AWS in Orbit: Modernizing Satellite Management

[ Music ]

>> Maria Varmazis: I'm Maria Varmazis, host of "T-Minus Space Daily," and this is "AWS in Orbit," Modernizing Satellite Management. Today, we're bringing you the next installment of the "AWS in Orbit" podcast series from the 40th Space Symposium. In this episode, I'm speaking with representatives from Arcfield Orion and AWS Aerospace and Satellite. And we're going to do a deep dive into the challenges that the AWS Modular Framework for Satellite Operations solutions solves. Gentlemen, thank you so much for joining me today. I think this will be the perfect time for us to do some intros. So, we're going to go from my left to right, if you don't mind. Could you please start?

>> Junk Wilson: Yes, happy to. Appreciate the opportunity to be here today. My name is Junk Wilson. I'm a Senior Vice President with Orion Space Solutions. We're an Arcfield company and I am really excited to talk to you today about one of the solutions that we've developed called Orbit IQ. A little bit about my background. I started my career in the Air Force. I was a graduate of the Air Force Academy with a degree in physics. Spent about 20 years flying fighters. So, I flew F-15s and F-117s. Fantastic airplanes. Really enjoyed that. Several overseas deployments. And then after I retired from the military, I got back into physics and realized that I really had a passion for understanding how the space environment is impacting a wide variety of things, from satellites to our civilization here on Earth. With Orion, I've had the opportunity to lead a team that we call Models and Applications. And within that team, we've developed some fantastic technology with the help of AWS and we're really excited to share that with you today.

>> Maria Varmazis: I'm looking forward to learning more about that today.

>> Ed Meletyan: Hey, I'm Ed Meletyan. I'm a Solutions Architect at AWS, part of Aerospace and Satellite, and I'm going to be talking a little bit about dynamic node orchestration on AWS. Excited to share that later on, but my background is mostly in mission design, mission planning, scheduling. I started my career at -- at Raytheon working on weather satellite programs, and I've eventually worked my way through different mission domains over to rendezvous proximity operations missions, something really dynamic, space situational awareness and some other kind of interesting projects before I came to AWS to help customers leverage the cloud to do things at scale in a way that's reliable and really push the envelope on what the technology can do.

>> Maria Varmazis: Very cool. Thank you.

>> William Duhe: I'm William Duhe. I'm a Senior Software engineer with AWS. I work for the AIML team. I like to consider myself a technologist, jack of all trades, software sourcer, love everything technology. My background's in physics and computational mathematics. Academically, I was trained in cosmology and particle physics. Transitioned from that into the software engineering world with Silicon Valley. I've worked with Maxar, Lockheed Martin and some of the biggest aerospace companies in the world that provide satellite imagery at scale to customers around the world. That's transitioned me here to AWS where I've leveraged those skills with the AIML team to develop pipelines, tools and a specific tool called Oversight ML that we're going to be presenting here at the conference that helps customers adopt the cloud for these mission critical software solutions.

>> Maria Varmazis: Yes, well, I'm looking forward to learning all about that, as well. So, Junk, Ed and Duhe, thank you all three of you for joining me today. So, why don't we sort of set the stage first? So, I know each of you sort of has a -- a component of the framework that you are going to be talking to me about. So, Junk, can we start with you? Tell me a little bit about Orion Orbital IQ.

>> Junk Wilson: Yes, happy to. So, you know, we are a space company. We build and fly our own satellites. And about ten years ago, we realized that the free products that are available from NOAA Space Weather Prediction Center or SWPC as it's known, really weren't meeting our needs completely. They're a great product and it's, you know, we're great to be, you know, partners with SWPC, but they didn't give us the fidelity of data that we needed about the Earth environment to really understand where our satellite was going to be and when was it going to be there. And that has a wide variety of implications, not the least of which was, you know, we have to buy antenna time in order to talk to our satellite. And so, if we are off by even a few seconds, those few seconds of antenna time that we buy are really expensive. And so, we realized that for our own purposes, we needed a better orbit propagation solution and we needed to really reduce the uncertainty in that math calculation, which the largest source of uncertainty below a certain altitude anyway, is drag, is neutral density and understanding how that neutral density is impacting it. And a lot of people think that when you are in low Earth orbit, you're in the vacuum of space and it's certainly lower pressure, but it's not actually a --a pure vacuum. And so, that neutral density and that drag force is actually really significant. We started investigating what it would take to run our own models. We are a company full of computational physicists, so we thought we could do our own pretty well and we realized that what was available in the market at the time, didn't have quantified uncertainty. You could make some guesses. You could say, "Hey, I think there's about this much uncertainty in these various numbers," but from a computational science perspective, that was deeply unsatisfying. So, we realized we had a problem for ourselves that we needed to solve. There was nothing in the market yet that could do it. So, we decided to -- to take that on as a challenge and see if we couldn't, not only solve our own problem, but maybe turn that into a solution that could be a -- a marketable thing that could be part of a -- of a larger enterprise. So, Orbit IQ builds on that legacy of, you know, we think of it as being, you know, by owner operators satellite, owner operators and for owner operators, right? So, we know what you need because we are you. We are people who fly satellites, and we have our own satellites and that's super important to us. One of the things that's really unique about Orbit IQ is that it runs a variety of ensembles. So, if you're not familiar with that term or what that means, it's very similar in nature to the hurricane weather forecasters that you see on TV. Right? You've probably seen those spaghetti charts where it looks like the hurricane could hit any one of a number of places.

