WEBVTT
1
00:00:02.000 --> 00:00:22.000
The hope is then we can begin to profile what the capabilities of the jammer are that we're seeing out in the field. With enough measurements from enough devices, we might spend weeks going through and analyzing that data, modeling it, trying to understand it, and then coming up with algorithms and techniques that allow us to either map out and interpolate the jamming areas or localize the jammer.
2
00:00:22.000 --> 00:01:19.000
This is the Convergence the Army's Mad Scientist Podcast I'm Matt Sanisper, Deputy Director of Mad Scientist, and I'll be joined in just a moment by Rachel Melling. Mad Scientist is a US army initiative that continually explores the evolution of warfare, challenges assumptions, and collaborates with academia, industry and government. You can follow us on social media at armymadsci or subscribe to the blog the Mad Scientist Laboratory at Madsci Blog Tradoc army mil Today we're talking with Dr. Sean Gorman, CEO and co founder of Zephyr XYZ, a developer of NextGen networked positioning technologies. We'll talk with him today about countering Russian jammers, how he came to work with the Ukrainian military, and commercial solutions in a GPS denied environment. As always, the views expressed in this podcast do not necessarily reflect those of the Department of Defense, Department of the Army, Army Futures Command, or the Training and Doctrine Command. Let's get started.
3
00:01:19.000 --> 00:01:22.000
Started. So Sean, welcome to the show.
4
00:01:22.000 --> 00:01:23.000
Hi, thanks for having me.
5
00:01:23.000 --> 00:01:36.000
All right, so we're going to talk today about some really cool stuff that you're working on with with the Ukrainians. But before we dive in, can you introduce yourself to our audience? Talk about who you are, what you do, and how you got to where you are today?
6
00:01:36.000 --> 00:03:04.000
Yeah, definitely. My name is Sean Gorman. I'm a CEO and co founder for Zephyr, and we build GNSS technologies to improve accuracy and resiliency of those systems and also providing countermeasures for detecting, jamming and spoofing and trying to localize where emitters are. I guess we came to this a bit of a of a long, circuitous route. I was originally an academic and we spun a company out of George Mason many years ago doing collaborative mapping, and we had a really core great team of folks within that, and we've kind of stayed together in different incarnations and done a couple of different startups. Most recently we were doing visual positioning systems for augmented reality and autonomy, and that ended up at Snapchat. And so we were there for a couple of years working on city scale augmented reality and mapping. So it was a bit of a diversion from a lot of the defense work we had done in the past. But we learned a lot about building large scale systems. But in the process of doing that, one of the things we realized that was super critical to doing visual positioning was having really high quality, accurate GPS positioning. And so when we left Snapchat, we thought we could take some of the techniques we had used from computer vision and AI and apply them to GNSS positioning and see if some of those new models could help us in some of the more traditional fields. And that's what brought us into doing Zephyr and then eventually into doing the, the work in Ukraine.
7
00:03:04.000 --> 00:03:22.000
That's really cool. We're very happy to have you here. And we're going to, as Rachel said, we're going to talk about some, some pretty cool things. So let's dive into that technology that you, you kind of alluded to there. Can you tell us about it? Can you talk about how it's able to mitigate GPS and comms denied environments on the battlefield? Yeah.
8
00:03:22.000 --> 00:06:01.000
So the core concept that the team came up with was that if we could get measurements from multiple devices and network them together, we could allow those devices to correct each other. So similar to like air tags with, with Apple phones, you use, you know, Bluetooth signals and they kind of triangulate with each other. We thought a similar concept could be used for, for mobile devices. So your phone can only see so many satellites and especially if you're in an area that is obscured, obscured by big buildings like an urban canyon or jammers that are attenuating and disrupting signals, once your phone is kind of occluded or precluded, you really start to run challenges with positioning. So the thought was if we could get measurements from lots of devices. So if a device is outside the jammed area or an open field that's not in an urban canyon, those high quality measurements could be used to correct a low quality measurements from another device that was being interfered with. And basically we create an ensemble optimization, taking measurements from a whole bunch of geographically dispersed devices and then allowing those to generate error corrections for the devices that we send back, and then they use those locally to be able to position with. So that was kind of the core concept that we came up with in a group from Ukraine came across the work we're doing and reached out and said, hey, could you help us out with Russian jamming? We have tons of problems with medevac and logistical operations that are really impactful. We said, well, we haven't really. We were doing purely commercial work at the Time. But we said we'd love to test out and see if we could learn and find a way to help. So we sent a couple of Android devices over there to start collecting data. And one of the things we realized, in addition to the phones doing a good job of adding some additional resiliency to the positioning, we found that the phones were also really good at mapping out where that jamming was happening or spoofing primarily. It was just about all jamming that we found in Ukraine. In other places like Israel, we'd find spoofing. So it kind of depended on the conflict. And then the other thing we found that in addition to kind of mapping out where that jamming was happening, that we could calculate kind of the angles of arrival of those signals and figure out where the emitters were coming from. And so as we kind of worked through this with the Ukrainians and kind of saw the iterative escalation of tactics for preventing positioning and then trying to get around obstructions, that maybe it was a bit of the Occam's razor. Right. Of what most direct approach. Figure out where the jamming is happening and avoid it, or figure out where the jammers are and get rid of the jammers. And so we started to invest more into working on that particular aspect of the problem.
9
00:06:01.000 --> 00:06:19.000
Yeah, so I feel like we've talked about the Ukrainian battlefield kind of being like a test bed for a lot of different technology, newer technology capabilities, and it seems like they reached out to you. But how did the Ukrainians find you? Like, how did, how did they come to you with this, with this problem set?
10
00:06:19.000 --> 00:07:24.000
Yeah, it was a bit of a circuitous path. A friend of ours who had been working on a standard for Bluetooth real time location tracking had written a really popular blog post and the Ukrainians had come across the blog post and reached out to her about the post and they said, hey, is this something that maybe could run in Ukraine? And she goes, actually, I'm in marketing. I, you know, I put together some code to demonstrate it, but I'm, you know, this isn't my full time profession, but you should talk to, to my friend Sean and Zephyr. They've been working on these technologies quite a bit and maybe they could help out. So she made the introduction for us and, and then that connected us up and, and they shared the, the problem set. We shared the research and prototypes we've been building. And so that's kind of when we got to the point of like, hey, let's, how can we help? Let's let's send some devices. And, you know, it's great with Andro that are, you know, electronic retailers all over the world, and you just have them, you know, buy five or six phones in the local market, throw the software on there, and then go out and do some testing.
