The Robot Report Podcast
00:39:38 · Mike Oitzman
[00:39:38] S1: I I mean, the other problem that that I've heard from some developers is... In egg, it's in the orchard. You get under the canopy. Maybe you're losing GPS or in your delivery robot example, you're in some concrete canyon between buildings in some urban setting and, again, signals bouncing around. Now suddenly things are a little bit confused as to exactly where you are. Does this solve that? Or how... Or how does that RTK and the other technologies evolve?
00:40:07 · Aaron Nathan, Eugene Demetri
[00:40:07] S2: Yeah. It's it's a great question. And this is why, like, Point One is this broader thing of a location company because RTK by itself does not magically make urban canyons work. It doesn't magically make, you know, the tree cover work. What it does do is it allows us to do what we call sensor fusion, which ultimately allows you to take other sources of information. So it could be, like, an inertial measurement unit that tells you kind of your relative motion. It could even be, like, a camera or a LiDAR scanner, a normal use perception we actually can use to help integrate that solution forward in time. And the idea is, like, well, how do you blend, you know, that with RTK? And that problem is actually much harder to do well. There's, you know, obviously, like, very simple ways that you could do it where you have something that just, like, switches, which is what a lot of people try, and then you end up with this, like, handoff problem. So you end up, kind of as you start to peel the layers of the onion realizing that this problem is way better if you could just have a magic RTK that worked all the time. And that's kind of our north star for point one is, like, how do you just deliver centimeter level accuracy regardless of the environment? And, that's really the path that, we're on and what we're delivering to our customers. [00:41:35] S3: You mentioned other technologies. How do you see RTK in relation to approaches like SLAM and Vision or newer ones like LEO?
00:41:44 · Aaron Nathan
[00:41:44] S2: Yeah. I mean, RTK is very unique in that it does provide you the absolute frame reference that basically no other sensor is able to do. And I would say that this is true of any satellite nav system where you are sort of, like, building one grid across the safe... Surface of the earth and able to put everybody in that same kind of grid. I like to think about it as like... It's like a language, and everyone's speaking the same language because you can reference each other and know what you're talking about when you're saying, hey. I'm at this position. Like, that means something to the people around them. That is... Sounds almost obvious, but it's, like, not by default how these other approaches you mentioned like SLAM work, where SLAM traditionally is a very egocentric way of navigation, meaning that, like, I know where I am relative to what I've seen, and I don't care about anything else. And so that's great. That's, like, how your, you know, robot actions work. But the reality is that we are not, again, of this mindset where you must always use RTK. The real answer is that you have to combine the capabilities that SLAM delivers with these absolute capabilities of technologies like RTK. And that's a lot of what we are exploring and working with our customers on. Very often, you know, our customer may have, like, a SLAM solution that they've implemented, and then they'll be using the RTK system that we have to be able to integrate those two. Same thing with, like, vision based navigation systems, like lane followers. To give maybe a concrete example, we're in production cars today. And so if you think of a production car, you'll basically have these, like, two stacks running in parallel. One of them is a camera looking at lanes on the highway and saying, like, keep the car in between the two lines of paint. That's basically, you know, the algorithm. And then the other one is saying, have this high definition map that came from, you know, a different company, like Ushr. Right? And that's effectively, like, GPS waypoints that define, like, the central line of every lane on the highway. And then they'll use point one to say, well, I know I'm exactly in Lane two on 280 in, you know, Mountain View, and now you have these completely two independent sources that tell you this is exactly where I am.
00:44:20 · Aaron Nathan, Eugene Demetri
And now that gives you a high degree of confidence to make a safe decision about should I change lanes or exit the highway or any of these, you know, actions. So, yeah, I guess, like, I'm a little bit of a broken record, but it's always about how do I use all the things together as opposed to be in competition with a different technology. [00:44:43] S3: Well, and you mentioned the absolute frame concept. What happens when many robots or systems are working from the same map? Does it work across platforms like drones that are moving at one speed and cars moving at another and delivery robots moving at yet another? Like, how how does that work together?
00:45:01 · Aaron Nathan
[00:45:01] S2: Yeah. So our technology is agnostic to the platform, but I need to be a little bit careful in how we say that because we do actually take advantage of the fact that we know that, you know, we're locating a drone or we're locating a tractor or a car, whatever the actual physics of the underlying thing is. The important part is this absolute frame is like a grid that we all agree on that defines the location and also things like the speed and direction of things. And the reason that's kind of a cool concept is if you think about, like, a port where you have, like, a bunch of, let's say, forklifts driving around, and they're all moving these big containers, they may not have, like, great visibility about where they're about to go. So your camera, your LIDAR may be blocked. There's this whole world where they actually can communicate in real time their positions and their directions and actually say, oh, I'm gonna slow down because I can see, quote, unquote, this collision happening purely just by observing, like, the almost video game view of the world. And I think that this is where, you know, the current state of robotics is that each robotic company goes to a site and builds a map. And that's, like, needed because we're so early. Right? And so the eventual solution, though, is there will absolutely be interop between these things. And I don't think it's gonna be just because, like, it's nice to have. I think it's actually going to drive, like, the core thing that robotics are supposed to do, which is to drive efficiency and throughput and all these things. So, yeah, I mean, in my mind, that's an inevitable future, and it's why, location to me, it's like it has to be on this shared common grid, and it can't be, you know, each robot company figuring out its own thing. It's just inherently less efficient.
00:47:07 · Mike Oitzman
[00:47:07] S1: Let's let's talk a little bit about some of the markets then that you're serving. We've we've dropped a few of them already. Tell us some of the other interesting use cases for for the technology. I don't know if you're able to share some of the companies that have deployed Point One Nav into their platforms.