Back
Episode 9

Autonomous Spacecraft: The Role of Software in Shaping the Future of Space Exploration

Space hardware is getting a software upgrade – and it’s changing the rules.

Podcast Space

In this episode, Mikkel Svold speaks with aerospace lecturer Andreas Stren and RF engineer Alexander Spaniol from Terma about how configurable tech like FPGAs is giving satellites a second life.

Smarter systems mean in-mission updates, greater resilience, and real-time adaptability in unpredictable environments.

For those tracking the edge of space infrastructure – from mission ops to defense contractors – this conversation cuts to the core of what's next in agile satellite design.

This is about what’s working, what’s coming, and why it matters.

In this episode, you'll learn about:

  1. Discover how software redefines space tech design and functionality.
  2. Understand FPGA's role in software-driven space applications.
  3. Learn the shift from hardware to software-defined architectures.
  4. Explore mission agility with software reconfigurability onboard.
  5. Uncover how AI enhances satellite autonomous decision-making.
  6. Future vision: virtualization and onboard data processing evolution.

Episode Content

00:11 Introduction to software's evolution in space technology
01:06 Historical shifts: From analog systems to software-driven solutions
03:29 Role of software-defined architectures in space systems
04:53 Hardware requirements in harsh space environments
07:24 Enhancements in mission planning and satellite lifespan
10:52 Implications of increased processing power for satellites
13:28 Benefits of software-defined systems for mission adaptability
17:56 Automation and AI in satellite decision-making processes
22:13 Influence of new space actors on software advancements
25:21 Future developments: Virtualization and onboard data processing
28:50 Practical applications: Wildfire detection and autonomous landings

Production

This podcast is brought to you by Terma.
This podcast is produced by Montanus.

Episode Transcript

Mikkel Svold (00:11): Yesterday, I looked up at the stars and what struck me was a little set of stars, it looked like a set of stars that were moving in line. And those of course being satellites. And I thought to myself, this is really part of this whole new space era where we suddenly look up at the stars and we see satellites. Satellites that we couldn't see before just I would say, what, five, 10 years ago, they were not part of what we could see. And that is actually what we're talking about today, this new way of doing space technology. Today, we are talking about software in space technology and how software and the approach to software is now changing how space tech is looking and how it's made and what's it's all about. And in the studio or on an online link I have with us Andreas Stren, welcome to you Andreas.

Andreas Stren (01:05): Thank you.

Mikkel Svold (01:06): And you, I have you down as the lecturer of aerospace engineering program at the University of Applied Sciences in Wiener Neustadt. And with us also we have Alexander Spaniol who is the radio frequency engineer here at Terma, of course in the space department. Welcome to you Alexander.

Alexander Spaniol (01:23): Thank you. Hello.

Mikkel Svold (01:26): I want to start out, I warned you about this before starting the recording, so I want to start out by asking you how have the approach to software, how has that changed over, well, let's say the last 20 years from, well, I was just about to say the first missions, that's a little bit more than 20 years ago, but how has software in space tech changed? Andreas?

Andreas Stren (01:52): It changed mainly that more and more tasks are taken over by the software itself, that were formally executed by analog parts for example. And together with this, not only the complexity of the software has risen, but also the possibilities we can have. For example, that we update something in the software during the mission. It isn't that we can go up there and change the hardware, for example, in the satellite, it's pretty hard.

Mikkel Svold (02:41): Just for me to understand, can you give an example of what that could be?

Andreas Stren (02:48): For example, I think we have also time to get a little bit deeper later on, for example, there is something called field programmable arrays. They can be updated with a software, and internally they change their configuration. So the same part has another function just by software update, that is an example.

Mikkel Svold (03:17): Okay. And how does it really work because you still need some hardware? Alexander, how does it work now?

Alexander Spaniol (03:29): I think Andreas already got the point quite well. So with this new space era, this prioritization of the space sector, what we can clearly see that we need new products, another product with a specific use case, it somehow must be such a platform with a general purpose and so on. And that cannot be done with old hardware that is very restrictive to one specific use case. We need new ways. And nowadays we often talking about this software-defined everything or a special thing for this software-defined architectures.

