0:04:44.400 of doing lots of different tasks hopefully yeah, to varying 0:04:44.480 --> 0:04:51.120 degrees of mediocrity because it's hard. It's not because you know, No, yeah, 0:04:51.160 --> 0:04:53.560 it's not because the people working on these aren't smart. Yes, 0:04:53.800 --> 0:04:56.600 because the job is a thousand times more difficult, right, 0:04:56.960 --> 0:05:00.080 because you have to anticipate lots of different things, a 0:05:00.160 --> 0:05:02.560 lot of different changing conditions. This would be the example 0:05:02.600 --> 0:05:05.880 of the darker robots we talked about previously, where UH 0:05:06.800 --> 0:05:09.520 challenge robots exactly. Yes, the DARPA Challenge robots, the ones 0:05:09.560 --> 0:05:13.520 that had to replicate a uh kind of a first 0:05:13.560 --> 0:05:17.440 responder situation, search and rescue kind of. Yeah. They to 0:05:17.480 --> 0:05:20.560 do like drive a car to a building, open a door, 0:05:20.720 --> 0:05:23.560 go inside, twist to lever, and cut a hole in 0:05:23.600 --> 0:05:26.240 the wall. They had to plug in a cable in 0:05:26.320 --> 0:05:30.320 one case. Yeah, they had to walk across rubble. They 0:05:30.360 --> 0:05:32.880 had to go up some stairs and then fall over. 0:05:33.600 --> 0:05:36.400 Uh didn't have to fall over, but a lot of 0:05:36.400 --> 0:05:39.680 them sure did. At any rate, we saw how difficult 0:05:39.800 --> 0:05:42.000 it is to design a robot that can do these things. 0:05:42.040 --> 0:05:44.840 And keep in mind, these were all tasks that the 0:05:44.920 --> 0:05:49.440 various teams knew about beforehand they were designing these. Yeah, yeah, 0:05:49.440 --> 0:05:51.279 I mean it was this was all stuff that they 0:05:51.360 --> 0:05:53.560 knew they were going to have, the robot was going 0:05:53.640 --> 0:05:56.440 to have to be able to do. Uh. So, building 0:05:56.440 --> 0:05:59.679 a robot that could end up anticipating all sorts of stuff, 0:06:00.000 --> 0:06:02.760 whether it's a single robot or a you know, a 0:06:02.960 --> 0:06:05.480 robot that can then design the next generation of robots. 0:06:06.360 --> 0:06:10.720 That's something that we really haven't perfected yet, but there 0:06:10.760 --> 0:06:14.320 are some really good reasons why we would want to. Clearly, 0:06:14.640 --> 0:06:17.400 being able to have a robot that could either adapt 0:06:17.440 --> 0:06:21.080 itself or adapt the next generation of robots would lead 0:06:21.120 --> 0:06:25.840 to much more efficient machines over time, and so this 0:06:25.920 --> 0:06:29.880 is something that we would really like to see in technology. 0:06:29.920 --> 0:06:33.080 It's not it's not just in robotics. We're also seeing 0:06:33.120 --> 0:06:37.120 it in computers. So there's they call it, you know, 0:06:37.160 --> 0:06:42.320 evolutionary computation or evolutionary computers or evolutionary robotics because that's 0:06:42.680 --> 0:06:45.839 essentially what we what we mean. We're talking about a 0:06:45.920 --> 0:06:50.160 machine that looks at the ability of another, like a 0:06:50.279 --> 0:06:53.960 next generation of machines to do a particular task, evaluates them, 0:06:54.279 --> 0:06:57.240 and then makes decisions on how to alter that generation 0:06:57.360 --> 0:07:01.360 to produce an even more effective generation after it. Uh 0:07:01.720 --> 0:07:05.080 the goal always being to come up with the most 0:07:05.240 --> 0:07:10.200 ideal design for whatever purpose you have in mind. And 0:07:09.920 --> 0:07:13.280 this isn't a super super new idea, no. Now, there 0:07:13.320 --> 0:07:14.880 have been people who have been working on this for 0:07:14.920 --> 0:07:17.760 a while. Back in two thousand eleven, researchers with NASA 0:07:17.840 --> 0:07:21.400 published a paper on evolutionary computation that was used to 