A 300 satellite constellation has about 580 alerts a year – manual manoeuvres cost €25,000 each (so up to €14m pa)
Satellites are a risky business (their history is littered with failed launches and business models), but they are rapidly becoming an everyday part of operators’ technology mix to provide connectivity and other services.
For one thing, they are part of the 5G standard and for another, lots of operators are looking to space to provide coverage to the many areas that have no or poor network coverage. Also, satellite comms are finally offering far higher transmission speeds than traditional systems and without needing expensive, specialist devices.
For example, in April, AST SpaceMobile, Rakuten Mobile and AT&T made a phone call via a low Earth orbiting satellite (Leo) using an ordinary Android smartphone. AST SpaceMobile’s mission is to provide connectivity, via space, to the half of the global population that lacks cellular or fixed broadband.
AST SpaceMobile has agreements and understandings with 36 mobile operators with a collective 2 billion subscribers, including Bell Canada, Orange, Telefónica, TIM, Saudi Telecom Company (stc), Zain KSA, e& (Etisalat), Indosat Ooredoo Hutchison, Smart Communications, Globe Telecom, Millicom, Smartfren, Telecom Argentina, Telstra, Africell, and Liberty Latin America, as well as Vodafone Group, AT&T and Rakuten Mobile. Vodafone Group is one of several with a financial stake in the satellite firm.
Not always the more the merrier
As ever more satellites and constellations are put into orbit, routes are busier and there’s more debris floating about – and things are only just getting going. ESA estimates that a 300-satellite constellation may receive about 580 alerts annually that trigger human intervention and satellite manoeuvres. As an emergency manoeuvre in LEO costs about €25,000, this can add up to a staggering €14 million per year, even though many of the manoeuvres are not necessary.
Enter Neuraspace. It was set up just over two and a half years ago in Portugal by two students who were studying machine learning and now has 25 staff. The founders wanted to reduce revenue loss for operators by providing space traffic management (STM) based on AI. Their STM system is being tested by some of the world’s biggest satellite operators to fully automate the process of dealing with space debris, orbital traffic, avoiding false-collision manoeuvres and better planning to avoid collisions. Last year the company raised €27 million.
Professor Dr-Ing* Chiara Manfletti is Director and COO of Neuraspace (pictured), joining when the company was a year old. She is also a professor for Space Propulsion and Mobility at the Technical University of Munich. Previous roles include the first ever President of the Portuguese Space Agency, heading the Policy and Programmes Coordination Department at the European Space Agency (ESA), and Programme Advisor to the Director General at the ESA’s headquarters in Paris.
Since then, it has raised €2.5 million funds from Armilar Venture Partners and €25 million from the EU’s NextGeneration EU Economic Recovery Fund project which enabled the start-up to get STM platform to market in 2022 for feedback from customers, offering service on a Software-as-a-Service model.
Neuraspace has signed up a paying customer and has seven other satellites operators are trialling the platform who collectively plan to launch 400-500 satellites into orbit over the next two years.
Leading with AI
Manfletti says, “We pride ourselves for being AI- and machine learning-based from the beginning: the first two women who were hired were actually AI specialists. The other technology or philosophy we have embedded in the company since the beginning is data fusion.”
She explains this is key because, “If you talk to different companies about domain or space situational awareness (SSA), some will tell you that radar is the best type of sensor there is, others that optical in-space is the best and still others that that optical sensors at the ground station is the best. In fact, these sensors are complementary: no single set of sensors can give you a full picture of what’s happening up there”.
Radar can’t reach middle-Earth or Geo-Earth orbits (MEOs and GEOs) and optical can only see during the day, while Numerica is looking at infrared sensors.
Manfletti continues, “Our goal is to bring data from different data sources together. We want to be an infrastructure-light company…to add value through analysing data and bringing the best insight for most rapid decision making to operators.”
Elliptical ellipsoids
She claims that Neuraspace is the first company to bring an AI and machine learning model to market, stressing that one of the biggest issues is that no-one knows exactly where the objects are in space, “And these uncertainties propagate in time,” Manfletti adds. “If you’re an object in LEO, as you move through the Earth’s atmosphere, the atmosphere’s density changes over time because of space weather and other elements, such as the fact that the Earth isn’t round.