(04:06): And that is one of the developments where we can see that the market goes and where customer requests us to deliver on this. So having, for example, this FPGAs, as Andreas already mentioned, is we have software which we can program, for example, this hardware piece. And doing this, we can do many different things just by using software or writing the software for it, and not just changing hardware, doing some physical work, it's all done digitally.

Mikkel Svold (04:45): But how do you create the hardware then because you still need hardware to run the software on?

Alexander Spaniol (04:53): That's the specific thing of FPGAs, it's a I should say a very, very complex thing. But I think that Andreas can talk a little bit about that, he had some experience with that. I'm more like the use of it. And Andreas already did some programming directly on the FPGAs.

Andreas Stren (05:19): Well, about this, we speak on a level where it is very close to the hardware, so it's also called embedded programming where we use integrated circuits that have a certain task or a certain function. And these new programmable arrays can change their position or state of transistors within the computer chip just by defining it by software. So that's actually a hardware change due to a program that is uploaded.

Alexander Spaniol (06:08): And maybe just to get the point to now cutting edge software technologies, how to influence our space industry today. A very well-known application for that is, for example, software-defined radio, which is part of software-defined architectures, where we combine this reconfigurable hardware with very powerful software. And that's the key point and why it works out. And that gives us many advantages. We have very flexible systems that can be used for several different things just by changing the software. We have reductions in costs because we are no longer dependent on a specific hardware that can just be used for one specific case. We have programmable hardware which we can control via software. And what we actually then do is we program our software to make the thing doing that what we want to do it.

Mikkel Svold (07:18): And what sorts of requirements is now set to the hardware on board?

Alexander Spaniol (07:24): On board of the satellite or let's say on an application? Yeah. Regarding the hardware or if we're now talking about software-defined radios or software-defined radios that are on a satellite, then we have to think about how we can withstand, with this hardware, this reconfigurable hardware, the atmosphere. We have a very harsh environment, we have radiation, we have high temperature fluctuations. And we have to manage that this little piece of hardware that we have that survives this atmosphere in order to get our software running on that. And Andreas, you can for sure maybe speak about what this hardware needs to withstand there out in the orbit.

Andreas Stren (08:16): Well, space is environmentally difficult to handle, not only that we can't just send a service team, but also often the operation time of such satellites can hit decades without actually maintenance. And therefore the systems that we can ... if we can control the systems easier by uploading for a new program, for example, the satellite itself can be maintained.

(08:57): And specifically on the environment, we have radiation, that's the most important impact for digital devices. But we also have thermal fluctuations, then we need to consider power needs. We need to consider that the power configuration must survive, also the radiation and temperature. And the combination of these requirements is hard to handle in terms of also how do subsystems speak to each other. For example, if a subsystem has a more capable software and chip, for example, then it can also act individually without being restricted to one or two commands that needs to be from a central computer. So there is also another advantage that evolves out of it, that each subsystem can handle itself more easily if the capability of the software rises.

Mikkel Svold (10:25): I was just thinking, when you have larger possibilities with software, you also require, say, processing power, you require more powerful systems on board. Is that a problem or how does this new, I'm saying new, how old is it, the software first kind of approach?

Alexander Spaniol (10:52): The software-defined, just to maybe to get the answer for the last question, software-defined architecture, software-defined radio and so on, that's already, it's not new, we have that for let's say several years or decades. But the actual application to run these systems with advantages over hardware defined systems, is there the test is now coming and can be used. And therefore we nowadays see the applications for that what was already developed in the past. So this idea already is quite old let's say.

Mikkel Svold (11:25): And what does its set of requirements to the build of, say, the satellite, does it require a larger power output?

Andreas Stren (11:34): Yes, definitely. I can go on this one. Definitely a larger power need. And it must be secured. So power is the most important thing for a satellite. If the power goes out, your mission is lost. And the second thing is communication, to actually interact with the satellite. And everything else comes afterwards.

(12:04): So also there you already see that the control of power management and the control of communication is the most important to also improve and make fail-safe. And therefore this more powerful systems as, for example, an FPGA then is capable of more redundancy I would say. For example, if the integrated circuit has a problem, we can reconfigure it and then the problem is solved. So the increasing requirements go hand in hand with increasing redundancy and also the fail-safe I would say.

