GEEKOUT: When Drones Descend with Jon Rogers

Jump to Podcast File.
Download File


Additional content from Jon Rogers on how drone technology is evolving at Georgia Tech.


[radio dial scanning, static]

[big band swing]

Steve McLaughlin:  It sounds incredibly complex—it sounds like—to build these that span—I’m really geeking out here.

[big band music]

[archival recording from Superman] —

>> Up in the sky!
>> It’s a bird!
>> It’s a plane!
>> It’s Superman!

Steve McLaughlin: Why can't I just fly my own drone and just start flying around the neighborhood and maybe fly to the next neighborhood? And then you start thinking of aircraft and, you know, the regulations that you're talking about that need to be in place for all this stuff to kind of work. And it does make sense that that's going to take a long time.

Jon Rogers: So the regulatory piece has not caught up with the technology, and we see that with a lot of emerging technologies. I mean look at what's been going on with Uber with, you know, New York regulating Uber licenses. We've seen that with cyber security, right? You know a lot of new technologies and the regulations aren't out there, social media what have you, drones are another instance of that where this technology is there; it's at people's fingertips. Can you go fly your drone into your neighbor's yard and set up a camera 6 feet off the ground? Well, certainly your neighbor wouldn't like it if you put a camera up on, you know, a tripod on their lawn. But people are flying drones into their neighbor's yards with cameras on them and they look at it differently. The regulations on drones are really just modified versions of manned aircraft regulations and they don't fit—a square peg in a round hole. We need we need people who understand the technology and understand the capabilities, probably to work with all makers to develop laws, and then we also need norms in our society that say, Hey, you know that may not have been against the law, but it was not right.

Steve McLaughlin: As you were talking about the me flying a drone into my neighbor's yard, there's obviously privacy considerations around that. Like you said, you know, in some instances people would view that as a huge privacy invasion.

[archival recording from Superman] —

>> Striding swiftly through the editorial department, Clark Kent steps into an empty storeroom and locks the door behind him.

Steve McLaughlin: I think it's a technical term, but it’s also because we're a Georgia Tech we talk about “swarms.” I think—do you call those swarm robots? You know where it sounds like the technical challenge you're talking about is the collaboration that needs to go on. Let's say there's a wounded soldier and you need three or four or five of these smaller—obviously those drones are talking to each other and need to coordinate. That's the kind of the technical hurdles around that collaboration between the drones?

Jon Rogers: That's certainly part of it. So how do five drones pick up a flexible payload like a person and ensure that they can maintain that stable flight, ensure certain guarantees in terms of stability? This whole idea of cooperative flight control, where they're actually physically manipulating the same object with flexible characteristics, is a very difficult problem, and it's been studied in some very controlled contexts. But when you have something as flexible as a person, it gets very complicated. And a lot of the stability guarantees we've relied on for traditional aerospace community cannot be made anymore.

Steve McLaughlin: You know I can I can imagine for drones picking up a heavy piece of iron or something that's not flexible is pretty easy. But as soon as you have—particularly if you have a patient who's a huge amount of duress, and that adds, that adds an awful lot.

Jon Rogers: We're not going to see it deployed anytime soon because there's still a lot of basic research problems that need to be solved in terms of constraints and how these vehicles fly the vehicle under the constraints of not trying to hurt the patient more than they already are.

[archival recording] —

>> Superman came from another planet, a planet called Krypton just before it exploded in space. A couple of years ago, a piece of the planet came to Earth like a meteor.

Steve McLaughlin: Are those problems that you need to solve, essentially mathematical problems and algorithms and things like devices? Is that the kind of research that you're talking about?

Jon Rogers: Yeah so there's two major areas; one is an algorithm side—so how do the vehicles communicate? What control systems are they running? What sensors do they need and how does that sensor feedback play into the control loop? And that essentially becomes a math problem, detailed complex math problem. And there's various algorithms to solve it. We've looked at bring in machine learning algorithms combining machine algorithms with more classical control algorithms. So we've done work in that area.

The other big area that needs to be investigated are mechanisms. And being a mechanical engineer, at least part of my research in mechanical engineering looking at how these drones actually connect to a person. How can they ensure that they can connect with nobody else to actually go physically connect them? That that connection mechanism has to be extremely reliable. It has to be able to be executed autonomously and disconnected autonomously and also not injure a person.

How does that machine connect to a person without possibly injuring him or her? That's another difficult question.

[archival recording from Superman] —

>> I’ll be on top of the world again—Kryptonite, golden slipper, Eddie.

Steve McLaughlin: An overly-used term—is “interdisciplinary” multidisciplinary? But clearly that's exactly what you were talking about. It’s not just about the math; it's about having students and yourself that have expertise in a broad set. I mean besides mechanical engineers, what are the kind of the skills of someone that would need to develop the kind of thing we're talking about?


