AI and space exploration: what purpose for artificial intelligence serve in space?

While Europe seems to be back on track in the space sector, particularly in the world of start-ups, artificial intelligence has already made its way into the industry and is now being put to use. But what impact does AI have on space exploration, and what role does it play?

Photo de Nasa sur Unsplash

In short

• The number of satellites and probes is increasing every year, generating a colossal volume of data to be analysed.

• AI can sort and interpret this data, detecting anomalies, events, and celestial bodies more quickly.

• It is already being used on Earth for weather forecasting, fire detection, and river monitoring.

• In space, AI gives stations and robots more autonomy, especially for complex landings.

• It also helps reduce launch costs and could play a key role in space colonization.

Artificial intelligence in space exploration: what are we talking about?

More and more objects launched into space

For several years now, artificial intelligence has had an essential application, albeit one that seems fairly basic in appearance: processing data collected by satellites, probes, rovers, and other space objects sent into space. To give an idea, "the total number of satellites actually deployed in orbit is very impressive: nearly 2,900 satellites. This represents a 17% increase compared to 2022, which is also a more modest increase than that seen in 2022 (36% compared to 2021)."* The launch of satellites, and more specifically in recent years of CubeSats, small, lightweight satellites that are becoming increasingly affordable, is increasing year on year. This is due to the increasingly low cost of access to space thanks to the reuse of rockets, pioneered by SpaceX, and the emergence of several new players in the form of start-ups, creating fiercer competition and thus driving down the price of access to space, relatively speaking.

*Un autre regard sur la terre ; L’année spatiale 2023 : le bilan des lancements orbitaux

An ever-increasing volume of data to decrypt

This increase in data is also reflected in the scientific field with all the probes and devices already in space: Voyager 1 and 2, Akatsuki, James Webb... all these devices continuously send raw data that is heavy and difficult to decrypt. One figure illustrates this point: "A quarter of a million terabytes of satellite data is what NASA's various Earth science missions are expected to generate in 2024, which is 25,000 times the total number of printed documents in the US Library of Congress."*

*LFDE ; L’IA & les données spatiales, un duo au service de la planète

AI and its processing capabilities at the service of terabytes of data to be analyzed

This is where AI comes into its own. Thanks to increasingly powerful and better-trained algorithms, AI-powered programs are capable of processing ever-larger amounts of data, enabling them to sort, classify, and catalog the various celestial bodies discovered at an ever-faster rate. If AI can do this, it can also more easily detect anomalies that fall outside existing models. This allows the scientific community to identify space events more quickly and efficiently, and to find more of them.

Data to be processed that can primarily help the Earth and societies

Before turning our attention to other planets, artificial intelligence is already enabling us to help Earth and its inhabitants. How? Thanks to the ever-increasing amount of data reported by various satellites and probes (including those orbiting Earth), we are learning more and more about our blue planet. In addition, weather and climate monitoring satellites can detect weather variations with increasing accuracy, take pictures of Earth to provide a global view of natural events, and more.

This ability of AI to analyze data for the common good is not new. In May 2024, "the Toulouse-based startup BWI succeeded in predicting river flows and levels ten days in advance."* Artificial intelligence is now essential for forecasting meteorological, geological, and hydrological data. Several other major powers, such as India and the United States, have already begun using AI for this purpose.

*La tribune ; Spatial : quand l'intelligence artificielle permet de prédire les débits des cours d'eau

Another example of the crucial role played by AI comes from the University of South Australia, which in 2024 developed an AI model capable of detecting forest fires by analyzing satellite images. Given that forest fires have doubled in the last 20 years, largely due to climate change, such a program is vital. AI, coupled with the satellite launched for this mission, could "detect fires up to 500 times faster than with traditional ground-based image processing."* This was made possible by integrating an artificial intelligence module directly into the satellite, capable of processing the photos taken by the satellite itself.

*Geo ; Incendies de forêt : l'IA à la rescousse, pour détecter immédiatement les flammes depuis l'espace

How is AI able to detect fires or changes in water levels in rivers? Through supervised learning, among other things. This involves feeding the AI with data that has already been sorted and "annotated" in order to train it for a specific purpose, such as detecting fires or changes in water volume. This is particularly the case with the AI model created in Australia, which was fed satellite images of recent fires and wildfires. This enabled the AI to differentiate between smoke from a fire and clouds in order to detect it and raise the alarm.

But image detection isn't limited to natural disasters. Geopolitics also plays a role. AI capable of recognizing visual changes can also be trained to differentiate between commercial ships and warships. This is a very useful feature for observing the movements of fleets or commercial ships.