11
00:07:24.000 --> 00:08:04.000
I think that's a great example of kind of how the modern battlefield looks and works right now. Usually we think about the transparent battlefield as being, you know, somebody able to see what's going on in a conflict, but it also works the other way around. You know, the, the parties involved in these conflicts are still connected to the rest of the world, and they're seeing adaptations and technology innovations, and they can pull them right into the conflict almost immediately. So let's talk a little bit about the testing that you're doing there. Can you tell us, give some examples or describe some of the tests or tell us, you know, really what you carried out or what the. I should say what the Ukrainians carried out on the battlefield. What kind of testing was it and what did they see?
12
00:08:04.000 --> 00:11:51.000
Yeah, we've done a couple of rounds of testing. The first round we really just wanted to see could we detect that jamming and was happening. And then also, could the network of phones improve the resiliency of positioning beyond just the standard GPS on the phone? And so for that scenario, we, you know, talking to the craniums of, hey, what kind of scenarios do you run into? Where would it be helpful? And we did a variety of kind of different modalities. So one was we just put phones on stands and put them out stationary out in the field near the front lines. Other ones, we'd mount them in cars and do a mounted scenario. Other ones we'd have folks that are on patrol walk around with the devices chest mounted. And then we also strapped them to drones, put them up in the air and flew them. And so one of the most interesting tests was we. We had some stationary phones set up around Donetsk near the front lines. And we had a phone mounted in a car, and we had a phone mounted on the drone. And then the drone flew above the car and. And it was, I guess, fairly obvious in hindsight, but the, the. The car was much more resilient to the jamming than the drone was because once the drone got up above any of the blockages from buildings and trees, it was able to have clear line of sight on the jammer. So it almost immediately started, you know, losing signal, whereas the car would kind of go in and out of being jammed and unjammed and and we, and we flew those kind of on top of each other. The drone would track the car so we could get an idea of, of what that jammer profile looked like across it. We did that both with a blue force where it was a Ukrainian jammer that they're using, and then also Red force where it was a big Russian omnidirectional jammer that was somewhere pushing out. And so that's what got us starting to think of, hey, if we have several of these jammers out there and we get this diversity of signals that are coming across and we understand the orientation of the phones, then that begins to give us some really interesting mathematical variables that we could plug into a model to try to triangulate back and figure out where that jammer was. And so in addition, we, you know, we built some models for, you know, building out some interpolations of the areas effect that we thought were happening from the jamming. And that tended to, to work pretty well. And then as we got interested in saying, hey, maybe we could localize in these things and figure out where the emitter is geographically, we went back and did a second set of tests where we, we set up some phones and we put a jammer in a car and drove it past the phones. But what we realized pretty quickly is that, you know, especially the Ukrainian jammers are very high powered. I think I covered that in a previous episode. They're really effective and they kind of were just so close they blew out the phones. And so we did a second test this past week where we, we set up, enforced some diversity of signals and tried to make it a little bit more realistic of saying, hey, there's going to be landscape that's out there that's going to be attenuating and interrupting these signals. And then let's set up phones and vehicles around that to collect those signals and see through that variation of signals. If we can calculate some angle arrivals and figure out that location happened. And the cool thing is that that hypothesis bore out in the testing that we did see that diversity and variety of signals that we think is going to give us a really good set of data to figure out where those jammers are located at. So that's kind of where we are. We just got that data set the last couple of days. So we're crunching on it and testing out the models and, and seeing how that kind of accuracy benchmark is going to work. So we have a whole bunch of ground truth for where those jammers were. We did four different scenarios mounted and Dismounted. And so now we're going to see how, how accurately we can predict where that jammer was across those, those location tests. And then if that runs well, we'll do a red force test and see how that works in a environment where we don't necessarily know where the, where the jammers at.
13
00:11:51.000 --> 00:12:01.000
Yeah, so I kind of wanted to ask you a question about that. So when the Ukrainians capture information using your technology, how long does it typically take you to get something usable back to them?
14
00:12:01.000 --> 00:13:51.000
Yeah, it kind of goes through a phase initially. You know, we're getting raw data that we've never seen before. And you know, we might spend weeks going through and analyzing that data, modeling it, trying to understand it, and then coming up with, with algorithms and techniques that allow us to either map out and interpolate the jamming areas or localize the jammer. But then once we have that model dialed in, like the interpolation of the area of effect of a jammer, we'll build that into the client on the phone and run that in real time. So kind of on, on the backside of doing that first Ukrainian test, we had a really wonderful opportunity to participate in the Insign propel program with PA staff out into paycom and go out and do demos and testing and talk to stakeholders and boots on the ground in Hawaii at Marine Corps based Hawaii and Hickam Air Force Base. And so we basically took everything we learned in Ukraine and then built a real time model that would run on the phone to interpolate jamming. And so we did some, some field testing out at Marine Corps Base Hawaii on the range there. And then we use like the mountains and the tunnels across Oahu as another interference source. Right. As you go through those mountains, you lose GNSS as you go through the tunnels. And then basically we wanted to see how well we could map out that interference as we went in addition to doing, you know, other kind of tests like in Ukraine and, and that all bore out really well. We could do that in real time. We could do it collaboratively with multiple devices and fuse those maps together and then that becomes a real time client that can run on the phone. So if you're disconnected, you can run those detections locally. And then as you get connectivity back to your comms network, you can sync up and share your data and then we can aggregate that into a collaborative map that provides that as a kind of situational awareness capability that could be integrated into other third party platforms. Although TACC is typically where we do our initial integration with, with the work that we've been doing with Air Force.
15
00:13:51.000 --> 00:14:05.000
Research labs, you kind of already touched on some of the things that you've learned through the Ukrainian testing. But can you kind of expand on some of the insights that you've gleane from the testing about Russian EW capabilities?