Mikkel Svold (13:06): When we are looking at the mission planning prior to a space mission, what does it mean now that we have different possibilities just by, I'm saying just by, I know it's harder than it probably seems, but just by updating the software, what does that mean to ... I'm guessing it means something to the lifespan of the satellite?

Alexander Spaniol (13:28): So the software-defined architectures, when we specifically look at the mission planning, there we can see the software has transformed how space missions are planned. The software make them more agile, data-driven and of course also efficient.

(13:51): And specifically the last point you have mentioned about the lifespan, since with software or when you have this software-defined approach or maybe also software-defined satellite, where most of the systems rely as much as possible on software, and we have the great possibility that we can react on certain events that happen in the orbit, can happen. And we can update and change our system to what we are required to. If we have a problem, for example, with the communication on that specific frequency or if that specific modulation type and so on, we are actually able by updating the software to change that. And that for example was in the past not possible with hardware-

Mikkel Svold (14:41): In the past, would you suddenly lose communication to a satellite and then mission is just gone?

Alexander Spaniol (14:47): Yeah, if you, for example, cannot change the frequency because it's just programmed into your hardware model, then the mission for sure can be lost.

Andreas Stren (15:04): I can also tell about the process to build a mission changed. Okay. So when you speak about mission planning, it's not only the moment the satellite is up there, so also the approach how to design the actual mission has changed. An example would be that the mission requirements define a task that the satellite has to do, for instance observation, communication, name it. And during the design process, you always had to consider, okay, which ground station am I using? Or which network, do I have to build an own ground station? And for example, if you build a ground station on your own, do you only tune it on that specific communication protocol?

(16:09): And with the SDI technology, we can actually, as space engineers, we can actually reduce this requirements and concentrate more on the actual mission of the satellite because we have more versatility.

Mikkel Svold (16:28): And you can also reserve more resource I'm guessing, to the actual goal of the satellite. Do you think this ... Yes, Alexander?

Alexander Spaniol (16:39): Maybe just a point to that, speaking about resources and so on. Having everything in software allows us to for sure also do automated mission design. So we have a software driven systems that help us on automating many aspects of the mission design. How we react on how we change the orbit, how we plan to do our trajectory, how do we configure our spacecraft, and especially also maybe regarding risk management.

Mikkel Svold (17:12): So basically setting up that if this event happens, then do this automatically.

Alexander Spaniol (17:18): Yes, that's possible with software. So we can automate processes where we don't need this, let's say, human power or human that decides how to do it. That can be automated now. And that for sure, since you mentioned it, saves resources and time.

Mikkel Svold (17:39): And just, I know that this is not, it's not necessarily or it's probably something fairly new, but this automation of processes, is that also a door into autonomous decision-making at the satellite itself?

Alexander Spaniol (17:56): Is it? Nowadays, everyone already have heard it for sure, that it's done by AI, so artificial intelligence, and also for machine learning. And with this, we are increasingly used to automate these tasks. So actually we have an onboard AI that makes the decision making for us. And at the point where it is required, if we, human, need to do the decision making for the satellite up there, we really have a problem. Because we need to get a communication window because the satellite now needs to be above a certain point above the Earth where we have to link with the ground station. And then we have to tell via an uplink the satellite body needs to do. And we can only do that when we receive the information prior.

Mikkel Svold (18:47): I think that problem is something that's very obvious to people working in the space industry. But it's something that it kind of baffles me every time. The fact that you have to wait for your satellite to come back around, to orbit Earth, and then once it's, maybe not directly above, but it's when it's within that kind of visible window or what you call it, then you can say something.

Alexander Spaniol (19:14): When it's over the horizon.

Mikkel Svold (19:16): When it's within the horizon. And then when it's outside again, well, you have to wait for, what, how long time does it take, I don't know, it depends.

