Flight School for Robots with Alejandro Stefan-Zavala

Who speaks in this episode?

• Alejandro Stefan Zavala (Guest)
• Lori Dah Jose (Host)

What is discussed, and when

1. 0:00 – 5:00: Caltech, Ingenuity, Alejandro Stefan Zavala, CAST
2. 5:00 – 10:00: Caltech, Ingenuity, JPL, El Sal
3. 10:00 – 15:00: Maraca
4. 15:00 – 20:00: Caltech, Base 11, Santa Monica College
5. 20:00 – 25:00: Caltech, Base 11
6. 25:00 – 30:00: Ingenuity
7. 30:00 – 35:00: Caltech, Ingenuity
8. 45:00 – 50:00: Ingenuity
9. 50:00 – 55:00: Ingenuity, Caltech, Perseverance

Notable quotes from "Flight School for Robots with Alejandro Stefan-Zavala"

• First year of college here at Caltech. I got my degree in planetary science and now I work here where I write stories for the Caltech magazine.

  about Lori Dah Jose · at 1:00

• Up in Venezuela and dreamed about building robots since he was a little kid. Now, he's a graduate student in Caltech's Center for Autonomous Systems and Technologies, also known as CAST.

  about Alejandro Stefan Zavala · at 2:03

• Okay, cool. So, for those of you who don't recognize it, that one's called Ingenuity, and it's also known as the Mars helicopter. And here's a real picture. Well, actually, I think that was CGI, but here's a picture of what it.

  about Ingenuity · at 3:57

• Right? Yeah, I know. I know. we actually have the tallest waterfall in the world. It's called El Sal. And that's where that's where up, that's where they were trying to go.

  about El Sal · at 9:29

• Enough about my upbringing. Now, let's make it back to California. well, or I guess to California. So, I started off my undergrad on a student visa in Santa Monica College. I…

  about Caltech · at 19:05

• Is we actually just build skinny little robots on Earth. So this for example is it's called the scarecrow. So what JPL does in order to test that in order to sort of make that…

  about JPL · at 26:30

• Wind at it. And I actually don't know how well JPL understood the wind back then. I do know though that Perseverance, the rover that went with Ingenuity, has dedicated like…

  about Perseverance · at 53:59

Who and what is mentioned in this episode?

People

• Alejandro Stefan Zavala — Person · Main speaker (featured engineer/guest) · 1 mention · first at 1:57
  Delivers the featured talk on robotics/autonomy, describing his background (Venezuela → community college → UC Berkeley → Caltech) and his research on predicting wind fields and planning safe autonomous drone flight using fan-array wind tunnels and ML models.
• Lori Dah Jose — Person · Host · 1 mention · first at 0:53
  Introduces the Science Journeys program and the featured speaker; provides her own Caltech background and frames the event’s goal of showing diverse STEM pathways.

Companies & orgs

• Caltech — Company · Headquarters · 10 mentions · first at 1:00
  Primary location where the speaker works and conducts research
• CAST — Company · Business focus · 1 mention · first at 2:16
  Center for Autonomous Systems and Technologies where the speaker conducts research
  Up in Venezuela and dreamed about building robots since he was a little kid. Now, he's a graduate student in Caltech's Center for Autonomous Systems and Technologies, also known as CAST.
• JPL — Company · Referenced expert · 1 mention · first at 6:31
  Jet Propulsion Laboratory involved in Mars missions
• Base 11 — Company · Case study · 3 mentions · first at 19:38
  Organization that helped the speaker transition from community college to Caltech
  Organizations like this. They're not alone, but they're the ones that did it for me. So, what Base 11 does is they create opportunities for students of underrepresented groups…
• Santa Monica College — Company · Setting · 1 mention · first at 19:15
  Speaker's initial college in the US
  Enough about my upbringing. Now, let's make it back to California. well, or I guess to California. So, I started off my undergrad on a student visa in Santa Monica College. I…

Products & tech

• Ingenuity — Product · Case study · 12 mentions · first at 4:01
  Mars helicopter used as an example of autonomous flight technology
• Perseverance — Product · Case study · 1 mention · first at 54:05
  Mars rover mentioned in passing

Places & events

• El Sal — Place · Setting · 1 mention · first at 9:33
  Tallest waterfall in the world located in Venezuela
• Maraca — Place · Setting · 1 mention · first at 10:09
  Speaker's birthplace in Venezuela
  Okay. So, within Venezuela, I was born in Maraka, which is a city. It's the second largest city. It's on the west coast of Venezuela, and it looks like this. this is a picture I took.

Full transcript with timestamps

0:10 Heat. Heat. My name is Lori Dah Jose and I will be your host today. So 14 years ago I was sitting right where you are starting my first year of college here at Caltech. I got my degree in planetary science and now I work here where I write stories for the Caltech magazine. We're so happy that you've joined us today for Science Journeys which is the latest iteration of a program that has had many names and is now in its 21st year. So our goal with science journeys is to show you how science and engineering impact all of our lives and also to show you how you might become a scientist or engineer one day yourself.

1:31 We want you to take away that there is no one single correct path to get into STEM. All of our speakers come from many different cultures and places and you get to hear about their challenges and experiences along their journey. All right, so raise your hand if you think you might want to be a scientist or engineer someday. Okay. Yeah, that's a lot of hands. Well, today you're going to get a peek into the life of a real engineer who works with robots. Our speaker today is Alejandro Stefan Zavala. Alejandro grew up in Venezuela and dreamed about building robots since he was a little kid. Now, he's a graduate student in Caltech's Center for Autonomous Systems and Technologies, also known as CAST.

2:16 Cast is a world leader in robotics and AI systems among many other things. And today, Alejandro is going to tell you all about his project to teach drones how to fly all by themselves with no human guidance. So, what's going to happen is we're first going to see a short animation about Alejandro's journey and then he's going to give a 40minute presentation and then afterward you'll all have time to come up and ask him questions during our time for Q&A. Does that all sound good? Yeah. All right. Well, I hope you enjoy our science journeys presentation today.

3:21 Hello. Hi everyone. Oh, that's a lot of people. Thank you all so much for coming. This means a lot to me. Um, why don't we just go ahead and get started? So, uh, welcome to my talk. I'm gonna start with a show of hands. So, does anybody in the audience recognize this? Does anybody know what that is?

3:44 Nice. Yes. Okay. Does anybody here have one? Yes. Oh. Oh, I think he's saying he wants one. Oh, sorry. Now you got to buy it. Okay. Next thing. Does anybody recognize that little guy over there?