Jon Rogers: Yeah. So we need folks—we need folks who know soft robotic mechanisms. So there's a whole class of robotics that deals with soft mechanisms that can interface with the human body without causing harm. There's computer science in machine learning techniques, so people—we need folks who know about Bayesian inference, who know about neural networks to handle that uncertainty that comes from that flexibility of the human. We need people who know about handling qualities of air vehicles so that we can understand, you know, we can understand how to ensure or guarantee stability in a cooperative framework. And we also need multi-agent control folks who know about tasking multi-agent systems and, you know, the ability—we don't have a central planner; we’ll have these vehicles that have to make decisions together.

[archival recording from Superman] —

>> Just mention one word to him—Kryptonite.

Steve McLaughlin: Why would we deploy Tarzan as opposed to drones? But maybe it's that persistence, the fact that the life span of Tarzan as you I think that's kind of where you started—months or years it could be out there entirely on its own. And that's really the advantage over it because clearly drones could probably do some of the same kinds of things.  Is that—

Jon Rogers: Absolutely. And a lot of folks have asked us, “Well, what's the benefit over a drone?” And the drone is certainly a capable platform, but the flight times are extremely limited. The ability of someone who doesn't know anything about a robotic system to operate one versus the other is different, right? So, you know, someone who has a large farm, they're going to be more comfortable flying an aircraft or deploying an aircraft around their farm and not knowing anything about aircraft or regulatory issues or putting this relatively cheap robot that just kind of swings along and it makes sense how it works. So we think that that barrier to entry leans a little bit more toward these types of cheap, brake-eating robots.

We're also focused a lot on algorithms, too. So we have a lot of work with partners and we're starting to grow that area of research in what's really called the stochastic optimal control, and it's like a mouthful of technical mumbo jumbo, but really it's about how should autonomous systems make decisions under uncertainty which happens all the time. It's actually been—these decision-making algorithms have been used in the finance community for years in terms of where to sell. But for autonomous vehicles, the autonomous vehicles of the future, they're not going to have Google Maps telling them where to go. They're going to have a vision sensor that’s going to have to figure out what they're looking at. And so they're going to need these control algorithms that say, Hey, I'm uncertain about where I am or what to do, but I still need to make a decision. What's the optimal decision I should make? So we're working on that.

And what I'm seeing is a lot of the ideas that we started in our lab and we worked on them a little bit, they're starting to pay off. Where we—we’re seeing those applications; we're able to push them toward experiment more. And that's been one of the best parts about, you know, moving into this seventh, eighth year my academic career is watching those early investments transition into, you know, actual hardware experiments and technologies and then, also having the students who have that capability and that expertise to make that happen.

[archival recording from Superman] —

>> Now, up with the window. Out and away!

[dramatic organ music]

Steve McLaughlin: You know, one of the things that I see is the barriers dropping between aerospace engineering, mechanical engineering. And all the time we get because, you know, a lot of high school students don't necessarily know what engineers do. And then, but they're good in math and science and so they, you know, gravitate towards engineering and then they have to decide, Should I be an electrical engineer or mechanical engineer or aerospace engineer? What would you say to those students? I know what I say to those students, but I want to hear what you have to say to the students. Does it really—does it matter to those students?

Jon Rogers: Exactly—it doesn't matter. I really don't think it does. I think about the students in my group. So many of my students have been mechanical engineers, and I put them on projects that are made more traditional aerospace engineering. They're smart people; they learn what they need to know. If they have a rock-solid understanding of the fundamentals, you know, they understand calculus, right, and they understand—my students understand circuits; they understand the basics of each discipline which is what we teach here at Georgia Tech the first and the second year, the basics of each discipline. Then the specialized topics, they’ll get a book and they learn it in a month or I work with them one-on-one. And so as we see engineering evolving, these silos are disappearing. What I look for in my group is a really smart person. I was a physics major undergrad, and when I got to grad school, I knew nothing about lift curves, drag curves, wings. I mean, I knew what an airplane did, flown many airplanes, but I didn't really know how it worked. And so I had to spend a few months learning; it wasn't that bad.


Steve McLaughlin: Well, I probably teed you up for that because I couldn't agree with you more because I think, you know, again, for the audience out there that's trying to figure out whether engineering's for them, it is. You know, it doesn't matter as much as it used to because those lines are really, really blurry. And it's just a creative mind and inquisitive mind that has some of these skills. There's just so many incredibly cool things out there. And you're talking about, you know, a thousand a thousand of them.

[archival recording from Superman]—

>> I'm going to find a way to find Superman tonight.

[dramatic organ music]

[marching band music]