Global Trends in Space: Innovation in the Age of Digital Engineering

Hello, everyone. So as I think about the space industry, if I were asked to summarize it in one word, I think that word would have to be innovation in this day and age.

And if I were then kind of peer behind that innovation and look at what's allowing it to happen, I would come up with digital engineering, which is a pervasive and accelerating technology trend in the last decade. My name is Ossi Saarela, and I'll be talking about global trends in space. So this talk will examine three broad topics.

First, I'm going to talk about the private space sector. And what I mean about the private space sector is the privately funded space segment, which has been such a major innovation driver in our industry that it makes sense to address it separately from government programs, which we'll touch upon next.

And finally, I'm going to talk about the current technology trends with a focus on digital engineering, as I mentioned before. And also treating a subset of it, AI and machine learning, which are emerging in our industry, separately.

So the big story of our generation really and the space industry are space startups. This video shows the first launch into space with a vertical soft landing. That was accomplished by Blue Origin in 2015. And SpaceX, of course, has since become famous for it. They've gone orbital with it, they've operationalized it, they've made it a big part of their business strategy.

But reusability isn't the only new technology to have come out in the last 10 years. We've also seen the emergence of satellite mega-constellations. And we've seen the privatization of technologies that formerly only large governments could do. I'm talking about things like synthetic aperture radar on smaller private satellites.

We've also seen some underlying development technologies go mainstream that are newer. For example, automated code generation for spacecraft flight software. I know of several human programs in development, including NASA's Orion capsule, which are utilizing automated generation for portions of their flight software.

I also know of multiple launch vehicles which are either in development or recently completed development that use automated code generation, which is-- it says something given that the launch vehicle sector is traditionally the most conservative portion of our industry.

We've seen some new attitudes brought into the industry by these startups. They're willing to take more risk. They're willing to do things that we kind of got away from for many decades, like using commercial off-the-shelf hardware, commercial grade hardware to fly their missions.

And they've also brought whole new ways to operate into the industry. Now traditionally, if you wanted to purchase a space-based asset, you'd buy one or maybe two satellites. And in order to protect that precious asset, you design things like redundancy on it. And then you do a lot of analysis for survivability, and radiation, and things like that. And that of course, added up in terms of the development cost of the satellite.

Now some of these newer companies, like Planet, for example, formerly Planet Labs, have kind of turned this on their heads. And they have enabled the use of much, much simpler and less reliable satellites, where now the reliability of the system is reliant on the constellation, right?

So these companies have developed a very good feel for what the average lifetime of their spacecraft is. And if it fails, they'll just launch another one. And this enables dramatically lower cost access to some of these space-based systems.

Now what kind of private investments does it take for these startups to do the kind of things they do? I'm going to start by looking at 2019, which is our last prepandemic year to kind of get an idea. And we can see that the amount pouring into private space companies was accelerating throughout the 2010s.

Now this chart does a little bit of a mind experiment that I like to play, where I imagine all of the private space investors and kind of what the size of their investments would look like if they were all under the umbrella of being their own space agencies. And of course they're not, but this at least allows us to take a look at how much money private space companies around the world are receiving compared to government agency investments.

And if we play this game, we can see that in 2019, private space investment, if it was a space agency, would have been the third most well-funded agency in the world. So this is real money. This is significant funding that these companies have begun to receive.

Now those numbers were for 2019. So of course we need to ask ourselves-- in today's age with a pandemic-- what happened in 2020? Well, in the first half of 2020, satellite and launch vehicle infrastructure development investments plummeted, as can probably be expected. And by infrastructure investments, I mean funding into the actual launch vehicles the rockets and the satellites themselves.

Now at the same time, however, investments into the services that the space industry provides increased dramatically. And if you think about it, that makes perfect sense. During the pandemic, we started ordering our groceries, our meals, and virtually all of our shopping using apps on our phone. And those apps all use localization.

And the on-demand delivery services use satellite navigation systems. And on top of that, demand on communications networks is up because all of our meetings now are virtual from home. So there's this huge demand on services enabled in part by the space industry.

And as a result, if you look at the combined growth of infrastructure and applications, we actually saw a positive number-- 4% in 2020 H1. And in a way, this confirmed the market demand for the services that the infrastructure provides. So maybe it's no surprise that infrastructure spending saw a strong rebound in the second half of 2020. In fact, strong enough that 2020 set the record for private infrastructure spending despite the pandemic, at almost $9 billion US.

Now in 2020, this was mainly led by investment rounds in two countries, the US-- and SpaceX Blue Origin and Relativity Space were the largest recipients there. But also strong investment rounds in China to companies like LandSpace and i-Space for launch, and Changwon Satellite Company.