3:57 Okay, cool. So, for those of you who don't recognize it, that one's called Ingenuity, and it's also known as the Mars helicopter. And here's a real picture. Well, actually, I think that was CGI, but here's a picture of what it actually looks like. Uh, in 2020, JPL put an actual, honest to God helicopter on Mars and flew it. It looked like this. This is the real footage from the first powered flight on another planet. Now, I don't know about you, but I see that and I think, literally, how did they do that? Um, well, actually, first of all, you know, we're on Mars. Um, well, at some point you have to be on Earth, right? Uh, and that's where it starts. So, before flying on Mars, you have to learn how to fly on Earth. And this is one of the early videos of a prototype of Ingenuity sort of learning to fly, right? I love this video. It kind of reminds me of like a baby deer, you know, sort of stumbling and trying

4:49 to learn how to walk when it's just been born. Uh, you're going to see it. It's going to do a little loop. Okay, so it it jumps up and then it bounces around and now it's going to go back and oh, and it crashes. It lost a blade. You know, it happens. But there's a lot more to it than that. This this video doesn't tell the whole story. So that video that that what you're seeing there is actually happening on the bottom of a giant chamber, a vacuumsealed chamber that simulates the Martian atmosphere. And actually on that chamber at one point it wasn't just like an empty space where you know Ingenuity was bouncing around and crashing into the walls and stuff. There was actually this whole contraption where over there that thing on the like upside down is actually one of the prototypes of Ingenuity, one of the more mature prototypes. And that big black wall is a giant grid of 441 computer fans. They kind of look like

5:41 this. Uh but like really beefy computer fans, but it's just computer fans. Uh that grid we actually call a fan array wind tunnel. and the software to control it. This is actually a screenshot of the early an early prototype of the software used to control this giant grid of fans, how to coordinate all of them. And uh if you look at the words on the left, there's all these weird names. Uh so actually uh Mckon is actually in honor of Beverly McKon, a former faculty here at Galsid at the aerospace labs at Caltech. Uh Mohar is a friend of mine.

6:15 America is just literally as in like America. It's just a funny word. Cass is a video game character. Actually, the software that controlled the wind tunnel that replicated Martian winds in the first powered flight in another planet was crammed with Easter eggs. We had things like JPL calling in because Charizard was malfunctioning. And that was entirely my fault I did that. That was that was me. That was my doing. I also made it so that you could draw on the wind. So in this picture, the blue squares represent areas where the fans are blowing slowly and the reduh the red rect u cells represent areas where the fans are blowing quickly. And that is a drawing of my lovely mentor Chris who had the misfortune of being my supervisor when I was in charge when I was you know doing this in the mission.

7:01 And for some reason he didn't like me doodling him as a pudgy dough boy on the wind. Um so how did I get there? How how did I end up cramming a Mars mission with Easter eggs in the, you know, Martian wind replication machine? Well, first off, now that I have your attention, that's you looking at me now because I showed you all this cool stuff. Before I give you the answer to all those questions and the other questions you have, such as how do you actually fly on Mars? I'm going to take a step back. So, I'm going to take a step back and I'm going to talk a little bit about what this talk is about. So, it's called Fly School for Robots, like I said. So, there's going to be robots.

7:40 Um, there's flight. So, we're going to be flying. Specifically, we're going to be talking about drones. And it's a school, so we're going to teach them. Specifically, we're actually going to use a bunch of computer fans. Unsurprisingly, I already told you about them. And we're also going to use a bunch of computer simulations. But this is just the the content of the talk in the sense that this is what I'm going to be showing you. But actually I want to emphasize the degree to which this talk is about you. See how I timed that? And what I mean by that is I I'm I'm dedicating this talk. I I want to make a point for the students in the audience.

8:19 Whether you're in eighth grade, whether you're in high school, whether you're in college, I want to say this explicitly, which is that as you see this talk, I'm going to just show you I'm going to show you a bunch of cool stuff and I'm going to try to make you laugh for 40 minutes. Everything that you see that you think is cool, I want you to see it with the entitlement of knowing that it's within your reach. And I'm saying that because when I was in your position and I looked at students who were in my position, I thought that the students in my position were just built different. And that's not true. I'm not special. And I'm going to show you I'm going to also along the you know along the way I'm going to give you some hints on how I made it to this you know to this position that I'm so privileged and thankful for that you can also use for yourself to sort of give you some ideas to guide you. Uh okay so

9:04 now that we've established that this this talk is entirely about you. Let's talk about me. So I was born in Venezuela which is over there. Uh and uh it's actually featured in Up. The movie Up was actually about flying to Venezuela. And you would think from a Disney Pixar movie they would sort of exaggerate Venezuela. They actually understated it. The place that they flew to actually looks like this, right? Yeah, I know. I I I know. Uh we actually have the tallest waterfall in the world. It's called El Sal. And that's where that's where up, you know, that's where they were trying to go.

9:37 They're trying to put the little house next to the waterfall. Uh we also have what I think are the most beautiful beaches in the world in Los San Rock. Uh and we also have the highest inflation in the world. Uh this is supposed to be a graph comparing inflations. Um you can't see the other countries. Here in the US people get worried about like 6% inflation and like you know fair enough. Uh we had 1.4 million% inflation whatever that means. So yeah Venezuela number one dude. Uh, okay. So, within Venezuela, I was born in Maraka, which is a city. It's the second largest city. It's on the west coast of Venezuela, and it looks like this. Uh, this is a picture I took from a really tall tree that I climbed on top of. Um, you can see there's a bunch of trees. What I want to highlight from that, one of my favorite features, half of them are mango trees. There's like infinite mangoes all the time.

10:30 I know. No, it's it's awesome. You just get free mangoes everywhere. Uh, next thing is, okay, imagine squirrels. Just think about the dynamic that you have with squirrels. We have that but iguanas like like like straight up the same thing but iguanas. I had a pet one. Her name was Bowser. And uh what else? Uh oh yes, we are the lightning capital of the world according to NASA and it's not an exaggeration. So we have this we have this uh phenomenon. It's called Elkatumbbo which I think is a that's an awesome name for lightning. Uh there's actually a part of our coast that is literally always lightning. It was called the natural lighthouse because uh sailors could use it to find the coast at night when there was no lighthouse. Uh this is a picture I took when last time I visited in 2018.

