Space is the ultimate environment for autonomy. Radio signals take between about 3 and 22 minutes to travel one way between Earth and Mars, depending on the planets’ positions, so a rover cannot be driven with a joystick. Satellites produce more data than can be downloaded or analyzed manually. And missions to the outer solar system must act on their own for long periods. AI and autonomy are therefore essential tools of space exploration.

Autonomous rovers

NASA’s Mars rovers have used increasing levels of onboard autonomy:

  • Autonomous navigation: the rover builds 3D maps from stereo cameras, identifies hazards and plans safe paths to a target chosen by mission planners. Perseverance, which landed in 2021 as part of the Mars 2020 mission, has an enhanced self-driving system that lets it navigate while driving, covering more ground per day than earlier rovers.
  • Autonomous targeting: the AEGIS software, first used on Opportunity and later on Curiosity and Perseverance, lets rovers select interesting rocks for laser analysis without waiting for instructions from Earth.

Landing

Perseverance used Terrain-Relative Navigation during descent: it compared camera images of the approaching surface with onboard maps to determine its position and steer away from hazards, enabling a landing in the challenging Jezero Crater.

Earth observation

Thousands of satellites image Earth daily. AI processes this data to:

  • map deforestation, crops, floods and wildfires;
  • monitor methane leaks and other emissions;
  • track ships and aircraft;
  • detect changes in infrastructure and urban areas;
  • support disaster response with rapid damage assessment.

Some satellites now run AI models onboard, analyzing images in orbit and sending down only useful results — for example, discarding images obscured by clouds.

Astronomy

AI helps classify galaxies, detect exoplanets in telescope data — as with the discovery of Kepler-90i using a neural network in 2017 — find fast radio bursts and transient events, and handle the enormous data volumes of new survey telescopes.

Mission operations

  • scheduling observations and communications;
  • detecting anomalies in spacecraft telemetry;
  • planning activities under power and thermal constraints;
  • managing collision risks from the growing number of satellites and debris objects in orbit.

Human spaceflight

AI assistants are being explored to support astronauts with procedures, diagnostics and medical guidance on long missions where Earth contact is delayed.

Challenges

  • Radiation-hardened hardware is far less powerful than terrestrial computers, limiting onboard AI.
  • Reliability and verification are critical: a software error can end a mission.
  • Limited training data for unexplored environments.
  • Space traffic: as constellations grow, automated collision avoidance becomes vital — and requires coordination between operators.

A day in the life of an autonomous rover

A typical Martian day for Perseverance shows how humans and autonomy share the work:

  1. Overnight on Earth, the science and engineering teams review images and data sent back from the previous day and agree on goals: drive toward a rock outcrop about 200 meters away and analyze a target on arrival.
  2. Commands are uplinked in the Martian morning as a plan for the whole day, because real-time control is impossible.
  3. During the drive, the rover takes stereo images, builds a 3D map of the terrain ahead, rates possible paths for hazards such as rocks, sand and steep slopes, and steers itself — re-planning as it goes.
  4. On arrival, it may use AEGIS to pick a promising rock for laser analysis without waiting for instructions.
  5. In the afternoon, it sends data back to Earth through orbiting relay satellites, and the cycle repeats.

The more the rover can do safely on its own, the more science it can return over a mission with a limited lifetime.

Lunar missions and the next decade

Plans to return humans to the Moon and to build permanent infrastructure there will rely heavily on autonomy: robotic landers and rovers scouting resources, systems that monitor habitats, and spacecraft that dock and navigate with minimal ground control. Commercial companies are developing lunar landers and rovers with increasing autonomy, and several have already attempted landings, with mixed success — a reminder of how hard operating in space remains.

What to watch next

  • Onboard AI processing on more Earth-observation satellites.
  • Greater autonomy for future Moon and Mars missions.
  • Automated collision avoidance and space traffic coordination.
  • AI assistants for astronauts on long-duration missions.

Key takeaways

  • Communication delays make onboard autonomy essential for Mars rovers.
  • Perseverance navigates autonomously and used terrain-relative navigation to land.
  • AEGIS lets rovers choose science targets without waiting for Earth.
  • AI analyzes satellite imagery for climate, disasters and monitoring.
  • Limited onboard computing and the need for reliability constrain space AI.

Frequently asked questions

Do Mars rovers drive themselves?

Partly. Mission teams choose destinations and plans; the rovers autonomously navigate around hazards and can select some science targets.

Do satellites use AI?

Many satellite data pipelines use AI on the ground, and a growing number of satellites run AI models onboard.

Why can’t rovers be driven in real time?

Signals between Earth and Mars take several minutes each way, making real-time control impossible.