How artificial intelligence is revolutionising space exploration

As humans begin a focused push to establish bases on the South Pole of the Moon, aiming at setting up colonies on Mars, and conceptualising interstellar voyages, AI is becoming an essential tool for realizing our cosmic ambitions. The vastness of space makes AI necessary for our robotic emissaries. Light takes eight minutes to travel from the Sun to the Earth, 14 minutes to Mars. Depending on the configuration of the planets, a signal to Mars can take between four and 22 minutes. Human controllers on ground stations cannot possibly micromanage every spacecraft operation, which is why we need to build-in the capabilities for machines to think for themselves.

On a frigid morning on 15 July, 2023, on the 854th sol or Martian Day of the mission, NASA’s Perseverance rover trekked across the Jezero Crater, its wheels leaving a mark on dust older than humanity. The rover was isolated on the Red Planet, 225 million kilometres from the Earth, tasked with finding signs of ancient life on a landscape of craters and dry riverbeds. The rover encountered a rock 35 centimetres in height that was too subtle for human eyes to spot from mission control. However, the rover was equipped with an advanced artificial intelligence system for just such a contingency. Even Perseverance had a co-pilot that allowed it to move around the rock.

The Perseverance Rover is equipped with an AutoNav system that was previously used on Curiosity, its predecessor, and initially demonstrated on the Deep Space 1 mission to a comet and an asteroid. This system allows the rover to pick out suitable paths through jagged rocks and sandy traps. A human on Earth would take about an hour to plot a 100 metre route, but the AutoNav system on Mars does it in seconds, doubling the range of the rover’s daily movements. Given an objective, the rover would make its own way to the goal. Bear in mind that there are no roads on Mars, and the rover is essentially offroading.

AI is steering spacecraft through the void of space beyond Mars. The Europa Clipper mission launched in OCtober 2024 uses AI to analyse data mid-flight, and will be looking for a subsurface ocean beneath the frozen crust of Europa. At 640 million kilometres from the Earth, the AI significantly speeds up the rate at which data is processed. Psyche, in orbit around an asteroid of the same name since August 2029 uses AI to sift through spectral readings, and identify minerals. Spacecraft can capture a deluge of data from instruments, which can be rapidly processed by AI.

The James Webb Space Telescope for example, a flagship deep space observatory generates so much data that astronomers rely on machine learning to spot patterns on all of its targets from distant galaxies, to exoplanet atmospheres. In 2024, an AI algorithm flagged a galaxy at a distance of 13.5 billion lightyears, pushing the timeline of the cosmic dawn. The Transiting Exoplanet Survey Satellite (TESS) has flagged over 6,000 potential exoplanets since 2018, and AI is helping confirm these discoveries. Scientists from MIT have developed a neural network called AstroNet that sifts through TESS data, cutting false positives by as much as 90 per cent. The approach allowed astronomers to identify a small, rocky world just 1.5 times the size of the Earth, around a star 50 lightyears away.

From variable stars that fluctuate in brightness, to the subtle dips in light caused by the transit of an exoplanet, to distant galaxies suddenly bursting with supernovae, to dark asteroids moving through the night sky, AI is supercharging the efforts of astronomers. Even the more mundane operations get a boost, with SpaceX Starlink satellites, now over 6,000 in number using AI to dodge debris in Low Earth Orbit faster than ground crews can. In 2023, an AI system on a Starlink cluster rerouted 12 satellites in under a minute, dodging a cloud of shrapnel from a Chinese satellite test in 2007. By 2024, the Starlink constellation satellites had executed over 5,000 collision avoidance manoeuvres. Millions of dollars of assets might have been lost without the split-second decision.

The Mars Helicopter Ingenuity used basic autonomy to scout ahead. NASA plans to dispatch its successor in the 2030s, packed with advanced AI technologies to map caves and ice deposits where humans or even rovers cannot venture. Robotic miners, guided by machine learning, will attempt to extract water and oxygen from Martian soil, while AI-driven greenhouses can grow crops on Mars. Elon Musk envisions AI rovers from Tesla laying the groundwork for a Mars colony, with a series of Starship missions to build up the population. The writing on the wall is clear, without AI, humanity’s aspirations in the space domain are dead in the water.