>> Maria Varmazis: Yes.

>> Junk Wilson: Every one of those lines of spaghetti is a member of an ensemble, and all of those lines of spaghetti together form an ensemble. And we do the same thing in space environment modeling. So, rather than just one -- run a single model of the space environment, we run over 100 models of the environment, and we take every one of those outputs and feed that into our orbit propagation software. Our orbit propagation software is also unique in the fact that it was designed from the ground up to be run like a Monte Carlo scenario. It's another type of ensemble, but the two ensembles are actually tuned to work together. So, we have an ensemble of space environments models, and we run that with an ensemble of propagation models. And together what that allows you to do is it allows you to sample the population of the ensemble at each step and give you measured uncertainty. So, some of the other products in the market will say things like, "Oh, well, we assume a, you know, a 7-kilometer uncertainty bubble around our spacecraft." We don't have to assume anything. We model it over 100 times in a row and we do it in parallel and we run the model and the orbit propagation together. So, we actually have a measured uncertainty --

>> Maria Varmazis: Calculated, yes.

>> Junk Wilson: -- bubble and it turns out that our uncertainty bubble is significantly smaller than what everybody else is assuming. So, from a -- a space traffic management --

>> Maria Varmazis: Yes, yes.

>> Junk Wilson: -- or from a conjunction assessment perspective, if you have a -- a whole bunch of relatively large bubbles, you're going to get an awful lot of false positives. You're going to think that things are running into each other when in fact they're actually quite far apart. And only with the measured uncertainty provided by Orbit IQ, with the help of AWS, of course, can you actually get that precision and that measured uncertainty, so that you can actually do the space traffic coordination and space traffic management that we all know is necessary in this modern space age.

>> Maria Varmazis: Yes, I was going to say that sounds transformative for space domain awareness, which you know, kind of a hot topic.

>> Junk Wilson: Absolutely. I mean, one of our favorite things to do right now, we are working with the -- the Space Domain Awareness Tap Lab here in Colorado Springs, and they have a whole variety of problems that they're trying to solve, not the least of which is trying to understand what's called camouflage, concealment and deception maneuver, CCDM. I think we all know that not everybody in space is a friendly actor anymore.

>> Maria Varmazis: Right.

>> Junk Wilson: And so, there are an awful lot of times where people are doing one thing, but they're saying another.

>> Maria Varmazis: Yes.

>> Junk Wilson: And the only way to separate those two is with very accurate understanding of the uncertainties associated with orbit propagation and orbit determination. So, we love to work with the SDA TAP Lab. It's a fantastic organization and we know AWS is embedded there, too. So, we're really excited to -- to continue that partnership there.

>> Maria Varmazis: Fantastic. Now, Ed, I'm going to move to you because we were talking about, you know, being able to understand better risk in different -- in -- in orbit. And now, I'm thinking about also how satellites are going to be communicating with each other. So, I feel like this is a good time to -- to introduce sort of your -- your component here.

>> Ed Meletyan: Yes, so I'm going to be talking a little bit about dynamic node orchestration on AWS. And really this came about from a need in the space industry to do physics and network simulations at scale. And what dynamic node orchestration on AWS, or I'll call it DYNO for short, what it does is it lets you -- customers build digital twins of very complex dynamic systems that may have thousands, tens of thousands or hundreds of thousands of moving nodes. And a node could be a network device, a router, could be a ground station, it could be a satellite, it could be a person or an aircraft. So, all these things can work together to accomplish a mission. Now, how well are these things going to work together to accomplish that mission? It's really hard to say in a very dynamic environment when none of the nodes really have any idea of what the other nodes in the system are doing. That's why we built this digital twin to bring all of those things in a virtual form into the cloud and be able to connect them and make sense of them at scale. And what DYNO provides is a -- is a platform for customers to do that by bringing in data about positions of all these nodes, about their capabilities, any constraints -- constraints that they have --

>> Maria Varmazis: Yes.