16
00:14:05.000 --> 00:15:32.000
Yeah, I mean, again, it's, you know, limited to the testing we've been able to do in those specific areas. But, you know, one of the things that we do is we'll, we'll do a calibration and profiling of the Ukrainian jammers that we're working against. So we have a free space model of what that area of effect will look like. And then when we do a blue force test, we'll collect that data and then baseline against the calibration against that jammer to see what it looks like in a real world scenario outside of that calibration model. And so that begins to give us a concept of what we see in the field, how that can equate to the capabilities of the jammer that we're receiving the signals from, especially as we get into calculating distance of how far away the emitter is and what kind of signal and disruption we're getting. The hope is then we can begin to profile what the capabilities of the jammer are that we're seeing out in the field with enough measurements from enough devices. But that's, you know, we've done that on the blue force side, but we haven't tried to extrapolate that over to the red force side yet. But that would be sort of the things obviously we'd be interested in doing in both in the Ukraine theater and in the Taiwan Pacific theater as well. That's one of the cool things, is as folks heard about the testing in Ukraine again to your point earlier, that with the clear space battlefield, folks see that capability and then reach out. So we've been talking to the folks in Taiwan that are interested in running similar kind of tests and understanding what that kind of EW landscape looks like.
17
00:15:32.000 --> 00:15:44.000
That's great. That's a perfect segue because I wanted to ask you about translating this technology over to the Indo Paycom alr. Is there any reason this wouldn't work there? Are you planning on doing further testing in the Indo Pacific?
18
00:15:44.000 --> 00:18:54.000
Yeah, definitely. That's the great thing about this is that, you know, we're just using run of the mill Android devices and really any GNSS device would work. Android makes it particularly straightforward because they provide the raw GNSS measurements off the phone. The Doppler the carrier phase, the automatic gain control, and all these things are omnidirectional antennas with very permissive models around the sensors that suck up as much measurements as they can. Because the reality is, like, your phone is kind of a crappy GNSS receiver. You know, the antenna is packed in with a whole lot of electronics. It's noisy, and as we know, with our little blue dot, it's not super precise many times. But the upside of that is since the antenna isn't the best, it needs to let in a bunch of signals to get the best possible positioning against it. So where many high end receivers, when they start to get radio frequency interference, they might shut down and stop positioning, whereas the phones will just continue to suck in these measurements even if they can't do a lot with them. And so that really makes them these great kind of sponges of EW measurements that we can grab and do interesting things with. And then the upside of that is there's so many of these devices that are out there that are capable and we can begin to pull those together. But one of the other things the team's done a lot of hard work on is not only supporting the Android phone specifically, but supporting the 5G 3GPP protocol that underlines how all GNSS receivers connect into telecommunications networks to do positioning. So while our test bed tends to be with Android phones because it's so easy to get them anywhere in the world, we have the ability and we've done integration into other GNSS receivers. So if you have, you know, receivers on drones or mounted or in a variety of other maybe IoT profiles, we can connect into any of those and use that bidirectional Supple LPP protocol within 3GPP to send those measurements across the network into our server for doing the exact same thing that we're doing with the phones. So that's very exciting as well. You know, some of the concept we've talked about with, with some of the partners in Taiwan is, you know, what's the potential of lighting up the whole island? Because if you have a 5G telco network that's providing connectivity to all the phones, there's nothing that's stopping you from turning every phone on the island into a sensor that's out there detecting EW events across, you know, the, the set of islands that are within Taiwan, which is another thing, I think a lot of folks aren't aware of that it's not only the main island of Taiwan, there's a whole set of islands that are much closer to China that are, that have populations sitting on them that could be part of that distributed sensor network. So that's some of the stuff, you know, that we get excited about of like really how can we scale this out? So, you know, tack is one wonderful vector that we can push into, but there's also a lot of public safety vectors because these are broad scale consumer technologies that are run. And if there's a good public safety or a good national security impetus for enabling these networks to become these defensive sensor networks, there's a great pathway to go and do that.
19
00:18:54.000 --> 00:19:11.000
So you've tested this with the Ukrainians and you'd like to test it with, in the Indopacom, aor, the Taiwanese. But how do you see this working for the US Army? And in your opinion, what should the army be doing to prepare for a future conflict in a denied environment?
20
00:19:11.000 --> 00:22:51.000
Yeah, it's a wonderful question. We're really fortunate that right around the same time we were doing the Ukraine testing that Air Force Research labs put out a Phase 2 SBIR for adding this ability to leverage signals of opportunity, for mapping out in a crowdsourced way a variety of different effects. And the one that we proposed was jamming and spoofing. And we're super lucky, lucky to win that sbir. So now we've been working with Air Force research labs to provide this capability into the TAC ecosystem, which is Wonderful. There's like 350,000 tac devices that are out there already. So that's a wonderful place to start as far as a large crowdsource sensor network within a controlled environment that we can begin to pull measurements from. But I think it's, you know, it's just one, one part of the equation. I think one of the keys to succeeding within this alt PNT environment is a layered approach that there is no silver bullet that's going to solve this. There's not this, you know, one technology that's going to go out saying, hey, we don't need GPS and gnss anymore because we have quantum or we have magnetic or we have visual positioning. And I think we learned this kind of in spades with our previous startup. We're working on visual positioning systems and we had the wonderful opportunity to work with, you know, a lot of the big players in AR that had some of the best scientists out there in the world and incredibly well funded to do it. And all of those visual positioning technologies are incredibly dependent on gps. And that was one of the reasons when we left we're like, hey, we need to make GPS better because it's so integral to all of these commercial technologies that it's really a limiting factor to what you can do with like the next generation of smart glasses and a wide variety of other things. And I think, you know, that that lesson of, of seeing, you know, even some of the best, most well funded scientists in the world struggling to have independence from gps. When we went and sat down with the Ukrainians, we saw the exact same thing, right? Because they're testing all these different techniques, they're reaching out to everybody that, you know, has something that they can put in the field and try out and inevitably it runs into, you know, very self similar issues. You don't have an initial known position for where you start from, whether inertial odometry or visual positioning or terrain based navigation. You don't know where to manage your error from that drift if you don't know where you're starting from. And there's very few things that would give me an initial known position other than a really robust GNSS system. So it's the interdependency from these things is so thick and interconnected. I think that trying to find this one thing that's going to solve the problem is almost impossible. It's really, how do you create a portfolio of investments that's going to give you a really layered approach to be able to tackle these things. So, you know, you know, your adversary might be able to take out one layer, but then you have four more layers that you can roll over to in an intelligent way. But that very much, you know, requires you to understand when are you being jammed? When was the last time you had a really good initial known position that you should be operating on? Because the one that you, you had five minutes later is corrupted. But you might think that was your last known position and you're already 50 meters off, off from where you actually are. And then that just, that error then propagates and all of a sudden you're, you know, you're way off kind of in left field. So I think that's, you know, kind of the way that we, we look at it is what's this kind of holistic approach that you can bring together to give you the, you know, the best onion layered approach that's going to be the most difficult for your adversary to be able to figure out how to corrupt and, and disrupt in the battlefield?