0:07:21.440 --> 0:07:24.600 design new efficient antenna's way back in two thousand six. 0:07:24.960 --> 0:07:26.760 So the paper was published in two thousand and eleven, 0:07:26.760 --> 0:07:30.120 but the actual project was yeah in two thousand and six. 0:07:30.600 --> 0:07:33.160 And what they what they were pointing out was that 0:07:33.240 --> 0:07:38.239 designing an antenna is really challenging. It requires that the 0:07:38.280 --> 0:07:41.560 builder to have a very detailed knowledge of how antenna's work, 0:07:42.040 --> 0:07:46.080 which already limits your pool of various builders, and even 0:07:46.120 --> 0:07:52.120 then it's just painstaking to create an actual efficient, working antenna. 0:07:52.360 --> 0:07:56.240 So what they did was they created evolutionary algorithms for 0:07:56.280 --> 0:08:00.440 a computer to design an antenna for spacecraft auto matically 0:08:00.480 --> 0:08:05.840 that the computer system went through various designs of antenna 0:08:05.960 --> 0:08:08.960 and essentially simulated tests of them to see which one 0:08:09.000 --> 0:08:12.720 would be the most effective. They then took the designs 0:08:12.760 --> 0:08:15.280 that were predicted to be the highest performing antennas by 0:08:15.400 --> 0:08:18.960 this computer program, and they took it to the same 0:08:19.000 --> 0:08:23.960 manufacturer who was already building the antenna intended for the spacecraft. 0:08:24.280 --> 0:08:26.600 Then they tested all three of the antenna, the two 0:08:26.640 --> 0:08:29.960 that were the two best performer ones in the computer program, 0:08:30.000 --> 0:08:32.880 and the one that people had designed for the spacecraft, 0:08:33.080 --> 0:08:35.440 and found that the two that the computer had designed 0:08:35.760 --> 0:08:39.160 were more effective. Yeah, so it was one of those 0:08:39.200 --> 0:08:42.760 examples of this approach actually working better than what we 0:08:42.880 --> 0:08:45.840 humans could do. Now, maybe you don't know the answer 0:08:45.880 --> 0:08:49.120 to this, but if you do, I'd be interested. In 0:08:49.160 --> 0:08:52.120 this case, we're the what what we might call the 0:08:52.240 --> 0:08:56.720 mutations introduced into the models in in the algorithm that 0:08:56.720 --> 0:09:00.960 that tested out antenna's where the mutations direct did where 0:09:00.960 --> 0:09:04.360 they programmed in? Or were they random mutations? Like? Was 0:09:04.400 --> 0:09:08.360 it really more like evolution where a random things thing 0:09:08.480 --> 0:09:12.000 happens and then the system tests is this any better? Well, 0:09:12.080 --> 0:09:15.400 as we've talked about in previous episodes, random with machines 0:09:15.520 --> 0:09:18.800 is really hard to do. It was more like they 0:09:18.840 --> 0:09:24.360 were the computer was given an enormous number of variables 0:09:24.520 --> 0:09:28.439 and started to test them in various configurations without going 0:09:28.480 --> 0:09:32.800 through every possible one. Because the goal of evolutionary computer 0:09:32.920 --> 0:09:36.320 computation or evolutionary robotics is to make sure that you 0:09:36.440 --> 0:09:40.240 come to the most ideal form of whatever it is 0:09:40.280 --> 0:09:44.760 you're going for without having to test every possible variation. Because, 0:09:44.760 --> 0:09:48.840 of course, thus things get more complex, those variations increase 0:09:48.880 --> 0:09:51.079 in number until it would take you till the end 0:09:51.080 --> 0:09:53.640 of time to test all the different ones to come 0:09:53.720 --> 0:09:58.280 up version one one seven eight nine seven, seven seven 0:09:58.400 --> 0:10:02.040 four three is the best. You would take forever to 0:10:02.080 --> 0:10:04.560 go through all of those. So what these are designed 0:10:04.600 --> 0:10:07.760 to do is to