11:20 Uh okay. Now from my childhood, those are my little brothers and me on top. So growing up, this is science journey. You guys are gonna you guys are going to see my whole journey. So growing up, I was good for three things. Climbing things, drawing things, and building things. and I'm going to tell you about them. So, climbing things, specifically things that shouldn't be climbed. Um, here's a video of me jumping off a roof to escape from campus security in high school. Here's me climbing the PE gym and like summoning it. You can It's three stories tall. You can sort of see the city skyline. Uh, here's me climbing the actual high school building from the outside, which was also three stories tall.

11:58 Uh, here's me climbing stadiums with my friend. So, uh, that's my buddy Ector. You're going to see him. We we would we would climb on the st like on the outside of the stadiums and then parkour on top of all the cool shapes. And the one that you saw that sort of has a slope. We would we would climb on top of it and then we would slide down the slope and we would like jump off to the to the parking lot and do like a thick parkour roll. It was awesome. Uh, and then here are some here are some pictures. Uh, here's me sort of enjoying the view from the top of that stadium. Uh, here's me vaulting on off of, you know, a little obstacle. Here's me aura farming.

12:33 So, this is what was going on in my head the entire time. Now, for you though, the takeaway from all this, literally do not do this, please. For the parents in the room, the kids, don't do this. Go to a parkour gym instead. Uh so here in the United States uh you actually have gyms are dedicated to this with padded surfaces, people who can teach you how to do everything, legal liability, you know, all kinds of awesome features. So you know, just go to a parkour gym if you think this is cool, which I think I would just agree with you. Okay, let's talk about drawing. Uh I've been drawing my entire life when I was one of my earliest memories, it might be the earliest memory, uh I was in preschool and I was told that I'm such a good artist, I should be an architect. And I just got really angry at the teacher and I looked at her and I said, "No, I'm going to be a scientist." Well, and here I am. So,

13:26 there you go. Uh, but no, I've been drawing my entire life. And, uh, actually, I use it to think. Uh, this these are pictures from my journals throughout my life. Um, I want to highlight this one actually because I I draw so much and basically my entire process of thinking consists of just drawing situations and conversations and stuff and like I mean I don't know I guess anime fights or something but you know and like conversations and whatever and they sort of the drawing starts seeping into my dreams. It's really cool. But this one is really meaningful to me cuz um I'm terrified of public speaking which is kind of inconvenient right now. And uh this one was the first conference talk I ever gave. I was so terrified I didn't sleep the whole night. Uh and I went there. I did a great job by the way. Uh but then right after I went to the hotel room, I fell asleep and I had this dream that I was this like Iron Man cyborg

14:19 fighting like Attack on Titan giants with lasers that shot off my hands. It was really cool. So, I doodle it and now it always reminds me of, you know, facing my fears. And the reason I'm showing you this uh is because I have another thing I want to sort of tell especially the kids. Um, and it's do art. I I I I like I take it so personally that I went through the trouble of handwriting it instead of typing it in the slides. Please, for the love of God, do some form of art. It will for for those of you who are about to who are going through school, who are going to go through college, art in any way, whether it's drawing, whether it's music, whether it's building things or making ugly sculptures out of soap, I just it will regulate your emotions. I made it through college thanks to my friends and my doodles. And I want you to know that. Okay, let's talk about building things. The third one. So, this

15:09 was actually from middle school. So, in middle school, I was obsessed with um uh circuit diagrams, like actual circuit diagrams, and I would sketch them and I'd make blueprints everywhere. And you might think, "Oh, Alejandro, you lied. You're actually built different." Cuz in middle school, you were making all these blueprints. They mean literally nothing. Like, no, like straight up I like this one, this one on the right, uh I was like, "This is the blueprint for a robot." And like I just put a bunch of arrows cuz I knew that blueprints had arrows, you know? Like that's the level we're at. But I just had a passion for it and I thought they were cool. So, I just did them. But there was one thing I did know and I want you to know it, which is that this is like the one thing about engineering that I actually understood in middle school. It's that if you take a battery and you take a DC motor, you don't need to know what that

15:56 is. You can look it up. But if you just take a little motor and uh if the motor has a plus and a minus and you match them, if they some of them don't and it doesn't matter, but if you connect the battery to the motor, you know, two terminals to two terminals, the motor's going to spin. If you flip it, depending on the motor you have, it'll spin in the opposite direction. If you put a switch or a button interrupting one of the wires, you can control when it spins without, you know, breaking the wires. And with that alone, with that piece, you can build an infinite amount of things. This is all I knew in middle school and I went hog wild with it.

16:27 Here's an example. So, this this blueprint actually does make sense. So, what I did is I took two motors like that. I took that circuit and I built it twice. I stuck them together. I put Lego wheels on them and then I put I made a really long cable with the two buttons. So, it was like a control, you know, like a remote control. And then it was a car. And if you pressed one button, you it would actually turn because only one of the wheels would turn. If you press the other button, it would turn the other way. And if you press both, it would go forward. And then I put like cardboard spikes on them. I built some for my friends and we had battle bolts.

16:58 And uh you can build all kinds of things just with this. So now you know it. So now you're an electrical engineer. Congratulations. Okay, this one's current. Wow, look at that. I'm already getting claps and I'm like a third of the way through. This is awesome. So, this one's actually a current one. So, now we're combining them. Um, so this is a current project. Uh, I'm actually working on a parkour exoskeleton to negate fall damage. Imagine that. So, this is a brace that has a pneumatic piston with a spring on it. Here's me dropping off the roof of my patio without it. And here's me dropping off the roof with it.

17:35 Uh, it actually works terrible right now. It's super uncomfortable and it like kind of hurts, but you know, it's it's a first prototype. We're working on it. What I want you to take away from it though, other than that, exoskeletons are really cool, uh, is that you can actually, the takeaway, you can just build stuff in school. So, what I mean by that is I wasted I blew my savings in undergrad buying terrible tools like like lowquality tools and and lowquality parts cuz I couldn't afford them to just build janky things in my dorm room, not knowing that basically all universities and some high schools have machine shops and maker spaces that are accessible to students. So if you think that engineering and building things out of passion for fun for projects that you want to explore yourself like me, you know, building an exoskeleton to not break my knees or maybe do break my knees, um you can just go and do it. So

18:34 I want you to know that that resource exists and that there's really nothing stopping you from using them probably. Okay, here's another example. Just one more. Uh, I'm actually working I'm on a pursuit to build the ultimate indestructible notebook and I'm using the Caltech machine shop for it. So, this prototype is built out of marine grade ultracorion resistant grade 2 titanium. It glows in the dark and the pages are waterproof and then I laser etch it. It's pretty cool. Okay. Okay. Enough enough about my my upbringing. Now, let's make it back to California. Uh, well, or I guess to California. So, I started off my undergrad on a student visa in Santa Monica College. I actually went to community college and uh in Santa Monica College. Uh well, first off, here's a picture of me at my job um in the international student office.