21
00:22:51.000 --> 00:23:18.000
That's fantastic stuff, Sean. You know, normally on this podcast we talk about the threat all the time and sometimes it can get a Little heavy. So we like to bring a little levity to these conversations. Not to say that this conversation was somber in any way, but we're going to switch now to our rapid fire questions to give you a chance to kind of let the audience know a little bit more about you. We ask these to every one of our guests. They're always the same and just whatever the answer is off the top of your head here. So what's a technology or a trend that keeps you up at night?
22
00:23:18.000 --> 00:24:27.000
I'd say, I guess to the, to the tenor of the, of the topic. You know, AI automation around kill chains is, is something that it's, you know, especially we've spent a lot of time recently of trying to integrate our technology into generative AI and LLMs for creating spatial awareness within LLM. So if it knows what you're looking at, it can connect that up to knowledge around it and just understanding kind of how janky these things can be and how much work needs to be done to prevent hallucinations and prevent it from kind of going off the rails at times. It's amazing and fabulous and totally nails it it. And with the, the right setup, you can really get these things to do. Amazing capabilities. But it's also just as easy for them to go off the rails and kind of go sideways. And so, you know, when you kind of think of the, the logical connection of some of the work that we're, that we've been doing, of how that begins to automate, you know, taking out a target, you know, just the amount of work that that's done in a really solid and predictable way. If there's one thing that kind of keeps me up at night, it's probably how easy it is potentially to mess that up and then what are the repercussions.
23
00:24:27.000 --> 00:24:45.000
Yeah, that's understandable. Yeah, there's a lot of work still to be done, but we probably wouldn't make any trigger man somebody who's prone to hallucination. So that should be the same for, for our generative AI. Question number two. What's something about you that most people might not know?
24
00:24:45.000 --> 00:26:12.000
That's a good one. I. This one's a little bit random, but when I was in grad school, I was doing some research that ended up raising some eyebrows within the intelligence community. And I think it was one of these, like, slow days. And it got picked up by the news. And so the reporter that was covering it went to the White House just during the Bush administration to date myself and Richard Clark, who was the White House cyber security czar asked him about, you know, what do you think about this dissertation and what its impact is going to be? And he has a bit of a penchant for salacious quotes and was like, should get his degree and then burn the dissertation. And I guess to. I guess it doesn't make sense without quickly saying what it was is that I'd been doing a bunch of research on Digital Divide where I'd mapped out fiber optic infrastructure around the globe. And post 9 11, people really interested in where that fiber optic infrastructure was because it connects financial exchanges and all these other things. And so we kind of pivoted the research at the time to looking at where there were vulnerabilities in the global and domestic fiber optic network and how it connected into these critical assets, which all of a sudden went from, you know, being kind of a Digital Divide issue to a national security issue. And we had really detailed maps, though, that freaked people out.
25
00:26:12.000 --> 00:26:16.000
And you successfully defended your dissertation and got your doctor it right.
26
00:26:16.000 --> 00:26:21.000
Yeah, and they didn't burn it either. It's still in the George Mason University Library somewhere buried.
27
00:26:21.000 --> 00:26:32.000
You came out the winner in the end. All right, Sean, this is our final question. Lots of times it's the hardest question, but it's the most fun for us. So what's your favorite movie?
28
00:26:32.000 --> 00:26:57.000
That's a. That's a tough one for sure. Probably a toss up between Inception and Lord of the Rings, the. The Two Towers as far as just a classic, fun adventure with lots of good tactics and strategy in it as well. And then Inception is just a great mind trip, creative, fun way to also escape from things, which are my favorite kind of movies.
29
00:26:57.000 --> 00:27:12.000
I agree wholeheartedly. My two boys and I, we just watched all three of the Lord of the Rings extended editions over the past couple weeks. We got through them and know several nights to get through one movie. Rachel, you've seen the first one.
30
00:27:12.000 --> 00:27:22.000
I was gonna say I'm the slacker. I've only seen the first one, but I've heard that the second one is the best. So I need to really setting me up here for. For a good time.
31
00:27:22.000 --> 00:27:27.000
And Inception, that was one of the. The best movie going experiences I ever had. I saw it in IMAX down in Atlantic City.
32
00:27:27.000 --> 00:27:28.000
Oh, nice.
33
00:27:28.000 --> 00:27:51.000
And it was, it was amazing. The. The Spinning top at the end and it goes straight to black. The whole crowd was into it. It was great. So, yeah, we approve those. All right, Sean, this has been a great conversation. We've learned a lot about jamming and potential ways to use technology that's been democratized to take advantage of finding out where jamming is occurring. So we want to thank you for coming on and talking to us and being on the Convergence.
34
00:27:51.000 --> 00:27:57.000
Thanks for having me. This is a wonderful conversation and really enjoy the opportunity to share the team's work with everyone.
35
00:27:57.000 --> 00:28:31.000
Thanks for listening to the Convergence. I'd like to thank our guest, Dr. Sean Gorman. You can follow us on social media me madsci and don't forget to subscribe to the blog the Mad Scientist Laboratory at madsci Blog Tradoc Army Mil. Finally, if you enjoyed this podcast, please consider giving us a rating or review on Apple, Spotify, or wherever you accessed it. This feedback helps us improve future episodes of the Convergence and allows us to reach a bigger and broader audience.