test ones that are a best 0:10:07.800 --> 0:10:12.400 guests already of being effective, measuring those against other best guesses, 0:10:12.640 --> 0:10:15.360 then combining ones. Like if you were to find a 0:10:15.440 --> 0:10:18.960 generation that works particularly well and another generation that works 0:10:19.000 --> 0:10:21.560 particularly well, you might say, well, what happens if I 0:10:21.600 --> 0:10:25.240 combine the best design elements of both of these into 0:10:25.360 --> 0:10:29.000 a new design. Does that increase the effect of the 0:10:29.160 --> 0:10:32.920 efficiency and effectiveness or does it decrease it? Because it doesn't, 0:10:33.320 --> 0:10:36.760 you know, adding to awesome things together does not guarantee 0:10:36.800 --> 0:10:40.280 you to get a third, even more awesome thing. Oh yeah, yeah. 0:10:40.000 --> 0:10:42.400 I was wondering because I'd read about the use of 0:10:42.440 --> 0:10:46.800 evolutionary algorithms before in the design of new planetary rovers 0:10:46.920 --> 0:10:51.720 like unmanned planetary rovers, and testing different variations on the 0:10:51.760 --> 0:10:55.119 models in a in a computer simulation that would naturally 0:10:55.240 --> 0:10:59.199 select the highest performing ones. Yeah. Yeah, it's very similar 0:10:59.200 --> 0:11:02.480 to that. And uh, the thing about the difference, you know, 0:11:02.520 --> 0:11:05.240 we're going to talk about a robot that works along 0:11:05.280 --> 0:11:08.240 these principles. The biggest difference between a robot and a computer, 0:11:08.280 --> 0:11:11.760 obviously is that a robot is working with actual physical matter, 0:11:11.920 --> 0:11:17.800 not just simulations. And so there's some practical limitations that 0:11:17.880 --> 0:11:21.320 you encounter in that case. Right, you have to work 0:11:21.360 --> 0:11:25.720 with real physical matter that has weight, it has mass, 0:11:26.040 --> 0:11:29.800 it occupies space, there's a limit to how heavy it 0:11:29.840 --> 0:11:34.080 can be. Uh, you know, you can't just magically increase 0:11:34.200 --> 0:11:37.360 the size of whatever it is in a simulation and 0:11:37.400 --> 0:11:39.559 then just see how it works. You have to physically 0:11:39.600 --> 0:11:44.000 put this thing together. And so it's a really interesting approach. 0:11:44.080 --> 0:11:46.920 And we have seen a couple of examples of people 0:11:46.960 --> 0:11:51.400 working with true physical matter, but uh, it's mostly been 0:11:51.440 --> 0:11:54.280 in things like working with a program that builds stuff 0:11:54.280 --> 0:11:57.200 out of lego. So there is an actual example of this. 0:11:57.280 --> 0:12:01.240 There were some evolutionary computational experts that design systems that 0:12:01.520 --> 0:12:04.319 would allow a machine to build other objects out of legos, 0:12:04.400 --> 0:12:09.200 but you had to build limitations into the computer program 0:12:09.320 --> 0:12:12.280 so that the machine would follow the rules of physics. 0:12:12.280 --> 0:12:14.920 In other words, you couldn't have to lego pieces occupy 0:12:15.000 --> 0:12:17.960 the same physical space at the same time, right, Sure, 0:12:18.000 --> 0:12:21.400 And some of these computations aren't going to I mean, 0:12:21.559 --> 0:12:24.680 it's hard as a programmer to build all of those 0:12:24.720 --> 0:12:29.200 limitations into something that has to work in reality, because 0:12:29.200 --> 0:12:33.520 the computer isn't going to understand I mean basic stuff, 0:12:33.920 --> 0:12:35.840 and so so it might when you take it out 0:12:35.880 --> 0:12:37.600 of the lab, out of the out of the computer 0:12:37.640 --> 0:12:41.120 simulation stage, it might operate very differently than you were 0:12:41.120 --> 0:12:44.000 expecting it. Then the computer was expecting it to. An 0:12:44.000 --> 0:12:46.280 analogy I would make and has nothing to do with 0:12:46.320 --> 0:12:48.920 computers or robotics, but an analogy I would make is 0:12:49.440 --> 0:12:52.040 if you were running a role playing game. You're you're 0:12:52.080 --> 0:12:54.080 the game master of a role play game, and you 0:12:54.160 --> 0:12:57.400 think you have anticipated everything your players are going to do, 0:12:57.840 --> 0:13:00.680 and you have made up a masterpiece of a module, 0:13:01.080 --> 0:13:03.880 and your players are going to have an amazing time, 0:13:04.120 --> 0:13:07.200 and three minutes in the players decided to do something 0:13:07.280 --> 0:13:10.960 you could not have possibly anticipated. It makes perfect sense 0:13:11.000 --> 0:13:16.240 within the context of the game. And then you and you, yeah, 0:13:16.280 --> 0:13:18.120 you gotta throw away all the stuff you worked on 0:13:18.160 --> 0:13:21.440 and say, well, well, I guess like it's sort of 0:13:21.480 --> 0:13:23.679 like if you imagine a horror movie where the characters 0:13:23.679 --> 0:13:25.560 all walk up to the spooky house, take one look 0:13:25.559 --> 0:13:28.800 and say no, and then walk away like, well, there 0:13:28.800 --> 0:13:31.600 goes the horror movie. Same sort of thing. Well, we 0:13:31.640 --> 0:13:34.840 wanted to specifically talk about some researchers from the University 0:13:34.880 --> 0:13:39.400 of Cambridge who were working with a robotic system that 0:13:39.600 --> 0:13:44.880 used evolutionary robotics. So imagine, essentially, you've got a robot 0:13:45.120 --> 0:13:49.480 that builds other robots and tests them to see which 0:13:49.720 --> 0:13:54.960 designs work the best, and then either UH either has 0:13:55.000 --> 0:13:58.720 a design go forward or eliminates designs and starts tweaking 0:13:58.720 --> 0:14:02.000 things to try and find the best physical design of 0:14:02.000 --> 0:14:05.680 a robot to complete a specific task. Now that sounds 0:14:05.720 --> 0:14:09.760 super cool. I need to explain some stuff first so 0:14:09.800 --> 0:14:13.520 we can manage our expectations. Yeah, yeah, okay, so so 0:14:13.520 --> 0:14:16.400 so performance of these of these baby bots, as we 0:14:16.480 --> 0:14:20.040 might call them UH in this case was how fast 0:14:20.120 --> 0:14:24.080 and far it could move, like kind of scuttle across 0:14:24.120 --> 0:14:27.200 the surface exactly. So it's not like it was performing 0:14:27.280 --> 0:14:30.640 open heart surgery. No, No, they decided to hold that 0:14:30.800 --> 0:14:34.400 for the next experiment. Perhaps, Now these robots all we 0:14:34.400 --> 0:14:38.360 were supposed to do. We're create a locomotion power that 0:14:38.400 --> 0:14:42.560 could move it from one point to another point, and 0:14:42.840 --> 0:14:46.800 uh they would. The experiment measured how long it took 0:14:46.960 --> 0:14:52.200 these little devices to move across this particular expanse they 0:14:52.280 --> 0:14:55.960 changed up. That didn't make the cut. They didn't they 0:14:56.000 --> 0:14:59.520 didn't continue on. Uh that they were taken apart. They 0:14:59.520 --> 0:15:02.400 were taking up bodily by the mamabot. Actually, to be fair, 0:15:02.440 --> 0:15:04.320 I think all of them were taken apart, but yes, 0:15:04.920 --> 0:15:09.280 uh uh yeah they were. They'd be scuttled, they'd be 0:15:09.320 --> 0:15:14.120 scuttled and then they'd be scuttled and and and harvested 0:15:14.160 --> 0:15:18.560 for their organs essentially. So that was the that was 0:15:18.600 --> 0:15:22.560 the criteria that the robot used. Um and here's how 0:15:22.600 --> 0:15:25.200 it gets here. Let's break this down because that's that's 0:15:25.200 --> 0:15:28.200 like an overview of what happened, but it's really interesting 0:15:28.200 --> 0:15:32.200 when you get