19:25 They had this on the wall. I had somewhat of a reputation. Uh there's one highlight though that I want to give for you that we're going to give you um of my time at SMC. Uh and it's how I went from SMC to Caltech and it was through an organization called Base 11. And there's a lot of organizations like this. They're not alone, but they're the ones that did it for me. So, what Base 11 does is they create opportunities for students of underrepresented groups, from students from places where they would normally not have access to things like a summer internship at Caltech, doing research at a Caltech lab, which is something that is mostly accessible to Caltech undergrads. and they actually do stuff like they they create a connection with Caltech and they actually subsidize the cost of having that student that new student come from outside so that they can get that research and so base 11 has

20:18 a bunch of programs they still exist. I don't know what they're what they're working on right now. I don't know if they still have the summer program that I got. Um but I want you to know this because I had no idea. I was just sort of sitting pretty in Santa Monica hoping that someday I get to do research and a CS professor just happened to comment me, you know, like tell me in a passing comment that B 11 existed and I applied the day it was due and then I got my summer internship and it made my entire career. So these things exist and if you want a career in research, you should look them up and pursue them. Um, by the way, fun fact, uh, the way I got it, like my application was basically I mean I had good grades back then. um not not the case for my entire undergrad.

20:58 Um and uh what I what I did my application was basically talking about this silly little program I wrote when I was learning how to code and I only knew how to output text. So I made a program for doodling using text. So you know more doodling. Um okay so uh this is actually a group picture of uh the summer undergraduate research fellowship. So that that internship that I got thanks to base 11. Uh, there's me. You're gonna you're gonna start seeing the pattern. So, I spent that entire summer building prototypes of what later became the fan arrays that you saw in the beginning. So, building the really janky prototypes of these wind tunnels made out of grids of fans. Uh, and I documented everything with very thorough research notes structured as comic books featuring the two research students who had the misfortune of being in charge of me, Chris and Marcel. And so everything was documented as the adventures of

21:55 Chris and Marcel. Um, I'm sorry, Chris and Marcel. Okay. Uh, real quick, I'm going to kind of speedrun this part because I really want to make it to the research. Uh, so then I transferred to UC Berkeley and that's where I graduated from. And I guess you may be asking, "Oh, Alejandra, there's all these schools. How do you make the principal choice of going to UC Berkeley?" I just didn't get into Caltech. Don't don't get me wrong, UC Berkeley is a phenomenal school, but by then I had done, you know, my internship at Caltech. Caltech was my home. Uh I also didn't get into MIT and for that matter kind of unrelated um but it's important for the point I'm going to make. Uh, I failed my fluids qualifying exam in grad school, which when you think like basically what that means is as an aeronautics PhD, I failed the aeronautics exam, which is like I had one job. Um, and and okay, the reason I'm telling you that I didn't get

22:50 all these things um is actually because you won't either probably until you do, but you'll be okay either way as long as you stay engaged. What I'm trying to say is that when you and I'm talking mostly to the students obviously um when you see somebody in my position which I used to do when I saw somebody in my position you don't see how the way up to being in this privileged situation is paved by rejection after rejection after failure after missed opportunity after not getting into the things that you wanted to get into. And you're never going to get it the first time. and what you're looking for will probably not look the way you imagine it now. I actually originally wanted to go to Calgary for some reason. Uh it ends up looking better sometimes. So what I'm telling you is just, you know, don't be discouraged. Uh you're actually supposed to fail in the beginning. Okay. Uh also, by the way, um that's me um at Caltech

23:50 while I was at Berkeley. Um after my summer internship, they just hired me as research staff. So I actually spent my entire undergrad at Caltech working in the lab anyway. So, you know, it kind of worked out. Uh, all right. Let's just move to Caltech now completely. You guys you guys can see me now. You you see the pattern now. The the student faculty programs office was like, "Aleandra, can you please stop ruining our promotional material? These pictures are really expensive." Yeah. All right, let's talk about research.

24:21 So, uh, we have, I guess I'm going to say two parts for research. First off, let's talk about Fannori wind tunnels. I mentioned them in the beginning. So, now we're getting now we're actually building up to what I promised I was going to, you know, give you answers for. Uh, so, um, how do you teach a robot to fly? Let's go back to that original question. Well, uh, one of the key ingredients is you actually need wind. Kind of obviously that sort of goes without saying, but you need the right kind of wind. And specifically, you need the wind that the robot will actually fly in when it goes out in the real world. So, for example, if you're a plane, the kind of wind that you see is really fast and really clean.

25:01 What that means is it goes Okay, here, let's let's look at a video. So, let's look at this one. Aside from looking extremely cool, um, look at sort of what it looks like around the jet. There's nothing. It's clear skies and it's going really fast. So, that's the kind of wind that that an airplane sees. It sort of all goes in a straight line and it's really fast. And so the wind tunnels, well, they're literally tunnels that you use to replicate the flight conditions of an airplane do that. So this wind tunnel for this airplane is going to give you wind that goes mostly in a straight line and is so fast, you will literally die if you stand inside how fast it is. A drone is the opposite. So the kind of wind that a drone flies in, when when I say a drone, I mean something like Ingenuity. I mean something like a quadcopter. I mean something like a you know I don't know a hexa a delivery drone hexacopter that

25:54 kind of thing they see wind that is slow and turbulent. So let's talk about that and actually let's talk about that with Mars as an example first because you know I promised you Mars. So let's think about the challenge of replicating the flight conditions in Mars in order to learn to fly on Mars. So first off there's gravity. Gravity is two-fifths that of Earth. Okay. So what does that mean? That means that the force that you have to overcome in order to move up when you're on Mars is two-fifths. So it's smaller than the force that you have to overcome when you're on Earth.

26:27 That's good. And the way we replicate that is we actually just build skinny little robots on Earth. So this for example is uh it's called the scarecrow. So what JPL does uh in order to test that in order to sort of make that gravity equivalent on Earth is they actually build these skeletons that weigh like have the force on Earth that they would otherwise have on Mars. Does that make sense? Like they just strip them down so they actually have they weigh two- fifths on Earth. What such that they weigh what they would weigh on Mars?