Crucial autonomy for spacecraft, space stations, and extraterrestrial missions made possible by AI

AI-controlled spacecraft and stations

In the future, the ISS, the International Space Station, will have to be deorbited and destroyed by 2030*. This will be an opportunity for various private players to showcase their expertise in creating space stations or parts of space stations for future installations. Given the geopolitical situation, we can assume that NASA will decide to create its own space station, perhaps in collaboration with European companies and private players. China has already created its own space station, and other emerging powers in the sector, such as India, may be able to do the same.

*La Tribune : ISS : SpaceX missionné par la Nasa pour détruire la station spatiale internationale après 2030

These multiple stations will therefore need high-performance and increasingly autonomous systems to manage them, as they could very well be left unmanned during certain periods. Artificial intelligence could then greatly help these future stations to manage themselves, in particular to continue experiments started by staff. In fact, the ISS has already had AI on board to supervise some of the station's systems since 2018, called Cimon (Crew Interactive Mobile Companion).

*Le Monde Informatique : Cimon, le robot IA d'Airbus et IBM, embarque dans l'ISS

Exploration made easier by AI-powered robots

If AI helps supervise human installations in space, it will also be useful for exploration. In fact, it would be more accurate to say that artificial intelligence is already being used for exploration.

We have already discussed AI models trained with images to recognize changes in terrain. This same technique is already being used and improved for space landers. Over time, this technology will enable robots to recognize terrain directly and land autonomously, rather than having a human pilot the object remotely.

This represents a significant gain in precision and reliability. Why? Sending a message or communication in the broadest sense from Earth to Mars takes 20 minutes to travel! In other words, if a Mars lander is in the approach phase, the data it sends will be received 20 minutes later by Earth, which is far too late to react accordingly. Robots capable of reacting to the terrain thanks to AI that can recognize the nature of the ground are, therefore a crucial asset for the future of space exploration.

Access to space becoming easier thanks to artificial intelligence

While AI promises to facilitate space exploration and colonization, it also promises to reduce the costs associated with accessing space. How? By designing the perfect rocket! More specifically, companies are already testing the application of artificial intelligence in the design of low-cost rocket models in order to reduce their costs. Democratizing this process would lower the overall cost of accessing space and even allow new players to emerge, further diversifying the sector.

This is particularly true of Rocket Factory Augsburg, also known as RFA. It is a German start-up whose goal is to offer low-cost launches for putting small satellites into orbit. The company intends to succeed thanks to artificial intelligence, using precise programs and specific requirements such as the best value for money for an alloy. The space channel Hugo Lisoir presents it in detail in its documentary, linked below.

Will AI make future space colonization possible?

All the advantages promised by AI can be applied to a large extent to space colonization. If we imagine a very long-term future with ships sent to the far reaches of space, the latter could be piloted entirely by AI to allow humans to remain in a kind of permanent sleep, preventing them from aging. We can also imagine ships without humans on board, sent first to prepare for human colonization by creating the essential infrastructure. AI would then be used to manage the ship throughout the journey and handle the approach phase without the need for a pilot, thanks to AI capable of detecting suitable landing areas in real time.

To explore one last new concept in this article, artificial intelligence could also be used to create machines known as "self-replicating." This term refers to machines capable of reproducing themselves using local resources. Thus, by sending a spacecraft with a limited number of resources and the right machine(s), the latter would then be able to reproduce using the resources found on site. In this way, a real colony could be created with robots and without the need for a gigantic spacecraft, which would be difficult to build and keep in one piece for the journey.

Artificial intelligence seems well on its way to continuing to carve out a place for itself in the space sector. This field still has a lot to show us. There is no doubt that space news combined with AI will offer us a very interesting spectacle!

FAQ

1. Why is AI important for space exploration?

Artificial intelligence processes huge volumes of data and helps automate tasks that are impossible for humans alone to manage. It is therefore crucial for the future of space exploration and the sector.

2. What are the uses of AI on Earth thanks to satellites?

Artificial intelligence is used for weather forecasting, climate monitoring, fire detection, and even geopolitical observation.

3. Is AI already being used in space stations?

Yes, for example with CIMON, an intelligent assistant that has been on board the ISS since 2018.

4. How does AI help missions to Mars?

It allows robots and landers to recognize the terrain and react in real time, without waiting for Earth-Mars communication delays.

5. Can AI really facilitate space colonization?

Yes, it could manage autonomous spacecraft, create infrastructure before humans arrive, and even develop self-replicating machines.