to the specifics. So what they wanted 0:15:32.240 --> 0:15:34.080 to do is they wanted to design a system that 0:15:34.120 --> 0:15:39.920 could design, test, and change robotic designs relatively rapidly. And 0:15:40.000 --> 0:15:44.120 it needed to be a a system that could work 0:15:44.160 --> 0:15:46.440 within the lab, but wasn't designed to be a practical 0:15:46.480 --> 0:15:50.000 system that could go through every single possible arrangement of 0:15:50.040 --> 0:15:55.640 the various pieces. For yeah, um, the pieces are essentially 0:15:55.680 --> 0:15:57.280 two different types of cubes that will get to in 0:15:57.280 --> 0:16:00.160 a second. So they wanted their approach to contain this 0:16:00.200 --> 0:16:03.160 is a quote from their paper paper, a limited number 0:16:03.240 --> 0:16:07.440 of evolutionary iterations, and that was, you know, the the 0:16:07.520 --> 0:16:11.520 idea of let's let evolution lead us to what should 0:16:11.600 --> 0:16:16.000 be the best version of these robotic designs, as opposed 0:16:16.040 --> 0:16:18.920 to just doing trial and error, measuring everything and then 0:16:18.960 --> 0:16:22.760 going with the top performer. Now, they did say in 0:16:22.800 --> 0:16:26.360 their paper that while this is really interesting and could 0:16:26.440 --> 0:16:29.520 in fact be a breakthrough in science and technology, you 0:16:29.600 --> 0:16:33.040 have to admit that there are limitations to this technology 0:16:33.160 --> 0:16:37.440 cannot necessarily be applied to mass production. Mass production relies 0:16:37.480 --> 0:16:42.520 heavily on automation. Automation is different from mass customization. So 0:16:43.520 --> 0:16:47.520 they said, you know, there there are certain practical applications 0:16:47.560 --> 0:16:50.080 for this kind of approach, but it's not gonna be 0:16:50.160 --> 0:16:52.920 like we're gonna design these robots and they are going 0:16:52.960 --> 0:16:58.160 to magically make all of our factories work at efficiency. Uh. 0:16:58.240 --> 0:17:01.640 It's uh. They said that there's a challenge to developing 0:17:01.640 --> 0:17:05.760 automas design of quote a large morphological diversity end quote 0:17:06.400 --> 0:17:11.200 morphological obviously meaning that the actual form of these things. Yeah. 0:17:11.400 --> 0:17:13.480 So the basic design of the experiment was to use 0:17:13.520 --> 0:17:16.280 a robotic arm that was the Mama butt, so it's 0:17:16.280 --> 0:17:18.600 a robotic arm, and had to gripper fingers, so just 0:17:18.760 --> 0:17:22.080 very simple robot um. One of the gripper fingers actually 0:17:22.160 --> 0:17:27.399 had a a little nozzle through which it could squirt glue, 0:17:28.400 --> 0:17:32.040 so they could glue pieces together. Uh, and it was 0:17:32.800 --> 0:17:35.240 connected to a computer that was running the evolutionary software 0:17:35.240 --> 0:17:37.800 to build the robots out of these little cube modules. 0:17:38.080 --> 0:17:40.919 And they had two types of modules. They had passive modules, 0:17:41.320 --> 0:17:44.000 which really just small wooden blocks that were painted black. 0:17:44.200 --> 0:17:46.360 The reason they were painted black was so that the 0:17:46.440 --> 0:17:49.480 cameras that the robot was using essentially as eyes, could 0:17:49.480 --> 0:17:52.680 easily pick them up. Let's see where the little passive 0:17:52.680 --> 0:17:55.199 ones were. So these couldn't do anything, right, they were 0:17:55.200 --> 0:17:58.560 just connectors. And then you had the active modules. These 0:17:58.600 --> 0:18:01.879 were a larger cube and one face of the cube 0:18:02.160 --> 0:18:04.919 was attached to a small motor that could rotate at 0:18:04.920 --> 0:18:07.800 a certain amplitude in frequency, so it could rotate the 0:18:07.800 --> 0:18:10.200 face of the