26:59 Okay. So that's great. That's awesome. What else? The air is a 1/100th the density. It's about a hundred times thinner. Think about what that means if you're trying to fly. So, think about what it means to fly. No matter what you're doing when you're flying, whether you're flapping your wings, whether you're a fixed wing, whe like an airplane, uh whether you're a drone spinning a propeller, you're always pushing mass of air down to push yourself up in some way. So, what happens when there's no mass to push? It's really bad is what I'm saying.

27:36 And so how do we replicate that? Well, that's where Chris and Marcel and me come in alongside the rest of the JPL team. It's in replicating that condition. So let's talk about that. So how do you make Martian winds? Well, first you make Martian air and that's what the vacuum chamber is for. So the space simulator chamber is just a big sealed chamber and they pump out all of the air. They make it a vacuum and then they pump in an imitation of the Martian air. So it has both same chemical composition and it is also 100 times thinner approximately. And then you get the Martian wind off of, you know, the fan array wind tunnel, the giant grid of 441 computer fans. And then you blow the wind, add Ingenuity, and you, you know, can now replicate both Martian airs and Martian winds. Okay, but how do you fly in the air? So we have the air. We still have the problem though that, you know, it's 100 times thinner and you push it

28:27 and you don't get any any force pushing you up. So, so you know like we're still here. So, what do we do? You spin really fast. That's the answer. Um, actually, so I don't remember the exact ratio. It's something it's like six or seven times higher than why why are you guys laughing? Why are you guys laughing? I don't get it. I got two claps before the talk ended. This is awesome. But it literally is six or seven times though. Okay. But but okay, now hear me out though. This is really cool. This is like profound. So you have this idea, right? You've Well, I mean, I didn't.

29:08 Chris and Marcel and the JPL team did. That was their part. But you you're in this situation where you know now that you need to spin really fast. Okay. How do you do that? Because spinning faster is harder. Except the atmosphere is thinner, so you have less drag. So it's easier to spin faster. So actually the challenge cancels out. What you need to do to make up for the thinner atmosphere is spin faster. And the only reason you can spin faster is that the atmosphere is thinner. I think that's awesome. So what I'm saying is Ingenuity's got this.

29:45 And so now here is another test with a more mature ingenuity. Compare it to the dinky little, you know, level one ingenuity. Look at him go. Now it can hover and it can control itself and it can fly. And this is in the one 100, you know, of the, you know, density atmospheric air. It's awesome. Okay, so here it is again, just because I think it's really cool. Here's Ingenuity flying again on Mars. All right, we won at Mars. Now, let's come back home. Let's go to Earth. So, let's talk about the challenges of flying on Earth.

30:19 Let's go over sort of the same exercise we did again. Uh, well, uh, the gravity is one times that of Earth. So that's that's great. It's awesome. Uh you want to guess what the air is? Yeah. Okay. It's one times that of Earth. Okay. So it's the same thing. So we're good. Everything's perfect. There's no challenge. Of course, there's a challenge. But what is the challenge? Well, remember what I said that for drones, the kind of wind that you need to replicate for drones is slower and more turbulent. And why is that? And the reason for that is because on Earth, there's all this stuff. It's full of stuff. and and specifically I mean it's full of like buildings and trees and mountains. It's full of obstacles that the wind runs into. So you have these obstacles, the wind that's you know going in a straight line runs into them, wraps around them, runs into itself and creates these wakes and these vortices

31:11 and you can't see them. You don't know they're there. So when you're flying doing just obstacle avoidance, you run into them and you crash. And that's the challenge. And therefore, that's what we need to replicate in order to fly on Earth. And the way we're going to do that is with CAST, the Center for Autonomous Systems and Technologies. That's a mouthful. Um, so actually this is our drone arena here on Caltech. This is our fan array here on Caltech. By the way, that one has 2,592 fans. So it kind of it kind of moss the JPL one. Uh, and it's sort of an analogy to the JPL chamber in that it's this it's this space where you replicate the flight conditions and you have the same kind of air because it's outdoors physically speaking. Uh, and it is, you know, you have your your sort of specialized device that gives you the wind with the right air. Okay, let's talk about what we can do with fan

32:01 arrays. Uh, so like I said, you can control those fan all of those fans individually, right? Well, look at this. You can do stuff like this. So here the color tells you the speed of each fan. And when I was coding fan arrays for cast, I made all these programs that made all these super cool patterns. And I was like, "Oh, I made these sick patterns. They're going to crash the drones cuz they're so cool and they're completely useless." Uh, but they're really pretty. Um, but not all of them are useless. A lot of them are very useful. So here's an example of one. Uh so this actually if you take the fans and you make them go faster as they go up in height at in a certain way in a specific way you can imitate what is called the atmospheric boundary layer which is the way in which as you go up in altitude air gets faster on average in a certain direction in a certain way and you can replicate that by just

32:54 telling the fans to do the right thing. In fact we have a lot of experiments that I can't tell you about because I only have 40 minutes. Um, but we have experiments where we can go places on Earth, measure the wind, bring it back, and then replicate it. Like just plug it into the fan array and replicate it. It's really cool. And also you can draw Chris in the fan array. And I I love that. I love drawing Chris on the fan array, and he hates it. All right, let's talk about flying now. So, we have the wind conditions. We can create the environment. Now, let's talk about actually flying. So, I'm going to show you a baseline first. So, here I have the fan array. This is in the drone arena. I have this obstacle. That's me up over there, very underslept. And that drone is flying autonomously, just going side to side. And there's no wind. You can notice that it can do it perfectly.

33:41 Now, I'm going to turn on the finer array and I'm going to blow wind at it. And look at what happens. So, there's the drone. There's a ghost of the previous drone. So, you can compare them. And look at it go. Now, it can't track the line. It just suddenly starts wobbling around and you can't see why. It just it just does it, right? there's an invisible obstacle, an invisible disturbance that's just pulling the drone around. Okay, here's a most a more extreme example. So here, this one's more turbulent and now the drone is getting sucked into this recirculation and getting slammed into the building.

34:09 And if this was a real building and that thing was a real drone, like a big drone carrying something or god forbid a a you know, an air taxi, that would be catastrophic. So you know, we need to deal with that. And and this is what's happening, right? Like the wind is blowing into the obstacle. It's wrapping around and it's creating all these invisible disturbances that are doing stuff to the drone that it cannot see. So, we need to figure out how to see it. Uh, first I'm going to clarify something before we go deeper. When I say the wind and I talk about seeing the wind, right?