1
00:00:02.000 --> 00:00:22.000
The hope is then we can begin to profile what the capabilities of the jammer are that we're seeing out in the field. With enough measurements from enough devices, we might spend weeks going through and analyzing that data, modeling it, trying to understand it, and then coming up with algorithms and techniques that allow us to either map out and interpolate the jamming areas or localize the jammer.
2
00:00:22.000 --> 00:01:19.000
This is the Convergence the Army's Mad Scientist Podcast I'm Matt Sanisper, Deputy Director of Mad Scientist, and I'll be joined in just a moment by Rachel Melling. Mad Scientist is a US army initiative that continually explores the evolution of warfare, challenges assumptions, and collaborates with academia, industry and government. You can follow us on social media at armymadsci or subscribe to the blog the Mad Scientist Laboratory at Madsci Blog Tradoc army mil Today we're talking with Dr. Sean Gorman, CEO and co founder of Zephyr XYZ, a developer of NextGen networked positioning technologies. We'll talk with him today about countering Russian jammers, how he came to work with the Ukrainian military, and commercial solutions in a GPS denied environment. As always, the views expressed in this podcast do not necessarily reflect those of the Department of Defense, Department of the Army, Army Futures Command, or the Training and Doctrine Command. Let's get started.
3
00:01:19.000 --> 00:01:22.000
Started. So Sean, welcome to the show.
4
00:01:22.000 --> 00:01:23.000
Hi, thanks for having me.
5
00:01:23.000 --> 00:01:36.000
All right, so we're going to talk today about some really cool stuff that you're working on with with the Ukrainians. But before we dive in, can you introduce yourself to our audience? Talk about who you are, what you do, and how you got to where you are today?
6
00:01:36.000 --> 00:03:04.000
Yeah, definitely. My name is Sean Gorman. I'm a CEO and co founder for Zephyr, and we build GNSS technologies to improve accuracy and resiliency of those systems and also providing countermeasures for detecting, jamming and spoofing and trying to localize where emitters are. I guess we came to this a bit of a of a long, circuitous route. I was originally an academic and we spun a company out of George Mason many years ago doing collaborative mapping, and we had a really core great team of folks within that, and we've kind of stayed together in different incarnations and done a couple of different startups. Most recently we were doing visual positioning systems for augmented reality and autonomy, and that ended up at Snapchat. And so we were there for a couple of years working on city scale augmented reality and mapping. So it was a bit of a diversion from a lot of the defense work we had done in the past. But we learned a lot about building large scale systems. But in the process of doing that, one of the things we realized that was super critical to doing visual positioning was having really high quality, accurate GPS positioning. And so when we left Snapchat, we thought we could take some of the techniques we had used from computer vision and AI and apply them to GNSS positioning and see if some of those new models could help us in some of the more traditional fields. And that's what brought us into doing Zephyr and then eventually into doing the, the work in Ukraine.
7
00:03:04.000 --> 00:03:22.000
That's really cool. We're very happy to have you here. And we're going to, as Rachel said, we're going to talk about some, some pretty cool things. So let's dive into that technology that you, you kind of alluded to there. Can you tell us about it? Can you talk about how it's able to mitigate GPS and comms denied environments on the battlefield? Yeah.
8
00:03:22.000 --> 00:06:01.000
So the core concept that the team came up with was that if we could get measurements from multiple devices and network them together, we could allow those devices to correct each other. So similar to like air tags with, with Apple phones, you use, you know, Bluetooth signals and they kind of triangulate with each other. We thought a similar concept could be used for, for mobile devices. So your phone can only see so many satellites and especially if you're in an area that is obscured, obscured by big buildings like an urban canyon or jammers that are attenuating and disrupting signals, once your phone is kind of occluded or precluded, you really start to run challenges with positioning. So the thought was if we could get measurements from lots of devices. So if a device is outside the jammed area or an open field that's not in an urban canyon, those high quality measurements could be used to correct a low quality measurements from another device that was being interfered with. And basically we create an ensemble optimization, taking measurements from a whole bunch of geographically dispersed devices and then allowing those to generate error corrections for the devices that we send back, and then they use those locally to be able to position with. So that was kind of the core concept that we came up with in a group from Ukraine came across the work we're doing and reached out and said, hey, could you help us out with Russian jamming? We have tons of problems with medevac and logistical operations that are really impactful. We said, well, we haven't really. We were doing purely commercial work at the Time. But we said we'd love to test out and see if we could learn and find a way to help. So we sent a couple of Android devices over there to start collecting data. And one of the things we realized, in addition to the phones doing a good job of adding some additional resiliency to the positioning, we found that the phones were also really good at mapping out where that jamming was happening or spoofing primarily. It was just about all jamming that we found in Ukraine. In other places like Israel, we'd find spoofing. So it kind of depended on the conflict. And then the other thing we found that in addition to kind of mapping out where that jamming was happening, that we could calculate kind of the angles of arrival of those signals and figure out where the emitters were coming from. And so as we kind of worked through this with the Ukrainians and kind of saw the iterative escalation of tactics for preventing positioning and then trying to get around obstructions, that maybe it was a bit of the Occam's razor. Right. Of what most direct approach. Figure out where the jamming is happening and avoid it, or figure out where the jammers are and get rid of the jammers. And so we started to invest more into working on that particular aspect of the problem.
9
00:06:01.000 --> 00:06:19.000
Yeah, so I feel like we've talked about the Ukrainian battlefield kind of being like a test bed for a lot of different technology, newer technology capabilities, and it seems like they reached out to you. But how did the Ukrainians find you? Like, how did, how did they come to you with this, with this problem set?
10
00:06:19.000 --> 00:07:24.000
Yeah, it was a bit of a circuitous path. A friend of ours who had been working on a standard for Bluetooth real time location tracking had written a really popular blog post and the Ukrainians had come across the blog post and reached out to her about the post and they said, hey, is this something that maybe could run in Ukraine? And she goes, actually, I'm in marketing. I, you know, I put together some code to demonstrate it, but I'm, you know, this isn't my full time profession, but you should talk to, to my friend Sean and Zephyr. They've been working on these technologies quite a bit and maybe they could help out. So she made the introduction for us and, and then that connected us up and, and they shared the, the problem set. We shared the research and prototypes we've been building. And so that's kind of when we got to the point of like, hey, let's, how can we help? Let's let's send some devices. And, you know, it's great with Andro that are, you know, electronic retailers all over the world, and you just have them, you know, buy five or six phones in the local market, throw the software on there, and then go out and do some testing.