>> Ed Meletyan: -- DYNO will orchestrate their motion, their connectivity, and build the system for you. And what that means is, if, for example, you have a satellite constellation, all these nodes are -- in the constellation are linked to each other via either optical links or RF links, and the operator has questions. What is my latency from node A to node B? Node A might be an aircraft and node B is a -- a satellite. What happens if one of my nodes goes down? How does my network self heal?

>> Maria Varmazis: Yes.

>> Ed Meletyan: How do I reroute around a specific geographic region?

>> Maria Varmazis: Yes.

>> Ed Meletyan: None of these nodes know about the nodes that they're connected to. So, it's important for someone to understand the -- the system holistically, to be able to take inputs from an operational environment and then be able to optimize the digital twin and from that, optimize their operational system.

>> Maria Varmazis: Okay.

>> Ed Meletyan: So, when a node says, "Hey --" is sending back unhealthy signals or not communicating, you're updating your digital twin to say, "Take this node out and send me new routes that I can proliferate through the rest of the network." And then sending those commands out to -- to everything so things know what they're going to connect to and when.

>> Maria Varmazis: That sounds like it would also save a lot of time for the customer, just listening to that. Yes.

>> Ed Meletyan: Yes, exactly right. For example, they don't need to -- it'll save time in terms of I know how to optimize which data path a specific maybe imagery collection needs to take. And I can orient my network in a way to optimize that path.

>> Maria Varmazis: Yes, yes. It makes a lot of sense. All right, now I'm going to move now to what happens with the data once it's acquired? So, Dewey, this is -- this is sort of -- turning to you now. So, we're talking about OSML. Tell me a bit about that.

>> William Duhe: So, Oversight ML, short OSML, is what happens at the very end of this pipeline that we're discussing here amongst the three of us. We have OrbitIQ and DYNO, which are helping to sustain satellites in space, manage them, keep them from running into other objects. But what happens when we get this data down to the planet and we want to orchestrate the release, dissemination, analysis of that data at scale? In the last five or ten years, there's been a proliferation of data that's being made available, increasing resolutions, increasing complexities, increasing formats. And customers have had a challenge to adopt the cloud as the place at which they do all the processing for this imagery. Processing can include normalization, release of that data to specific customers, the application of AIML, which we'll be kind of focusing on in these demonstrations here with Oversight ML. Oversight ML is a broad framework of solutions and tools that are all open source. We provide them through GitHub --

>> Maria Varmazis: Oh, cool.

>> William Duhe: -- for customers to download and use in order to increase their velocity in adopting the cloud as the ecosystem they use for processing imagery. Model Runner, which is our focal tool of this conference, specifically empowers customers to process overhead imagery against computer vision models efficiently at scale. When you have thousands, tens of thousands of high-resolution images coming in from these satellites, how do you manage that data flow of taking an image and getting insight out of it? If we look back to the '50s, '60s, '70s, when we were first starting the process of collecting overhead imagery, we had hand analysts going in and circling object of interest, recording that in a table, a notebook, a database, sharing it with their colleagues and then making strategic decisions based off of that.

>> Maria Varmazis: Right. Yes.

>> William Duhe: With the advent of computer vision and AIML, all of that's being automated away. How do we efficiently take these behemoth sized images and distribute them to computer vision so that we can go from image to insight in a velocity that makes sense for our customers?

>> Maria Varmazis: That -- the fact that that is automated is -- I mean sort of table -- I -- I mean, it's just amazing to me that we're -- that is where we are at. The insights that have to be delivered from that are just a -- and the speed at which it's coming, it's just -- it's got to be game changing.

>> William Duhe: It really is. And especially when you consider National Defense agencies, governments around the world, people that want to respond to natural disasters.

>> Maria Varmazis: Yes.

>> William Duhe: And sometimes going to insights in minutes or hours --

>> Maria Varmazis: Yes.

>> William Duhe: -- makes all the difference in an organization's ability to effectively respond to a situation.

>> Maria Varmazis: That's quite amazing. All right. Well, gentlemen, we've -- we've talked about sort of each of the -- the components here. And I'm wondering if you can give me sort of a sense of like a -- a through line for the framework here, just to -- to make sure that you know, we're sort of aligned here. We can give that a shot.

