This Is How AI Discovers New Planets and Navigates Rovers – AI in Space

With the aid of AI, scientists discovered two exoplanets, Kepler-90i and Kepler-80g. But AI not only reveals planets: it also maps the universe, makes decisions for Mars rovers, and avoids satellite collisions. 

The Kepler-90 system compared to the Solar System; Source: NASA JPL

Almost 3’000 light-years from home resides a Sun-like star with a planetary system that feels remarkably similar to our own, with eight planets orbiting the host star in a configuration as we see it in the Solar System: the inner planets are rocky, protected by two outer gas giants. 

However, the planets orbit so close to their star that they could fit well within the space of Earth’s orbit. Seven planets were already found
orbiting Kepler-90 when the discovery of the 8th planet, Kepler-90i, came in 2017. 

By feeding data from the Kepler space telescope into machine learning algorithms, AI found this new, hidden planet that human scientists would have missed. AI helped similarly in the discovery of Kepler-80g, and we can only expect that AI will help in a much more significant way in the future. 

The age of AI is here. However, in space exploration, it has already played a significant role when the first rover landed on Mars almost 30 years ago and now steers rockets and satellites. Today, let’s take a look at the diverse use of AI in space, and how it has helped discover new planets and galaxies. 

The tiny Sojourner rover photographed by the Pathfinder lander; Source: Lights in the Dark

The rovers decide: How AI makes decisions on other planets

Which planet is merely inhabited by robots? You’re right: Mars. But it’s also a planet where AI and robots are united to become rovers. 

The first rover touched down on the Red Planet in 1997 when the Mars Pathfinder landed with Sojourner on board. The microwave-sized rover looked much different from how we know rovers today, and equally basic were its AI systems. 

The rover had the necessary equipment to ensure a safe trip to its destination, as well as systems to communicate with Pathfinder and the scientists on Earth. Sojourner had, like all rovers to follow, hazard avoidance systems.

This was a super advanced rover at the time, and NASA said in 1997, celebrating the first landing of a rover on a different planet, “For the first time, a ‘thinking’ robot, equipped with sophisticated laser eyes and automated programming, is ‘thinking’ and reacting to unplanned events on the surface of another planet.”

The next two rovers that landed in 2003, Spirit and Opportunity, were much better in comparison to their precedent. The twin rovers could plan their routes autonomously and analyze images – very convenient for Opportunity, which exceeded its expected life span by eight Martian years (15 Earth years) and produced more than 200’000 raw images. 

Indeed, it was Spirit and Opportunity that discovered the fact that the Martian climate was once much different with water on its surface. Spirit got stuck in sand and lost contact with Earth in 2010. Opportunity operated well into 2018, but then failed to reboot after a global dust storm. 

Self-portrait taken by Curiosity; Source: Space.com

However, the next operational rover was already on Mars: Curiosity. 

If you were to compare Martian rovers to phones, I’d say Sojourner would be a Nokia and Curiosity a modern smartphone, equipped with all types of technologies to facilitate its exploration of Gale crater in the Aeolis quadrangle. 

Equipped with the Autonomous Exploration for Gathering Increased Science (AEGIS) system, Curiosity can help scientists on Earth directly with what it discovers. Curiosity’s near-twin, Perseverance, which landed in Jezero crater in February 2021, has the same system on board. 

With AEGIS, the rovers can collect valuable data on the geologic features on Mars. Sometimes, the system can recommend what the rover could do on a day depending on the terrain, energy usage, and scientific value. 

As for Perseverance, the rover also has systems to measure the weather, being able to measure the highest and lowest temperature, atmospheric pressure, humidty, and ultraviolet radiation – an important feature that can provide valuable data for future crewed missions. 

Check out NASA’s website to see what the weather’s like today! (I personally always find that very fascinating)

When it comes to searching for traces of ancient life on Mars, machine learning has also proven helpful. A research paper reported last year that AI can help rovers search for life with 90% accuracy.

In tests, the AI was able to differentiate between living and non-living samples. Perhaps soon we get to find the first traces of alien life from Mars itself! 

All kinds of satellites dominate LEO; Source: Earth.com

Satellites avoid collision and analyze data from space 

AI also plays a crucial role in space closer to home, namely in Low Earth Orbit (LEO). LEO is a busy place, with thousands of satellites scurrying over our planet, providing communications, scientific data, and reconnaissance. 

These satellites produce an immense amount of data. A human alone could never analyze all the data from the satellite, so AI has helped make the process more efficient. 

AI processes data and imagery, and is able to pick data that may be interesting for further study. This data is then sent to Earth, where humans take a look at the data and continue to work with it. 

This means that AI has the responsibility over many types of data: from meteorological and climate data to reconnaissance imagery. 

AI not only processes satellite data, it also helps the satellites themselves. For example, a remote satellite health monitoring system can improve decision-making and detect possible issues with the satellite itself. AI also aids in avoiding satellite collisions, to minimize the danger of space debris and losing valuable resources. 

However, we not only take photos of one planet from orbit. Here comes the Mars Reconnaissance Orbiter (MRO), which has sent a ton of data to scientists on Earth to study the Red Planet. In fact, it processes six megabits of data per second, thanks to the help of AI. AI filters out important data and has been trained to recognize key features from billions of images of Mars’ surface. 

Ring galaxies were thought to be the rarest type of galaxy – until the Galaxy Cruise project found 30'000 of them! Source: Universe Today

How AI finds new planets and galaxies

As noted in the beginning of this article, AI helped discover the exoplanet Kepler-90i, a super-Earth with temperatures as hellish as on Venus, orbiting its star much closer than Mercury. Another exoplanet AI discovered is Kepler-80g. This planet is just a tiny bit bigger than Earth, and it orbits its red dwarf star at the same distance as Kepler-90i.

The Kepler space telescope monitored about 150’000 main-sequence stars in the constellation Cygnus over nine years, leading to the discovery of almost three thousand planets, with more discoveries surely to follow. 

With these vast amounts of available data, AI has enhanced the discoveries, as seen with Kepler-90i and Kepler-80g, with 96% accuracy!

Looking further, AI has helped discover new objects in deep space. Using neural networks, AI has been trained to identify new galaxies in photos taken from space. By now, these neural networks can classify galaxies with 98% accuracy. 

The Galaxy Cruise citizen science project has also shown that AI and humans can work together really well – thanks to this project, AI and citizen scientists discovered a total of 430’000 galaxies all across the universe!

It’s fascinating to see how much astronomy and space exploration have improved thanks to AI. Whether it’s navigating rovers and rockets or identifying previously unseen exoplanets, AI has become a fundamental part of space. In the future, who knows what else AI can do?