11
00:07:24.000 --> 00:08:04.000
I think that's a great example of kind of how the modern battlefield looks and works right now. Usually we think about the transparent battlefield as being, you know, somebody able to see what's going on in a conflict, but it also works the other way around. You know, the, the parties involved in these conflicts are still connected to the rest of the world, and they're seeing adaptations and technology innovations, and they can pull them right into the conflict almost immediately. So let's talk a little bit about the testing that you're doing there. Can you tell us, give some examples or describe some of the tests or tell us, you know, really what you carried out or what the. I should say what the Ukrainians carried out on the battlefield. What kind of testing was it and what did they see?
12
00:08:04.000 --> 00:11:51.000
Yeah, we've done a couple of rounds of testing. The first round we really just wanted to see could we detect that jamming and was happening. And then also, could the network of phones improve the resiliency of positioning beyond just the standard GPS on the phone? And so for that scenario, we, you know, talking to the craniums of, hey, what kind of scenarios do you run into? Where would it be helpful? And we did a variety of kind of different modalities. So one was we just put phones on stands and put them out stationary out in the field near the front lines. Other ones, we'd mount them in cars and do a mounted scenario. Other ones we'd have folks that are on patrol walk around with the devices chest mounted. And then we also strapped them to drones, put them up in the air and flew them. And so one of the most interesting tests was we. We had some stationary phones set up around Donetsk near the front lines. And we had a phone mounted in a car, and we had a phone mounted on the drone. And then the drone flew above the car and. And it was, I guess, fairly obvious in hindsight, but the, the. The car was much more resilient to the jamming than the drone was because once the drone got up above any of the blockages from buildings and trees, it was able to have clear line of sight on the jammer. So it almost immediately started, you know, losing signal, whereas the car would kind of go in and out of being jammed and unjammed and and we, and we flew those kind of on top of each other. The drone would track the car so we could get an idea of, of what that jammer profile looked like across it. We did that both with a blue force where it was a Ukrainian jammer that they're using, and then also Red force where it was a big Russian omnidirectional jammer that was somewhere pushing out. And so that's what got us starting to think of, hey, if we have several of these jammers out there and we get this diversity of signals that are coming across and we understand the orientation of the phones, then that begins to give us some really interesting mathematical variables that we could plug into a model to try to triangulate back and figure out where that jammer was. And so in addition, we, you know, we built some models for, you know, building out some interpolations of the areas effect that we thought were happening from the jamming. And that tended to, to work pretty well. And then as we got interested in saying, hey, maybe we could localize in these things and figure out where the emitter is geographically, we went back and did a second set of tests where we, we set up some phones and we put a jammer in a car and drove it past the phones. But what we realized pretty quickly is that, you know, especially the Ukrainian jammers are very high powered. I think I covered that in a previous episode. They're really effective and they kind of were just so close they blew out the phones. And so we did a second test this past week where we, we set up, enforced some diversity of signals and tried to make it a little bit more realistic of saying, hey, there's going to be landscape that's out there that's going to be attenuating and interrupting these signals. And then let's set up phones and vehicles around that to collect those signals and see through that variation of signals. If we can calculate some angle arrivals and figure out that location happened. And the cool thing is that that hypothesis bore out in the testing that we did see that diversity and variety of signals that we think is going to give us a really good set of data to figure out where those jammers are located at. So that's kind of where we are. We just got that data set the last couple of days. So we're crunching on it and testing out the models and, and seeing how that kind of accuracy benchmark is going to work. So we have a whole bunch of ground truth for where those jammers were. We did four different scenarios mounted and Dismounted. And so now we're going to see how, how accurately we can predict where that jammer was across those, those location tests. And then if that runs well, we'll do a red force test and see how that works in a environment where we don't necessarily know where the, where the jammers at.
13
00:11:51.000 --> 00:12:01.000
Yeah, so I kind of wanted to ask you a question about that. So when the Ukrainians capture information using your technology, how long does it typically take you to get something usable back to them?
14
00:12:01.000 --> 00:13:51.000
Yeah, it kind of goes through a phase initially. You know, we're getting raw data that we've never seen before. And you know, we might spend weeks going through and analyzing that data, modeling it, trying to understand it, and then coming up with, with algorithms and techniques that allow us to either map out and interpolate the jamming areas or localize the jammer. But then once we have that model dialed in, like the interpolation of the area of effect of a jammer, we'll build that into the client on the phone and run that in real time. So kind of on, on the backside of doing that first Ukrainian test, we had a really wonderful opportunity to participate in the Insign propel program with PA staff out into paycom and go out and do demos and testing and talk to stakeholders and boots on the ground in Hawaii at Marine Corps based Hawaii and Hickam Air Force Base. And so we basically took everything we learned in Ukraine and then built a real time model that would run on the phone to interpolate jamming. And so we did some, some field testing out at Marine Corps Base Hawaii on the range there. And then we use like the mountains and the tunnels across Oahu as another interference source. Right. As you go through those mountains, you lose GNSS as you go through the tunnels. And then basically we wanted to see how well we could map out that interference as we went in addition to doing, you know, other kind of tests like in Ukraine and, and that all bore out really well. We could do that in real time. We could do it collaboratively with multiple devices and fuse those maps together and then that becomes a real time client that can run on the phone. So if you're disconnected, you can run those detections locally. And then as you get connectivity back to your comms network, you can sync up and share your data and then we can aggregate that into a collaborative map that provides that as a kind of situational awareness capability that could be integrated into other third party platforms. Although TACC is typically where we do our initial integration with, with the work that we've been doing with Air Force.
15
00:13:51.000 --> 00:14:05.000
Research labs, you kind of already touched on some of the things that you've learned through the Ukrainian testing. But can you kind of expand on some of the insights that you've gleane from the testing about Russian EW capabilities?