Andreas Stren (19:27): Here it also depends on your mission or communication architecture. There are several types of that. Maybe you've at some point heard geostationary, that would be a different approach to swarms of satellites that are very close to Earth. And therefore the orbit design also defines your communication architecture. So if I, for example, want to have communication with a satellite that is outside line of sight, I either need another ground station that I connect to, or we can, for example, relay the signal via different satellite and make a sort of chain.

Alexander Spaniol (20:25): That would require a constellation so you need more than one satellite.

Mikkel Svold (20:29): So that's what SpaceX would be able to do with their-

Alexander Spaniol (20:35): Yeah.

Andreas Stren (20:35): Yes. They have an interconnected close range common application in between each satellite. And if you take thousands of them, you can reach satellite number 500 at any time because it relays over several others.

Alexander Spaniol (20:56): And that is controlled using AI, that cannot be done by humans, it's not possible.

Mikkel Svold (21:03): So that linking, and it's actually, I'm guessing, that's not even hard AI, that's what ... Google Maps solved that kind of AI a long time ago. And even that algorithm is probably, what, it's probably from the forties or something like that. So that's just a shortest distance kind of algorithm, right?

Andreas Stren (21:20): But also it raises respect for the people who built this in the first place.

Mikkel Svold (21:26): It does, it does, absolutely, yeah, yeah. I can't but think, but in your opinion, do you think that this push for software-defined radio, is that a natural thing, is that a natural development? Would it have happened anyway? Or is that the result of more private actors entering the space market where you have not, well, well, I'm guessing that Elon Musk and Jeff Bezos, they more or less have unlimited resource. But let's just say that it's not unlimited resource, someone, you can't just keep on spending. Does this push come from there mostly or would it have happened anyway?

Alexander Spaniol (22:13): I think there are two reasons for that. One is for sure this development to this new space era with the privatization and so on. Where we definitely look more into how much money we actually spend on the system. That is one point.

(22:29): And the other point I think is we are always interested in advancing our technologies. And of course if we have a technology that works for a specific use case and we have the opportunity to advance it in a way that it works for several use cases just by reconfiguring it, then we for sure do that. And we also would've done it in the past. But there the tech wasn't here. But nowadays with the technology we have, with the FPGA networks on the ground with high speeds, low latency, we are now finally really able to get this technology working. And that is, from my point of view, the second point where we now have this breakthrough and this push to software-defined everything.

Mikkel Svold (23:13): I'm also guessing that these missions that are not near Earth but the deep space missions, they're actually only meaningful if you have some sort of autonomous decision-making at the satellite because you can't wait for two days to answer because there's no-

Andreas Stren (23:34): Well, in most cases we just do.

Mikkel Svold (23:37): You just wait, yeah, you just wait. You send out the signal to whatever satellite that is that's on its way out of the ... and then you just wait.

Andreas Stren (23:45): Yes. Yes. Well, if you're also talking about new space here, in old space or big space, they're not really terms, but to think about it, it was like, yes, if you want to call a satellite that's 40 light minutes away, the answer will take 80 minutes.

Mikkel Svold (24:06): Yeah, because you need the communication to first get out there, you'll have it do its job and then send back that, yeah, confirmed or ...

Andreas Stren (24:14): Yes, with the first few Mars rovers, it was exactly like that. That someone in the operation room, for example, sees a stone on the surface, tries to navigate around, makes an input, and 40 minutes later there was like, "Oh, okay, did I hit it or did I miss it?"

Mikkel Svold (24:35): Whereas that now it will just, well, drive around.

Andreas Stren (24:40): Yes, and we can watch the decisions afterwards and make adjustments to them.

Mikkel Svold (24:49): And teach them new things, how do we want you to react next time.

Andreas Stren (24:53): Yes, exactly.

Mikkel Svold (24:55): And that of course both goes for Mars rovers, which is really cool, but also satellites and, well, telescopes and everything. I'm wondering, what do you see coming ahead, what's the new, what's on the horizon development-wise in software-defined, well, not software-defined radio, but software in space? And then we'll deep dive into software-defined radio in the next episode.