“So if you try to propagate the orbit of an object in space, uncertainties about where it is become larger and larger. In attempts to predict a conjunction or collision that’s going to happen in five or six days, these uncertainties become huge.”
She continues, “Ellipsoids generally describe where objects are, with an object at the centre…but ellipsoids gets bigger over time. The more frequently you observe an object, the smaller the uncertainties. If you can’t revisit where an object is, the other option is trying to predict how the uncertainties will evolve…when we calculate the probability of collision, the uncertainty or co-variance, goes into that calculation.”
An advantage of this approach is to give operators insight as to whether they can delay deciding what, if any, action they take so they don’t have to decide five or six days in advance. As the possible event gets closer and the uncertainties aren’t diminishing, a dedicated tracking service could be brought into play. As uncertainties are greater when space debris is the object in question, dedicated tracking with a telescope or radar could be the best option.
Simulated situations
To build its model, Neuraspace simulated a simulation of a constellation of about 250 satellites, at an altitude of 700km, and computed that it could expect about 4,000 conjunctions weekly for a miss distance of below 5km, which is the usual parameter. Having to make decisions about so many is a huge burden on operators.
Manfletti explains, “We calculated that of those 4,000 about 4% are high risk, with a miss distances below 1km and manoeuvres may be required to avoid collisions as often as two out of every three days – a total of about 250 per year. These are big numbers from our simulation for one specific orbit.”
The manoeuvres themselves have challenges too: “As a satellite operator, you want to ensure that you’re not manoeuvring over high density areas…We also try to go beyond just the probability of a collision, to assess the risk,” she says. “We provide tailored advice to operators and how they could best manoeuvre, for example, based on if the satellite is for telecoms or Earth observation. Does it have chemical propulsion, electric propulsion or no propulsion? What are attitudes, boundary conditions and any other preferences the operator may have? It’s a global solution tailored to customers’ assets.”
Neurospace says the other systems mentioned here aren’t automated, so space track, for instance, provides conjunction data messages that are examined manually to compute what if any manoeuvres are needed. Neurospace argues that it’s just too hard and there are too many of them to sustain this practice.
Manfletti comments, “I’ve spoken to telecom operators and…If you have to do 24/7 operations and hire people dedicated to this task, it has a huge impact on the cost. For the constellation we simulated, we calculated HR costs of €6 million to €10 million – just to staff collision-avoidance manoeuvres.”
Massive potential
Most solutions designed to address STM are ground based, but she thinks, “We can optimise it further, putting some of this intelligence in orbit…and inter-satellite links are being developed. I dream of a place where satellites are talking to each other and saying, ‘Careful I’m coming by next Wednesday’.”
Manfletti also thinks the industry needs “some kind of minimum data protocol so that satellites are aware and avoid collisions. This is with a bit of a naive hat on, assuming there are no defence activities and people are happy to tell other people where their satellites are. On the other hand, we so they can be tracked and if a ship turns it off, and is trying to go unseen, then they’re up to no good.”
She concludes, “Of course, we want to commercially successful, but there’s a fundamental, underlying [belief] that this is something we must do: I think it is also a responsibility, reaching out to these constellations and these big leader… to see how it can be advantageous.”
Integrated infrastructure
She notes that a speaker from the US government said at conference in Washington in March that the Chinese reckon that space is not a $1 trillion economy, but a $10 trillion one. Manfletti comments, “If that is true…I think it’s going to be our ability to integrate space and non-space, which will be the key to making that happen. So I think we still have a ‘space bubble’ – when I was in Portugal, I was talking to telecom services providers and many saw ground infrastructure as being in competition with satellite or space-based infrastructure.
“It was very hard to get across the message that the two things are complementary, and can boost boost each other to create more economic success. For example, the Azores are in the middle of the ocean [about 1,400 km or 870 miles west of Lisbon] and so are connected by very long cables to mainland Portugal. When they were discussing renewing the infrastructure, I suggested it may be less cost intense to put satellites in orbit and sell services to the rest of the world as well, rather than new cables. That’s what I meant with integrated infrastructure.
The more we build in space-based data services into our daily lives, for urban planning, mobility or whatever else, the space economy is going grow.
* The German equivalent of a PhD in Engineering