16
00:14:05.000 --> 00:15:32.000
Yeah, I mean, again, it's, you know, limited to the testing we've been able to do in those specific areas. But, you know, one of the things that we do is we'll, we'll do a calibration and profiling of the Ukrainian jammers that we're working against. So we have a free space model of what that area of effect will look like. And then when we do a blue force test, we'll collect that data and then baseline against the calibration against that jammer to see what it looks like in a real world scenario outside of that calibration model. And so that begins to give us a concept of what we see in the field, how that can equate to the capabilities of the jammer that we're receiving the signals from, especially as we get into calculating distance of how far away the emitter is and what kind of signal and disruption we're getting. The hope is then we can begin to profile what the capabilities of the jammer are that we're seeing out in the field with enough measurements from enough devices. But that's, you know, we've done that on the blue force side, but we haven't tried to extrapolate that over to the red force side yet. But that would be sort of the things obviously we'd be interested in doing in both in the Ukraine theater and in the Taiwan Pacific theater as well. That's one of the cool things, is as folks heard about the testing in Ukraine again to your point earlier, that with the clear space battlefield, folks see that capability and then reach out. So we've been talking to the folks in Taiwan that are interested in running similar kind of tests and understanding what that kind of EW landscape looks like.
17
00:15:32.000 --> 00:15:44.000
That's great. That's a perfect segue because I wanted to ask you about translating this technology over to the Indo Paycom alr. Is there any reason this wouldn't work there? Are you planning on doing further testing in the Indo Pacific?
18
00:15:44.000 --> 00:18:54.000
Yeah, definitely. That's the great thing about this is that, you know, we're just using run of the mill Android devices and really any GNSS device would work. Android makes it particularly straightforward because they provide the raw GNSS measurements off the phone. The Doppler the carrier phase, the automatic gain control, and all these things are omnidirectional antennas with very permissive models around the sensors that suck up as much measurements as they can. Because the reality is, like, your phone is kind of a crappy GNSS receiver. You know, the antenna is packed in with a whole lot of electronics. It's noisy, and as we know, with our little blue dot, it's not super precise many times. But the upside of that is since the antenna isn't the best, it needs to let in a bunch of signals to get the best possible positioning against it. So where many high end receivers, when they start to get radio frequency interference, they might shut down and stop positioning, whereas the phones will just continue to suck in these measurements even if they can't do a lot with them. And so that really makes them these great kind of sponges of EW measurements that we can grab and do interesting things with. And then the upside of that is there's so many of these devices that are out there that are capable and we can begin to pull those together. But one of the other things the team's done a lot of hard work on is not only supporting the Android phone specifically, but supporting the 5G 3GPP protocol that underlines how all GNSS receivers connect into telecommunications networks to do positioning. So while our test bed tends to be with Android phones because it's so easy to get them anywhere in the world, we have the ability and we've done integration into other GNSS receivers. So if you have, you know, receivers on drones or mounted or in a variety of other maybe IoT profiles, we can connect into any of those and use that bidirectional Supple LPP protocol within 3GPP to send those measurements across the network into our server for doing the exact same thing that we're doing with the phones. So that's very exciting as well. You know, some of the concept we've talked about with, with some of the partners in Taiwan is, you know, what's the potential of lighting up the whole island? Because if you have a 5G telco network that's providing connectivity to all the phones, there's nothing that's stopping you from turning every phone on the island into a sensor that's out there detecting EW events across, you know, the, the set of islands that are within Taiwan, which is another thing, I think a lot of folks aren't aware of that it's not only the main island of Taiwan, there's a whole set of islands that are much closer to China that are, that have populations sitting on them that could be part of that distributed sensor network. So that's some of the stuff, you know, that we get excited about of like really how can we scale this out? So, you know, tack is one wonderful vector that we can push into, but there's also a lot of public safety vectors because these are broad scale consumer technologies that are run. And if there's a good public safety or a good national security impetus for enabling these networks to become these defensive sensor networks, there's a great pathway to go and do that.
19
00:18:54.000 --> 00:19:11.000
So you've tested this with the Ukrainians and you'd like to test it with, in the Indopacom, aor, the Taiwanese. But how do you see this working for the US Army? And in your opinion, what should the army be doing to prepare for a future conflict in a denied environment?
20
00:19:11.000 --> 00:22:51.000
Yeah, it's a wonderful question. We're really fortunate that right around the same time we were doing the Ukraine testing that Air Force Research labs put out a Phase 2 SBIR for adding this ability to leverage signals of opportunity, for mapping out in a crowdsourced way a variety of different effects. And the one that we proposed was jamming and spoofing. And we're super lucky, lucky to win that sbir. So now we've been working with Air Force research labs to provide this capability into the TAC ecosystem, which is Wonderful. There's like 350,000 tac devices that are out there already. So that's a wonderful place to start as far as a large crowdsource sensor network within a controlled environment that we can begin to pull measurements from. But I think it's, you know, it's just one, one part of the equation. I think one of the keys to succeeding within this alt PNT environment is a layered approach that there is no silver bullet that's going to solve this. There's not this, you know, one technology that's going to go out saying, hey, we don't need GPS and gnss anymore because we have quantum or we have magnetic or we have visual positioning. And I think we learned this kind of in spades with our previous startup. We're working on visual positioning systems and we had the wonderful opportunity to work with, you know, a lot of the big players in AR that had some of the best scientists out there in the world and incredibly well funded to do it. And all of those visual positioning technologies are incredibly dependent on gps. And that was one of the reasons when we left we're like, hey, we need to make GPS better because it's so integral to all of these commercial technologies that it's really a limiting factor to what you can do with like the next generation of smart glasses and a wide variety of other things. And I think, you know, that that lesson of, of seeing, you know, even some of the best, most well funded scientists in the world struggling to have independence from gps. When we went and sat down with the Ukrainians, we saw the exact same thing, right? Because they're testing all these different techniques, they're reaching out to everybody that, you know, has something that they can put in the field and try out and inevitably it runs into, you know, very self similar issues. You don't have an initial known position for where you start from, whether inertial odometry or visual positioning or terrain based navigation. You don't know where to manage your error from that drift if you don't know where you're starting from. And there's very few things that would give me an initial known position other than a really robust GNSS system. So it's the interdependency from these things is so thick and interconnected. I think that trying to find this one thing that's going to solve the problem is almost impossible. It's really, how do you create a portfolio of investments that's going to give you a really layered approach to be able to tackle these things. So, you know, you know, your adversary might be able to take out one layer, but then you have four more layers that you can roll over to in an intelligent way. But that very much, you know, requires you to understand when are you being jammed? When was the last time you had a really good initial known position that you should be operating on? Because the one that you, you had five minutes later is corrupted. But you might think that was your last known position and you're already 50 meters off, off from where you actually are. And then that just, that error then propagates and all of a sudden you're, you know, you're way off kind of in left field. So I think that's, you know, kind of the way that we, we look at it is what's this kind of holistic approach that you can bring together to give you the, you know, the best onion layered approach that's going to be the most difficult for your adversary to be able to figure out how to corrupt and, and disrupt in the battlefield?