Alexander Spaniol (25:21): I think that the next big thing or one application for having this software-defined approach is now this big term of virtualization. Meaning that we have a small hardware like, for example, we already talked about it several time, a software-defined radio device, which gets then the analog waveform of the signal I receive from a satellite. And then I directly more or less directly digitize it and then I already have it in the digital domain. So I virtualized it. And using that and by having it in, let's say for example, hosted in the cloud, we can access this information we got from the satellite, let's say, more or less from everywhere. And that's the big topic I see or we see where we have development going to. So this specific application for this virtual software-based systems.

Andreas Stren (26:23): Also developments I've seen that are already ongoing, and if we speak about the future is, for example, the virtualization connected to the capability of the satellite, that the satellite can interpret data on its own. So in the past there was always, okay, we get the data, we get it down to Earth, we take a look at it and interpret it. And all these processes can be outsourced to the satellite itself so that actually the post-processing can be handled on the satellite itself before even the data reaches Earth.

(27:08): Where it is implemented already is wildfire detection, for example. That the image the satellite takes can be processed onboard. And if it detects a wildfire, can give a warning signal even though the data hasn't even arrived at any human interface.

Mikkel Svold (27:29): Yeah, I was just about to say, what are the practical usages of that? Why would you want to throw the processing power up in the air so to say or up in space where you can't upgrade the processor or you can't upgrade the hardware at all, why would you want to do that, what's the benefit?

Andreas Stren (27:52): There are two that I can speak of. Wildfires I already mentioned. Many of these applications concentrate on image observation, for example. Earth observation is the top topic. But also for example, if you want to land something on another celestial body, as we said, the latency of communication signal leaves no time to actually react something that happens during a landing. So the interpretation of data from the spacecraft itself to react to the actual situation right now, that is one of the major points that are really improved by now and will improve.

Alexander Spaniol (28:50): Maybe also something about the self-observation thing and this wildfire, why just throw up the processing power of this up to the satellite. We also have to think that these images are high-resolution images and they have a vast amount of data. And we have to really transfer this data down to Earth and then process it there. And that really can be challenging with very high data rates, very high bandwidths. We may also have problems with adjacent signals that influence our signal quality. And we may, for example, not be able to get a link or get the full image that we would require in order to further process it on Earth. And if we directly do that on the satellite, and then just give the result down to Earth, we are only speaking of one hundreds of thousands, whatever, just those numbers, what you have to send down to Earth, for example, in that case. That is one thing.

Mikkel Svold (29:47): It's basically the exact same problem as when I upload, well, this video to YouTube, this is probably going to be like a 50 gigabyte video, so if my connection is just a little bit off, it will take forever or crash. And that's the same thing.

Alexander Spaniol (30:04): Yeah. We are speaking here of data rates for a gigabit, one gigabit, for example, let's say 100 mbit.

Mikkel Svold (30:08): Wow. Okay.

Alexander Spaniol (30:09): But with a satellite link, I don't know, if you have an high data rate model, you may be able to achieve 100 mbit per second. But normally we speak about eight mbit, let's see. And there you will really have to wait a long time to upload gigabytes.

Mikkel Svold (30:30): Our time is up, but in the next episode we'll talk more into this software, we'll talk more about software-defined radio, more deep dive into that particular type. And I think that's going to be very exciting as well. Also, probably very geeky, just a warning sign for everyone out there. But I'm looking forward to it anyway.

(30:51): Andreas Stren and Alexander Spaniol, thank you so much for joining this conversation. And I look forward to talking to you again in the next episode.

Andreas Stren (31:00): Thank you.

Alexander Spaniol (31:00): Thank you.

Mikkel Svold (31:02): And of course to you listeners out there, thank you so much for joining this conversation. And if you have any questions or any topics that you want us to dive into, just reach out to us on podcast@terma.com. That was podcast@terma.com.

(31:17): And of course, if you liked this episode and want more of it, just hit that subscribe button and a little bell notification always helps a little bit. And of course, share it with your friends and colleagues or whoever might be interested as well. That really helps us quite a lot in spreading the word and spreading the knowledge, which is obviously the point of us talking here at all.

(31:37): So I think with that, all I have left to say is thank you so much for listening.

Andreas Stren (31:44): Nice to be here. Thanks.

Alexander Spaniol (31:45): Bye.