AI Controls Satellite Orientation for the First Time in Space
For the first time, scientists have successfully used artificial intelligence to autonomously control the orientation of a satellite in orbit — a breakthrough that could revolutionise spacecraft operations and safety in space.
A Major Leap in Satellite Autonomy
The experiment, led by researchers at Julius-Maximilians-Universität Würzburg (JMU) in Germany, demonstrated how an AI-driven system can independently adjust a satellite’s attitude — the term used to describe its orientation relative to space. Normally, these manoeuvres are handled by human operators or pre-programmed routines, both of which require time, resources and prior knowledge of possible scenarios.
The project, known as the In-Orbit Demonstrator for Learning Attitude Control (LeLaR), showcased the first real-world instance of a satellite changing its own orientation without human input. Using a nanosatellite called InnoCube, the system applied deep reinforcement learning, a branch of machine learning that enables software to learn through trial and error.
Testing AI in Orbit
Before being deployed, the AI was trained using a high-fidelity simulator at JMU to prepare it for real orbital conditions. Once in orbit, researchers tested the system by setting a target attitude for the satellite and allowing it to adjust itself. Mechanical reaction wheels — the devices that physically rotate a satellite — were controlled directly by the AI.
The first in-orbit test took place on 30 October, when the InnoCube successfully achieved the desired orientation autonomously. The team repeated the test several times, confirming consistent results.
“This successful test marks a major step forward in the development of future satellite control systems,” said Tom Baumann, a research assistant at JMU and a member of the LeLaR project. “It shows that AI can not only perform in simulations but also execute precise, autonomous manoeuvres under real conditions.”
Transforming Satellite Operations
Attitude control is critical for satellite missions, ensuring instruments point in the right direction, maintaining thermal balance and enabling navigation adjustments. Traditionally, such control requires extensive ground communication, which can delay responses and increase costs. The new system could help satellites make instant decisions in complex or unexpected situations, reducing human oversight and operational expenses.
While other agencies have used AI to automate satellite tasks — such as NASA’s dynamic camera targeting or the US Naval Research Laboratory’s Autosat project — none have previously demonstrated full motion control in orbit.
Professor Sergio Montenegro, who leads aerospace information technology at JMU, described the achievement as a pivotal moment. “It’s a major step towards full autonomy in space,” he said. “We are at the beginning of a new class of satellite control systems — intelligent, adaptive and self-learning.”
This advancement opens the door to the next generation of fully autonomous spacecraft, capable of managing themselves with minimal human intervention — a vital step for future deep-space exploration.
with inputs from Reuters