21
00:22:51.000 --> 00:23:18.000
That's fantastic stuff, Sean. You know, normally on this podcast we talk about the threat all the time and sometimes it can get a Little heavy. So we like to bring a little levity to these conversations. Not to say that this conversation was somber in any way, but we're going to switch now to our rapid fire questions to give you a chance to kind of let the audience know a little bit more about you. We ask these to every one of our guests. They're always the same and just whatever the answer is off the top of your head here. So what's a technology or a trend that keeps you up at night?
22
00:23:18.000 --> 00:24:27.000
I'd say, I guess to the, to the tenor of the, of the topic. You know, AI automation around kill chains is, is something that it's, you know, especially we've spent a lot of time recently of trying to integrate our technology into generative AI and LLMs for creating spatial awareness within LLM. So if it knows what you're looking at, it can connect that up to knowledge around it and just understanding kind of how janky these things can be and how much work needs to be done to prevent hallucinations and prevent it from kind of going off the rails at times. It's amazing and fabulous and totally nails it it. And with the, the right setup, you can really get these things to do. Amazing capabilities. But it's also just as easy for them to go off the rails and kind of go sideways. And so, you know, when you kind of think of the, the logical connection of some of the work that we're, that we've been doing, of how that begins to automate, you know, taking out a target, you know, just the amount of work that that's done in a really solid and predictable way. If there's one thing that kind of keeps me up at night, it's probably how easy it is potentially to mess that up and then what are the repercussions.
23
00:24:27.000 --> 00:24:45.000
Yeah, that's understandable. Yeah, there's a lot of work still to be done, but we probably wouldn't make any trigger man somebody who's prone to hallucination. So that should be the same for, for our generative AI. Question number two. What's something about you that most people might not know?
24
00:24:45.000 --> 00:26:12.000
That's a good one. I. This one's a little bit random, but when I was in grad school, I was doing some research that ended up raising some eyebrows within the intelligence community. And I think it was one of these, like, slow days. And it got picked up by the news. And so the reporter that was covering it went to the White House just during the Bush administration to date myself and Richard Clark, who was the White House cyber security czar asked him about, you know, what do you think about this dissertation and what its impact is going to be? And he has a bit of a penchant for salacious quotes and was like, should get his degree and then burn the dissertation. And I guess to. I guess it doesn't make sense without quickly saying what it was is that I'd been doing a bunch of research on Digital Divide where I'd mapped out fiber optic infrastructure around the globe. And post 9 11, people really interested in where that fiber optic infrastructure was because it connects financial exchanges and all these other things. And so we kind of pivoted the research at the time to looking at where there were vulnerabilities in the global and domestic fiber optic network and how it connected into these critical assets, which all of a sudden went from, you know, being kind of a Digital Divide issue to a national security issue. And we had really detailed maps, though, that freaked people out.
25
00:26:12.000 --> 00:26:16.000
And you successfully defended your dissertation and got your doctor it right.
26
00:26:16.000 --> 00:26:21.000
Yeah, and they didn't burn it either. It's still in the George Mason University Library somewhere buried.
27
00:26:21.000 --> 00:26:32.000
You came out the winner in the end. All right, Sean, this is our final question. Lots of times it's the hardest question, but it's the most fun for us. So what's your favorite movie?
28
00:26:32.000 --> 00:26:57.000
That's a. That's a tough one for sure. Probably a toss up between Inception and Lord of the Rings, the. The Two Towers as far as just a classic, fun adventure with lots of good tactics and strategy in it as well. And then Inception is just a great mind trip, creative, fun way to also escape from things, which are my favorite kind of movies.
29
00:26:57.000 --> 00:27:12.000
I agree wholeheartedly. My two boys and I, we just watched all three of the Lord of the Rings extended editions over the past couple weeks. We got through them and know several nights to get through one movie. Rachel, you've seen the first one.
30
00:27:12.000 --> 00:27:22.000
I was gonna say I'm the slacker. I've only seen the first one, but I've heard that the second one is the best. So I need to really setting me up here for. For a good time.
31
00:27:22.000 --> 00:27:27.000
And Inception, that was one of the. The best movie going experiences I ever had. I saw it in IMAX down in Atlantic City.
32
00:27:27.000 --> 00:27:28.000
Oh, nice.
33
00:27:28.000 --> 00:27:51.000
And it was, it was amazing. The. The Spinning top at the end and it goes straight to black. The whole crowd was into it. It was great. So, yeah, we approve those. All right, Sean, this has been a great conversation. We've learned a lot about jamming and potential ways to use technology that's been democratized to take advantage of finding out where jamming is occurring. So we want to thank you for coming on and talking to us and being on the Convergence.
34
00:27:51.000 --> 00:27:57.000
Thanks for having me. This is a wonderful conversation and really enjoy the opportunity to share the team's work with everyone.
35
00:27:57.000 --> 00:28:31.000
Thanks for listening to the Convergence. I'd like to thank our guest, Dr. Sean Gorman. You can follow us on social media me madsci and don't forget to subscribe to the blog the Mad Scientist Laboratory at madsci Blog Tradoc Army Mil. Finally, if you enjoyed this podcast, please consider giving us a rating or review on Apple, Spotify, or wherever you accessed it. This feedback helps us improve future episodes of the Convergence and allows us to reach a bigger and broader audience.