So as you can see, the story is actually better than expected in terms of the market for private companies. Now what about government agencies? How are those doing?

Well, government interest in space remains very strong. Much of it is actually focused on human space flight these days. Now in some cases, that investment remains strong through the pandemic. China is a good example of that.

In others, we saw that governments actually were not quite as resilient as the private sector, surprisingly, to the pandemic. India, for example delayed their space program investments and funneled that money into the pandemic response instead. In the US, I think government interest remains strong. But the funding for the Artemis program was not as robust as many had hoped for.

And consequently, when the human landing system awards were released just a brief while ago, instead of multiple winners, which I had expected-- I had expected NASA to pick two winners-- they're only able to select one company, SpaceX, to move forward. So we're seeing a budgetary impact in the US, whether it's entirely pandemic related or not can be arguable. But I think overall, government interest globally remains strong.

Now I think it's important to note that when we think about the space industry, you can't really think of it as government versus private. There are some media articles that like to portray these two subsegments kind of in competition with each other.

But I think largely that's not true. And a good example would be NASA and SpaceX working closely together on the commercial partner model for-- on the commercial crew vehicle shown here in this picture with the Crew Dragon.

Likewise in China, we've seen substantial private sector development where the government is encouraging private company growth and innovation. And India has also taken steps in that direction just in the last year or so. So we have this model of government agencies, especially NASA, transitioning from a design establishment, if you will, to becoming more of a nurturer of private sector development and innovation, while, of course, doing a lot of the higher risk research and development themselves.

And this is a model that the European Space Agency of course has employed for quite a while. But it's an interesting landscape for emerging nation states, like Australia and the United Arab Emirates, that are developing a space capability, because now there's this option of either kind of going at the government route or using government as a nurturer of innovation of a robust private segment. And I think there's success to be had through both models.

Which brings us to technology trends. I'm going to argue that digital transformation is a common trend between private and public players. And one of the primary reasons that digital engineering has accelerated in the last 5 or 10 years is because these private space companies that are unencumbered by old ways of doing things have established strong digital engineering practices and now are bringing them back to the public sector.

So we know companies and governments are investing in digital engineering, but what does that actually mean? It's a buzzword that seems to get used a lot, but is defined less frequently. So I'm going to use a survey performed by the Boston Consulting Group to break down the concept of digital engineering a little bit further.

So what the survey did is it interviewed several hundred aerospace executives from around the world, and asked them which technologies they were actually investing in as part of their digital transformation initiatives. You can see number one, 3D printing for prototyping. The space industry is a strong leader in the 3D printing area. In fact, there's 3D printed parts on the Juno spacecraft in interplanetary space right now.

Data storage in the cloud came in at number two. Three and four, simulation-based design and big data and analytics, are areas that I would argue are really MathWorks' core competencies. And those are actually areas that we're going to hear more about later on today.

Now a simulation-based design, or as we call it MathWorks, model-based design addresses design complexity. Our industry is undergoing an evolution that's been going on for decades. In the 1950s and '60s, and even '70s and onward, we started out as a very hardware-centric industry.

But the hardware quickly became too complex to control without software-- actually, a lot of software. When I was working on the International Space Station program, there was actually a running joke that the space shuttle had just enough software to run all the hardware, whereas the Space Station had just enough hardware to run all the software. and to some degree, that was true. And honestly, some organizations are still coping with how to deal with this software centrism that's emerged.

Now more recently, we've seen this trend towards a model-centric approach. Again, this is driven by our systems, including the software systems, continuing to evolve in complexity and becoming prohibitively complex to understand without models in many cases.

Now what do I mean when I say models? Well, models are a way to explore systems before they're built. But they're also a way to provide abstraction from the system for understandability. And they can provide different views that promote understanding to different domain experts and specialists.

And models don't have to be complicated, they can be as simple as this state chart on the right here. I throw this up as an example because to a software engineer working on the code, on the left might be the right environment for them to spend their time. But for an algorithm designer or a system engineer that's trying to understand how that code fits into the bigger picture, I would argue that the model on the right is vastly more easier to understand than the code on the left.

Now model-based design also addresses reconfigurability. This is a picture of one of NASA's Voyager spacecraft. They were launched in 1977 and the mission still continues. The design life of a typical large satellite today is on the order of 15 to 20 years. And if you look at how fast technology is moving, that satellite is doomed to be obsolete in terms of its technology by the end of its life.

So there's a tremendous interest in making our space systems more reconfigurable, which is actually theoretically made possible by systems that are more software-centric because software is something that you can modify in space. The hardware design, not so much.

But this raises the question of, how do you develop and test new capabilities for hardware that's not here on the ground where you can test it-- that's out there in space? And that's where a digital model on the ground becomes important.