>> Ed Meletyan: Yes, sure. I mean this is a modular framework. So, all these solutions kind of exist on their own and they provide their own value. But here we're bringing them all together to show how customers of ours can -- can do space management at scale and Orbit IQ is kind of where it all starts. This is the -- a foundational capability that almost every space mission needs to have knowing where satellites are to a very high degree of accuracy.

>> Maria Varmazis: Yes.

>> Ed Meletyan: And it's not just the satellites too, right? Like Junk mentioned, there's the ground station contacts. Like when can I talk to my -- to my assets? The quantified uncertainty is a big piece of that. It also extends to space situational awareness, knowing where everything else is and -- and so, just about everything is going to use that Orbit IQ piece.

>> Maria Varmazis: Yes.

>> Junk Wilson: Tell me if I'm wrong, but I'm betting for DYNO, you really want to know where every node in your network is, right?

>> Ed Meletyan: That's exactly right. So, when DYNO comes in and we have thousands of assets that we're modeling, those uncertainties stack up, right? So, we need that high fidelity propagation to make sure our ground contacts are optimized. We're not losing any time there. If things are drifting out of position and maybe we aren't getting a contact that we expect, is something nefarious happening or -- or something else? And Orbit IQ provides that -- that backbone. And as customers of ours who operate many different Earth observation satellites, as they optimize their fleets using DYNO, for example, we have a use case of, "How do we image every city on Earth every hour?" Well, what do we do with those images is the next -- the next question, right? And that's where OSML comes in is --

>> Maria Varmazis: Yes.

>> Ed Meletyan: -- once we're piping all that data through the inner satellite link constellations and sending that out to operators, we need the capability to do inference at scale in a very reliable way. And OSML is, especially the model runner piece, it's very good at that. So, customers have a much easier time getting value and insights out of the imagery that they collect.

>> Maria Varmazis: Fantastic. Well, thank you. Thank you for tackling that question. I know that was a -- it's a -- it can be a tricky one to answer. And since this is a WS in orbit, I have to ask because I am genuinely also curious. When we think about each of these components working on their own or also In -- in coordination with each other, what do they enable customers to do in using the cloud and AI with AWS? I don't know if anyone wants to -- who wants to tackle that one? Yes, Junk, you want to take that one? Yes.

>> Junk Wilson: I can at least start from our perspective. It is -- it is uncommon to have a team of astrophysicists like we have at -- at Orion and have that be -- have that group also be present in a satellite owner operator. Normally, satellite owner operators are a bunch of very talented engineers. Those are very different skill sets. So, I think what this overall package provides satellite owner operators is this, you know, that classic question of do you build it or do you buy it? It is possible like some other satellite company could go out and -- and hire a bunch of PhDs in astrophysics. Pay them their annual salary --

>> Maria Varmazis: I was going to say, notoriously cheap, right?

>> Junk Wilson: -- yes. You know, spend a couple of years developing and testing and validating all of these models. Or I think what AWS offers now is, "Hey, this is now a ready to go package solution. Buy it off the shelf. Because it's packaged by AWS, install it right in your own personal AWS account and feed it what you want to feed it." So, you could -- you could hook it up to Internet data. You could hook it up to your own satellite data if you have your own small constellation of satellites. You could run it inside your own little protected network, if you're worried about, you know, data security. You have an infinite number of options there because AWS gives you all those options. And this is one of the solutions that seamlessly plugs into that architecture.

>> Maria Varmazis: It's very cool. Anyone -- yes --

>> William Duhe: And if I could expand a little bit --

>> Maria Varmazis: -- please?

>> William Duhe: -- on what Junk is saying here, I think that scale is one of the focal points of what our customers need as we begin to expand what we do in space. When we look at the last 20 years, the amount of components and hardware and devices that we have in space, has gone up exponentially.

>> Maria Varmazis: Yes.

>> William Duhe: Space is quickly becoming crowded, I think is a phrase that I've heard --

>> Maria Varmazis: A lot.

>> William Duhe: -- in the industry, right?

>> Maria Varmazis: Yes, yes.

>> Junk Wilson: Well, one of my favorite analogies for that is more mass has been launched in the last five years than had ever been launched in the history of humanity up until five years ago. That's an amazing --

>> Maria Varmazis: Yes.

>> Junk Wilson: -- thing to think about.