34:39 Uh, what you might think is, you know, what you see in the the weather forecast, which is like, you know, 20 miles an hour coming, you know, from the west, uh, which would be something like this, right? If that's the city. That's not what I'm talking about. What I care about is this wind. Is this picture? This is what I'm trying to see. Because those those spaces where the wind wraps around, that's what those those white lines are sort of following the wind. Uh those spaces are the places where if you fly through them and that's the pattern, you might crash. So that's what we need to know. We need something that paints that picture for us. Okay. So let's go over it again. How do you fly? And I'm going to focus on a city. So how do you fly in a city? Well, you have the city.

35:19 There it is. Uh, you need the wind. You need to figure out the wind, like I said. So, you need to paint this picture. And once we have it, we just fly around it. That's it. So, once we have the wind and we have an understanding of how our drone will interact with the wind, we can actually make an informed decision on how to fly through it and fly safely. So, here's the plan. We're going to predict the wind, which means we're going to model it. And that's a very important question, very important term that I'm going to define for you, um, because it means a lot in science. And then we're going to plan a path and then we're going to crash because that's just how it works. And then we're going to learn from it. We're going to crash and learn and then we're going to fly for real and it's going to work and it's going to be perfect. So let's talk about that word and this is the last term I'm going to teach you. So this

36:01 word model is very important in science. Um this if you guys go into STEM uh if if you study any STEM field at some point you'll have a professor who's going to give you this quote which is that all models are wrong but some are useful. And that's very true. Uh and I'm going to explain what that means. So in science when we say model, what we actually mean is a simpler version of the real thing. That's I'm using that very loosely because it's a very broad concept. So for example, if you have an airplane, if you're designing a whole airplane with, you know, all of its avionics and stuff and it's really big and it's really complex and then you build this basically toy like a 100 times smaller airplane that doesn't work. It's literally just a model airplane and you put it inside a wind tunnel and you blow wind at it and you measure how the wind interacts with it. If you know your aeronautics, you

36:49 can actually from looking at how the wind interacts with this little toy airplane, you can know how the wind will interact with the real one without having to build it. That's the power of a model that although it falls short of the real thing, is more manageable. Now, for us though, the models that we're going to be talking about are all math. Don't worry, I'm not going to show you any actual math, but they are math. Specifically, I'm talking about machine learning. You're gonna that's a buzz word that everybody hears. So, I'm going to demystify it a little bit. So, uh in short, machine learning means models from data through optimization. You don't need to understand this right now.

37:27 I want to say it because that's the accurate quote that I go by and it's true. What I do want you to remember though is that when you hear a model as in machine learning, when you hear AI, when you hear, you know, artificial intelligence, what that actually means is a formula. It's just a formula. That's it. It's it's like 2 plus 2 equals 4, but algebra. So there's like letters instead. It's just a really big formula and it has the quality that you can fit it to data. So you have this big formula. You collect a bunch of data and then you can tune the formula to match the behavior of your data. you can use it instead of the thing that's giving you the data. And I'm going to give you an analogy for it using ditto. So this ditto is our our machine learning model.

38:11 It's our formula that adapts to the real thing. And this Pikachu is the real thing. So this Pikachu is the wind in the cities with all of its infinite complexity. And what we're going to do is we're going to collect data. So we're going to give it a bunch of p pieces of Pikachu. I didn't chop a Pikachu in pieces. Just a bunch of chunks of Pikachu. That's the data collection. And then we're going to do a bunch of math to fit the DTO to the Pikachu. And then with the power of math, we're going to create a perfect indistinguishable replica of the Pikachu.

38:43 Remember what I said? All models are wrong, but some are useful. So that's what we're going to get. We're going to get a janky copy that is useful. So what does our model do specifically? Well, we take buildings, then we take our model, we feed it. So, our model is a it's a formula that you put a mathematical representation of what the buildings are into, and it paints that picture. It gives you what the wind around the buildings look like. That's the model. We need data for that. So, this it's kind of hard to see um just because the data is kind of complex, but what this actually is is it's a top view and it's called a heat map where the colors tell you the wind around cities. And so me and my colleague Peter are generating this giant data set of computer simulations of wind around a bunch of cities throughout the world that we're using to train our model. So those are the Pikachu bits and then we fit it. So this

39:36 is the model training. So that's the real thing in the sense that that's the computer simulation wind that is considered sort of the the gold standard that our model's trying to fit. And then this is the model's prediction as it trains with more and more data. And it's actually doing pretty good. So now let's use it. So, now we're gonna go back to our original challenge and we're gonna use our model. Are you ready? All right, let's you let's do it. Wait for it. WAIT FOR IT. OH, LOOK AT IT. Now you can see the wind. And actually the parts where those vectors that tell those those arrows that tell you how the wind is flowing, where they curve around, it matches where the drone gets gets, you know, wiggled. So now you can actually see the invisible thing that's throwing the drone around.

40:22 here. Let's go back to this one. Let's do it again. Wait for it. H there it is. And actually in that dark turbulent region when the drone enters that, that's where it gets sucked into the recirculation zone. So now we can see that invisible thing that throws the drone around that in the past the drone could only react to after being hit. So now what? Well, we can see the wind. Let's plan a path. Now we can actually execute a PL planning algorithm. So we have this sort of flow around the city. Now we can see it. So now we have the city, we have the wake. Let's say that we want to go from between those two orange dots. Well, if we didn't know the wind and we were just doing obstacle avoidance and looking for the shortest path, we would go this way. That way is bad. You don't want to do that. Uh but now we execute a path planning algorithm that takes into account the effect of the wind and we go around this way,

41:18 which is good. So, we did it. Success. So, are we done? Did we win? Did we Did we win at science? No. No, we did not. We have to actually fly it for real. And we haven't. So, so now now what? Well, actually, you're caught up with my research. Uh specifically uh this figure that you saw uh me and Peter got like we just happened to get our model to the point where we executed the path planning algorithm and generated this figure literally yesterday and then I rushed to put it in the talk so I could show you that's how like hot off the press this is. So you're caught up with my research.

42:00 So thank you. You have any questions? Thank you. I got three claps. That was awesome.

42:12 Okay, I think we're doing questions now. Thanks so much, Alejandra. That was awesome. Thank you.

42:24 Thank you, guys. So, I think we're going to have a couple microphones up here for Q&A. So, we're going to invite students to come up and ask your questions. Um, while they're getting those set up, uh, I have a question. Is Pikachu your favorite Pokemon?

42:45 Absolutely not. No, no. It's so generic. I'm not going to go for Pikachu. Oh, wow.