cube, which is kind of the basis 0:18:10.280 --> 0:18:13.359 of the locomotion of the piece. Right, So, so imagine 0:18:13.400 --> 0:18:16.520 a Rubik's cube where only one side can rotate around, 0:18:16.800 --> 0:18:19.040 but it does so on a motor. That's essentially what 0:18:19.080 --> 0:18:23.479 we're talking about. And so different cubes would have faces 0:18:23.520 --> 0:18:26.840 that rotated at different speeds, and the all of the 0:18:26.880 --> 0:18:29.760 information was contained within the cube so that the robot 0:18:29.800 --> 0:18:31.879 would be aware of that. It wasn't like the robot 0:18:31.880 --> 0:18:34.960 was randomly picking blocks and some could turn at a 0:18:35.000 --> 0:18:37.240 certain speed and others would turn at half speed or 0:18:37.240 --> 0:18:40.400 twice speed or whatever. The robot was actually quote unquote 0:18:40.400 --> 0:18:43.600 aware of which cubes could do what and uh and 0:18:43.640 --> 0:18:47.760 by by fitting them together in different combinations, uh yeah, 0:18:48.000 --> 0:18:49.960 you could you could get the box to kind of 0:18:50.080 --> 0:18:52.399 kind of wiggle, to kind of hop and wiggle across 0:18:52.400 --> 0:18:56.600 the table. It reminds me um of the motion of 0:18:56.680 --> 0:19:01.080 some of the the less mobile toys and toy story 0:19:01.200 --> 0:19:03.280 um kind of just sort of like scuttling along it 0:19:03.359 --> 0:19:06.160 ones that didn't have legs. Yeah. Yeah, it's flipping adorable, 0:19:06.200 --> 0:19:09.000 you guys. I am going to see if we can 0:19:09.040 --> 0:19:11.720 find video that's share able to share with you because 0:19:11.760 --> 0:19:13.879 it is so cute. It makes me think of a 0:19:13.920 --> 0:19:17.480 lot of like, you know, wind up toys that that 0:19:18.119 --> 0:19:22.080 move because they're doing this repetitive motion. Uh, and that's 0:19:22.320 --> 0:19:26.320 enough to propel them across a surface. Uh, not with anything. 0:19:26.480 --> 0:19:29.120 They can't necessarily steer or anything, but they can move 0:19:29.440 --> 0:19:31.040 the same sort of thing. Like the idea was that, 0:19:31.080 --> 0:19:34.720 all right, well, if we pair this passive module with 0:19:34.760 --> 0:19:38.920 these two active modules in this configuration, will that create 0:19:39.040 --> 0:19:42.480 movement that's faster and more effective than this other design. 0:19:43.080 --> 0:19:45.800 So the computer they used was a regular desktop PC, 0:19:46.080 --> 0:19:48.919 nothing particularly special about it, but it was running a 0:19:48.960 --> 0:19:52.679 controller program that was using the matt Lab language m 0:19:52.720 --> 0:19:55.600 A T L A B and that's used for interactive 0:19:55.640 --> 0:19:59.159 environments of developing algorithms, among lots of other stuff. The 0:19:59.240 --> 0:20:02.280 robotic arm could easily grip and rotate any module it shows, 0:20:02.440 --> 0:20:04.520 and could stick two of them together using what they 0:20:04.560 --> 0:20:09.040 called a hot melt adhesive or h m A. Yeah, 0:20:09.160 --> 0:20:12.479 essentially like hot glue, like glue gun glue is more 0:20:12.560 --> 0:20:15.200 or less what that was, And it could build robots 0:20:15.240 --> 0:20:19.560 by combining multiple modules together. Each module had what they 0:20:19.600 --> 0:20:23.359 called a gene, which essentially described the type of module 0:20:23.400 --> 0:20:26.800 it was and the motor control of that module, and 0:20:26.840 --> 0:20:31.320 also included the construction parameters. Of the module construction parameters. Essentially, 0:20:31.359 --> 0:20:34.000 it was things like basic rules for the robot to 0:20:34.119 --> 0:20:37.159 understand so that it could effectively build a robot. In 0:20:37.160 --> 0:20:39.840 other words, if you want to build a robot, you're 0:20:39.840 --> 0:20:42.560 probably gonna need to put a larger piece down first 0:20:42.640 --> 0:20:44.720 before you put a smaller piece on top. It was 0:20:44.760 --> 0:20:47.879 building them vertically, so it needed to know, Hey, if 0:20:47.920 --> 0:20:50.040 you try and build a robot this other way, this 0:20:50.119 --> 0:20:52.560 thing's gonna fall over. So you need to be able 0:20:52.600 --> 0:20:55.439 to build them in you know, these are the general 0:20:55.520 --> 0:20:57.720 rules you need to follow, essentially the rules of physics, 0:20:58.160 --> 0:21:00.840 so that your robot will be built the way you 0:21:00.880 --> 0:21:04.720 intend to build it. Uh. So a robot was made 0:21:04.760 --> 0:21:06.480 up of these modules, and it was said to have 0:21:06.520 --> 0:21:10.080 a genome. It was the collection of these genes and 0:21:10.480 --> 0:21:13.440 uh these genes would either work or not work together. 0:21:14.080 --> 0:21:19.000 And robots could have between one to five genes modules. Yeah, 0:21:19.160 --> 0:21:22.399 now they if it's having five. The max was three 0:21:22.400 --> 0:21:25.360 active too passive. You could not have more than three 0:21:25.359 --> 0:21:28.320 active simply because it would make the robot too heavy 0:21:28.359 --> 0:21:31.040 for the gripp or to grip. Right. Yeah, that's that's 0:21:31.040 --> 0:21:33.719 one of those limiting factors in reality there. Yeah, so 0:21:33.800 --> 0:21:36.439 if it were more than fifty grams of mass. The 0:21:36.520 --> 0:21:39.000 robotic arm, the gripper just couldn't maintain a grip. It 0:21:39.040 --> 0:21:42.680 would drop it and it would break. And so practically 0:21:42.920 --> 0:21:45.800 you could have a maximum of three active and two 0:21:45.800 --> 0:21:50.159 passive pieces connected altogether. They didn't necessarily have to have 0:21:50.200 --> 0:21:52.760 that many. Some of them were only three or two 0:21:52.960 --> 0:21:58.440 uh two blocks large at large two blocks in total, 0:21:59.080 --> 0:22:02.479 but at any rate, the construction parameters were these very 0:22:02.520 --> 0:22:06.600 simple directions and it allowed a lot of flexibility. So 0:22:07.440 --> 0:22:10.560 the robot arm essentially could decide which modules to use 0:22:10.680 --> 0:22:13.760 and how to connect the two together or three or 0:22:13.920 --> 0:22:18.359 four however many together, and then test it so that 0:22:18.640 --> 0:22:20.960 the what would happen after it builds one of these, 0:22:21.000 --> 0:22:24.760 after it squirts the glue and everything. Um, you could 0:22:24.760 --> 0:22:28.359 actually send it a genome essentially a recipe saying here 0:22:28.400 --> 0:22:30.439 are the modules I want you to use and the 0:22:30.480 --> 0:22:32.320 configuration I want you to put them in, and this 0:22:32.359 --> 0:22:34.600 is going to represent the first generation. You could do that, 0:22:35.240 --> 0:22:39.640 or you could make it uh more of a random approach. Well. 0:22:39.680 --> 0:22:43.080 Once built, the mother robot would lift the finish modular robot, 0:22:43.440 --> 0:22:45.879 move it to the testing area, which is just a 0:22:45.880 --> 0:22:49.520 flat surface for the robot to crawl across, also known 0:22:49.560 --> 0:22:54.320 as the arena. In some cases it was a hard surface. 0:22:54.400 --> 0:22:57.000 In other cases they used a carpeted service, and at 0:22:57.080 --> 0:22:59.280 least in one they used a foam surface. Yeah. They 0:22:59.280 --> 0:23:02.640 they slowly moved towards the foam surface as they realized 0:23:02.680 --> 0:23:07.399 throughout the experiment that the robots were having undo trouble. 0:23:07.840 --> 0:23:09.880 I was sad to learn that the foam they were 0:23:09.880 --> 0:23:12.000 talking about was like, you know, kind of like mattress 0:23:12.040 --> 0:23:15.359