>> William Duhe: It really is. And as that scale increases for the amount of hardware, collection and data that we have available for our customers, how do they operate in a way that allows them to scale as those continue to increase for the next two, five, ten years, as the amount of hardware increases required to do the compute for the sophisticated orbital modeling, for the dynamic traffic simulation, for the data analysis on the back end? AWS is kind of the king of scale in the market. And by plugging into these types of ecosystems, it allows customers to adapt to a rapidly evolving techno landscape, which is space operations.

>> Maria Varmazis: Yes, absolutely.

>> Ed Meletyan: Yes. In terms of the orchestration piece too, AWS provides a lot of flexibility, so customers can bring tools that they're familiar with that they can let AWS handle the orchestration of things that most customers don't make money off of in terms of, you know, IT operations and scaling and having to figure out, you know, how much hardware do I need? How much staff do I need to maintain that hardware? AWS kind of handles all of that, right? And so, if you bring on a new mission, or maybe you have new customers asking for your data, you don't have to go back to the drawing board and figure out, "Well, how much -- how many more servers do I need?" These serverless offerings that we -- we offer, they -- they scale on demand. So, if you need more capacity, you just -- you get it and -- and you don't have to worry about it.

>> Junk Wilson: And -- and I can speak to this too, because not only am I part of this team here, but Orion is a customer of AWS.

>> Maria Varmazis: Right, right, right, right.

>> Junk Wilson: You know, I can speak to the fact that Orion doesn't want to get into the business of cloud computing. That's not what we are best at --

>> Maria Varmazis: Why not? No, I'm kidding.

>> Junk Wilson: -- right? Like we're astrophysicists. That's hard enough. I don't need to be, you know, a bunch of computer science guys too. I would much rather use the power of AWS and take all of that off my hands.

>> Ed Meletyan: Not to mention the economics around how quickly is this hardware going to depreciate?

>> Maria Varmazis: Right.

>> Junk Wilson: Oh, definitely.

>> Ed Meletyan: How quickly am I going to need to buy the next generation of GPUs, the next generation of compute processors, FPGAs, all these different --

>> Junk Wilson: Oh, yes.

>> Ed Meletyan: -- complex and quickly evolving hardware that we're using in space operations? AWS offers a framework for implementing all of these on the fly.

>> Maria Varmazis: That's fantastic. Gentlemen, given that you all have such varied careers and perspectives on the space industry, I -- I just have to ask, what are you most excited about when you look towards the future of what's coming? I mean, even just the capabilities that you all have described to me today are things that I think five years ago I would not have believed that we were doing.

>> Junk Wilson: Yes.

>> Maria Varmazis: So, as you look forward, what -- what's got you jazzed? What are you thinking about? I'll start left to right. Yes.

>> Junk Wilson: Oh, yes, I've got a lot. I'm a total space nerd.

>> Maria Varmazis: Where to start, right?

>> Junk Wilson: Yes. So, I'll tell you one of the things that gets me really like, I can't wait to see this is when we get commercial suborbital rocket travel. Like how cool is that going to be?

>> Ed Meletyan: I guess for me it's the -- the connectivity that's always increasing. So, as we continue to launch new -- new satellites that provide connectivity to very remote places on Earth and guaranteed communications to the rest of the world, I think -- that's really exciting to me. I -- I camp a lot, so when I'm out in, you know, in the forest and maybe I need to get a topographical map, it'll be really nice to go and find one on the Internet when I need it.

>> William Duhe: And so, my interest in space and where I think it's going and where it's going to impact humanity is a very cross-intersection with philosophy and our understanding with our place in the universe. It's great to think about all these practical applications that space is going to offer us as we begin to be able to, for instance, ride rockets to the beach on Saturday --

>> Junk Wilson: Oh, yes.

>> William Duhe: -- or, you know, communicate with one another more efficiently and quickly.

>> Maria Varmazis: Well, gentlemen, it's been a pleasure. Thank you so much for speaking with me today. I appreciate it.

[ Music ]

That's it for this episode of "AWS in Orbit" by N2K Space. We'd love to know what you think of this podcast. You can e-mail us at space@n2k.com or submit the survey in the Show Notes. Your feedback ensures that we deliver the information that keeps you a step ahead in the rapidly changing space industry. This episode was produced by Laura Barber for AWS Aerospace and Satellite, and by N2K Producer Liz Stokes and Senior Producer Alice Carruth. Mixing by Elliott Peltzman and Tre Hester, with original music and sound design by Elliott Peltzman. Our Executive Producer is Jennifer Eiben. Our Publisher is Peter Kilpe. And I've been your host, Maria Varmazis. Thank you for joining us.

[ Music ]