Now of course, in order to do this effectively, you have to deploy digital engineering comprehensively across the full lifecycle of a program. So basically, the goal of model-based digital engineering is to enable the use of connected models throughout the lifecycle to digitally represent the system in the virtual world. And we have this concept of a digital thread which serves as a single source of truth for all of your design artifacts, data, design, test, et cetera.

If this concept is applied consistently, the natural outcome is that you end up with a digital twin. Now for the purposes of this discussion, I'll define a digital twin as a system in digital form that's so close to reality that it can be used in place of the real thing to analyze and assess the system. So that enables that concept of being able to redesign and reconfigure hardware that you don't have access to currently.

And clearly, this full adoption of digital engineering that I'm talking about here, it's a step beyond model-based system engineering where you have a static model of your design, maybe in SysML or some other modeling language. Or even model-based design that's constrained to one or two subsystems that you can execute standalone.

It really requires merging your model-based system engineering and your model-based design across multiple subsystem domains and across the lifecycle of the program. And this is kind of the next step in digital engineering that a lot of companies and organizations are currently trying to wrap their heads around.

Now most of our startup customers use MATLAB and Simulink, which are digital engineering tools, because it helps them develop more quickly and efficiently. And they care about things like, how quickly can I get to market? How flexible are my designs to changing requirements?

So for example, Virgin Orbit used our help. We helped them perform their stage separation analysis for the LauncherOne vehicle. And the main reason they adopted are our tools MATLAB and Simulink to do that analysis, because they were able to reduce their simulation time from days to hours, so faster development cycles.

Now a lot of our government and contractor customers, like NASA, for example, with the Orion program, they also like the reduced development cost because it makes them more competitive. But what they really care about is increasing quality due to fewer defects, right? As systems become more complex, especially human-rated systems, maintaining that quality with more complex systems is crucially important.

Now we've invested in working with space agencies around the world, including NASA, including the European Space Agency, to ensure that our model-based design workflows that you'll hear about can be used to comply with space standards. Now this page is an eyeful, I don't expect you to read it. But on our website, you will find a mapping of our software engineering workflows to both the NASA and the European Space software standards.

Now we've talked a lot about systems that operate in space. Let's touch a little bit about the consequences of this increased access to space that we've established. Basically, one thing that's resulted is that there is a lot of data coming down from space all the time.

And this now goes back to the data analytics, digital engineering investment topic that I brought up earlier. And this is actually where the industry has seen its first really practical use cases for AI and machine learning. Machine learning has been deployed on ground segment applications for several years.

Spacecraft health monitoring is through telemetry outlier detection, is one use case. I've seen the European large fleet operators in particular make some investments in these types of systems. They're of course motivated by having to monitor sometimes a couple of dozen satellites for their customers.

And they want to rightsize their staffing, reduce their staffing costs, but also make sure that the human operators catch off-nominal trends. And they've discovered that some of these algorithms that they've developed can actually catch off-nominal behavior and satellite telemetry faster than a ground operator would.

Now the other huge use case in space, of course, on the ground side, is geospatial analytics. We are getting tons and tons of Earth observation imagery, hyperspectral data, all kinds of data from space. There's just too much of it to be able to analyze by human being alone.

So machine learning and deep learning have been used again for several years now-- many years-- to solve this big data analytics problem for Earth observation. And there is a lot of different things that are being done, as I'm sure many of you are aware. Anything from agriculture and mining, assessing the health of fields or looking for minerals on Earth, to business-type analysis, maybe counting the cars in the parking of a store before a holiday to assess the health of the economy. All kinds of applications for this. And we'll hear more about this particular technology later today as well.

Now just in the last couple of years, we've also seen machine learning algorithms emerging in space. A classic problem for satellite imaging is the cloud detection problem. If you're trying to take pictures of the Earth's surface, then clouds are essentially a useless waste of bandwidth if you try to downlink images of clouds.

And the European technology demonstrator, PhiSat-1, has now solved this in orbit-- it's flying today. And Thales Alenia Space is working on a production implementation in the hyperspectral domain that they want to fly in a Sentinel spacecraft. So this is going from technology demonstrator to potentially production very quickly.

Another use case for machine learning could be rendezvous and proximity operations. This is something that NASA has studied a little bit. They flew a convolutional neural network aboard this little Seeker spacecraft. It was trained to detect and localize the Northrop Grumman Cygnus vehicle that was deployed in September of 2019. And this algorithm, from what I understand, performed well.

So to summarize, the outlook for space continues to be very exciting, despite some of the challenges we live with. There are high-demand applications that are fulfilled by our industry. We're seeing tremendous innovation and increased complexity that's being enabled by digital engineering tooling. And we're also seeing practical applications for machine learning. Thank you for listening.