42:50 No, no, no. I like I like Scissor. I like Charizard, but only the shiny one. The black one. That one's really cool. Uh, and I also had a I had a shiny coughing when I was little, and then and then my Game Boy got stolen. Um, and then there was one more. Um, Kabuto tops and Dragonite.

43:09 Gotcha. I'm going to pretend I know what any of that means. The real ones know.

43:13 The real ones know. So, we've got the microphone set up. So, again, students, please feel come up and ask your questions.

43:21 Guess I'll ask you another one while folks are coming up. Um, is there a building that you'd particularly like to parkour off of? Anywhere in the world?

43:29 Anywhere in the world? Yeah. Anywhere in the world. Without legal liability. Right. I want to climb this one. It's

43:37 pretty good building. This one's pretty cool. I want to climb it. Oh, we have a line. We have people. Okay, I'm going to start here. Just just just start here. Hello. Hi.

43:50 Okay. So, how are you able to improve the drone so that way it didn't go smashing into the building?

43:57 Okay. Wait. So, there's there's different ways to do it. So most research uh is there's a lot of research on improving the drone itself. So ways to make the drone such that when it gets hit by that gust, it can actually respond quickly enough or it's built in such a way that the gust doesn't do anything to it. The way I'm trying to do it is I'm trying to be able to see the gust like I showed you and then well gust or wake or whatever. I'm trying to see the thing in the wind that would you know take it down and I'm actually trying to tell the drone how to fly through it. So the way it would actually go is kind of like that drawing that you saw, that plot that you saw at the end where what would actually happen now is I tell the drone to go from one side of behind the building to the other, right?

44:39 But this time it flies a little bit back to a part of the wake where it wouldn't get sucked into the building. That's what it would do using my approach.

44:49 Does that make sense? Yes. Good question. Let's do this side. There there are a lot of drones in the marketplace and they fly.

44:59 Yes. So how is your training different or in what way you're trying to improve that?

45:05 I understand you did it for Mars but but how what So that's the thing right most of the drones in fact all of the commercial drones that I'm aware of they they are like there's a lot of research that goes into making that drone really good and really robust but they don't take into account the wind. And for something like I mean focusing on the example of a city because if you're actually trying to deploy drones systematically through cities doing actual missions the standard of safety and reliability makes it such that not being aware of the wind is unacceptable because in most cases you're going to get away with it but there will be a gust that will run you into somebody's house. And so that is the contribution is that we would provide a framework by which to account for the wind and fly around it. And the thing is in a in a it the groundwork is already there like we know we have 3D models of cities right so we know the

46:00 geometry we have weather forecast so we know the boundary conditions and buildings are crammed with animometers already so we have sensors that also give us feedback on what the wind is around them so there's really no excuse why we shouldn't just know what the wind is and plan the paths and we're not doing that that's what's missing yeah this side hi

46:18 is ingenuity run by AI I or by humans. Okay. So, yes and no. So, so the thing about controlling robots on Mars, this is why this is why JPL is so advanced in autonomy. So, if you were trying to fly Ingenuity like you would fly a drone here with just a remote, the delay, the the lag between your command being sent and received by Ingenuity is 7 minutes. So, imagine playing a video game where you have seven minutes of ping. That's what it would be like. So you cannot control it as like that. So which means that there is AI and ingenuity. The thing is the AI and ingenuity is in a sense a lot more conservative than the AI that you would find in cutting edge things uh like like robots here on Earth because of how high the stakes are because if something fails on Mars, it's just done forever. So it has these really wellthoughtout like like compositions of relatively old algorithms that work really well. But

47:20 the way it works is JPL actually tells Ingenuity fly up in this way and then Ingenuity has to figure out on its own you know how to do it and how to account for whatever wind it's going to run into and then it has to scan the ground and then it has to land. So there's a what I'm saying is there's a mixture there's human supervision but ingenuity has to make its own decisions. It's not entirely self-sufficient. Does that answer your question?

47:45 Yes. Yes. Okay, cool. Thank you. Let's see this side. Hi. Uh, does anyone stop you when you climb?

47:54 All right. So, I feel free to tell you this because all of this happened in Venezuela, so it's outside of the jurisdiction. Um, so, so, uh, yes, I have been stopped from climbing, uh, given the fact that I like climbing things that shouldn't be climbed. Uh the funniest story is uh back in Venezuela uh my buddy Ector who you saw in the video and I we were running around looking for abandoned buildings to like you know go into and climb and we found this really janky building and we were like bro this is perfect. This is like a destroyed building. We're going to climb into it. Um and we did and it turned out to just be the world's ugliest functioning police department.

48:31 Thankfully, we realized that before anybody saw us, BUT WE DID CLIMB into it and then out. Uh, and then and yeah, we have had things like, you know, security guards getting angry at us. Sometimes they're polite. We obviously like we're respectful of the places that we climb on. So, we will climb in places that are already destroyed or we will climb on places where people let us climb or well, in all cases, we will it's kind of like um it's kind of like hiking, you know, you leave nothing but footprints and you take nothing but pictures. So, we try to never damage the environment that we're climbing. We are respectful of it. So, but yes, I have been stopped.

49:08 Hi. Hi. If it if you accidentally like get a piece of sher or a Pikachu, what will happen to the end?

49:20 Can you repeat that? If you accidentally get a wrong pieces of the Pikachu, what will happen?

49:26 Oh, I see what you're saying. So basically what happens if I have bad data what actually there there's a hierarchy when you're doing machine learning. So when you're doing these models off of data there's a hierarchy of the decisions you have to make. So for example you have to choose what the cost function is which basically means what you're going to literally tell your algorithm to learn to do. Um there and it has to be very carefully worded in a sense. There's a choice for what kind of formula you're going to use and then there's a choice of what data you're going to use. And actually the choice of the data is the single most important one. So if you have bad data, there's nothing you can do. If you have if your data is bad enough, your algorithm like your model will be garbage. However, all data is flawed. So the question really is how bad the data is going to be. And that's actually that's that's a really

50:14 good question because that's one of the many reasons why all models are wrong. It's because all models are built off of data and that data collection, that building of the pieces is always imperfect. So you can never get a perfect picture of the thing you're trying to imitate. So it makes your model worse and if the pieces are bad enough, your model is useless. That's the answer.

50:37 Okay. Hello. Hi. Hi. Um when you were when your teacher asked or like said that you should like be a

50:52 artist architect. Yeah. No, an artist when you grow up. When you said no, I'm going to be a scientist. Which scientist did you want to be?

51:02 Okay. Okay. Okay. That's that's a good one. So, um, I grew up watching Jimmy Neutron, so I kind of went down that path, but actually the way it went was I wanted to be whatever scientist Jimmy Neutron was. I don't know if you've seen Jimmy Neutron, but whatever. He was like a genius kid who like builds stuff. Eventually, uh, I decided I wanted to be a physicist. And then one day I had a conversation with my dad. I was I was hiking with my dad and my baby brother. And my baby brother goes, "Dad, what does Ally want to be when he grows up?"

51:28 And my dad said, he said, "Well, your brother Al is really smart and he likes science, so he's going to be a physicist. Too bad he's going to starve to death." And that made cuz physicists in Venezuela don't make a lot of money. So, uh, I decided not to be a scientist at that point and I studied computer science in Santa Monica and at Berkeley. But then I got this internship at Caltech and I just ended up a scientist again. So the short answer is first a physicist then I tried not to do that because I thought I was going to starve and now I'm an I'm an aeronautical engineer but okay there's a lesson behind this which is that as if you are studying STEM you're not limited to the name of the thing you're studying that's really really important I studied computer science in my undergrad and then I went to do my PhD in an aerospace lab you can do you can just do that and in the aerospace lab I met I met a guy, his

52:24 name's Chris Row. He's one of my mentors. Uh he studied entomology in undergrad. He studied bugs and then he went into the aerospace lab. I don't know how he did that. So the short answer is first physicist then aerospace. The important thing for you is it doesn't matter what you pick. You can always continue to find do what you want as you find it. Okay, that's the next one. Thank you. You guys are so you're such a good audience.

52:52 Hi What's your favorite ice cream flavor? Vanilla. Vanilla all the time. Tieuk vanilla bean. It's the best one and it cures headaches. I swear to God. Uh, let's do this one. Hi. How do you know what the like how do you know what the like climate is on Mars so you can like like create data to tell how to fly?

53:25 Yeah. Yes. Okay. That's a very exciting question actually. So the first time we didn't so for the first Ingenuity mission we had no idea. All we knew was the chemical composition. So we knew we basically knew what the air on Mars was. Uh and I think we had some idea of the wind speeds. I'm actually not sure. Um so that's what we did. Like we basically replicated like our experiments for the first ingenuity mission. There's going to be another one, but for the first ingenuity mission was uh we knew exactly what the atmosphere was made of. So we got the air right and then we just blew wind at it. And I actually don't know how well JPL understood the wind back then. I do know though that Perseverance, the rover that went with Ingenuity, has dedicated like really high precision wind sensors on it. So now we have a I know for a fact that we have a weather station on Mars. So now we're actually collecting data and it

54:19 would be I do actually know that there is work on rebuilding that fan array again. So you know, maybe we will actually be able to literally measure and replicate the wind from Mars. But the for the first mission, it was good enough to just know that there was wind and kind of guesstimate it. As far as I know, there's papers though. You can read them. They're accessible. They're not that hard to read. Um, did that answer your question?

54:42 Did Okay, good. Cool. All right. We have time for one more question. Okay. So, I have two questions actually. Number one,

54:51 is that cheating? Is he allow You're allowed. I You know what? I call the shots. You're allowed to do it.

54:54 Okay, fine. Um, so first one is if the drone flew into an environment that you had not studied beforehand, would it be able to gather data about the environment and actually make its own prediction? Bro, why do you have to get ahead of my research like that? That's literally the next plan. So yes, so the actually the current iteration of the algorithm does not take into account sensor feedback. It only takes into account boundary conditions. So you give it you tell it what the buildings are and you tell it what the incoming wind is and it just gives you a prediction and that's it. The more advanced version we're working on after that's more of a like spring kind of thing uh is actually one that can respond based on sensor. So it can basically take in sensor feedback and then adjust its prediction to make it consistent with it. And one of those sensors can be the drone itself. So if you understand the drone dynamics, if

55:52 you like if you understand how the drone interacts with wind, you can infer from how the drone got pushed what the wind speed was. So if you're in a situation where you tell the drone the wake looks like this, so fly this way and the drone is flying around like you told it. And then it runs into a wake it wasn't expecting, you can actually use that feedback as feedback to know, oh the wake is actually shaped like this and you know redo that. So that is the plan. Was there another question? Uh, the second one was just where are you flying the where are you going to test the model that you

56:21 Oh. Oh, that's all in the drone arena. So that's that's what you saw. So the this So basically what I'm doing is in that that big space that I showed you that I call the drone arena. I I basically have a bunch of wooden boxes where I'm building like a little city and then I just blow at them. I make like different test scenarios and then I test the algorithm against arrangements that it has never seen before. That's really important because you want to know how your algorithm generalizes. So yes, can I give him a question? I really want to give him a question.

56:46 Sure. Hi. Hi. Here's my question. How fast did the fastest drone fly? Oo, hold on.

57:01 Is Ivon here? Okay. Ivonne, who was outside, is literally trying to break the record. I wanted to ask him how fast his flew. He's actually building his own jet engines to do it. So I can tell you the Wait, do you know the speed?

57:21 He told me it was 450 mph. 450 mph. But I mean, so it is a drone in the sense that, you know, like by the broader term of a drone, a quadcopter, the fastest quadcopter that I know of personally could fly 60 miles an hour, which is still pretty fast. Yeah.

57:45 Well, thank you all so much. Thank you. Thank you so much. Heat. Heat. Heat.

Questions about this episode

Who speaks in "Flight School for Robots with Alejandro Stefan-Zavala"?

Alejandro Stefan Zavala (Guest), Lori Dah Jose (Host).

What is discussed in "Flight School for Robots with Alejandro Stefan-Zavala"?

In order: 0:00 Caltech, Ingenuity, Alejandro Stefan Zavala, CAST; 5:00 Caltech, Ingenuity, JPL, El Sal; 10:00 Maraca; 15:00 Caltech, Base 11, Santa Monica College; 20:00 Caltech, Base 11; 25:00 Ingenuity; 30:00 Caltech, Ingenuity; 45:00 Ingenuity; 50:00 Ingenuity, Caltech, Perseverance.

What is mentioned in this caltech episode?

11 people, works and ideas are mentioned, including Alejandro Stefan Zavala, Caltech, Ingenuity, CAST, Lori Dah Jose, JPL. Each one on this page links to its own page with every episode that mentions it.

How long is "Flight School for Robots with Alejandro Stefan-Zavala"?

The episode runs 58 min; the transcript is about 11,150 words in 100 timestamped sections.

More from caltech

• All analyzed episodes
• Books mentioned on caltech
• People mentioned on caltech