On October 30, the InnoCube satellite performed a maneuver in Earth’s orbit. This event might have gone unnoticed, but it was controlled by artificial intelligence for the first time in history. The test was successful.

InnoCube satellite in the laboratory.
Source: phys.org

Artificial intelligence-based satellite orientation controller 

As a real milestone on the path to autonomous space systems, a research team from Julius Maximilian University of Würzburg (JMU) has successfully tested an AI-based attitude control system for satellites directly in orbit — a world’s first experience. The test was conducted on board the 3U nanosatellite InnoCube.

During the satellite’s passage between 11:40 and 11:49 Central European Time on October 30, 2025, an AI agent developed at JMU performed a complete orbital orientation maneuver, fully controlled by artificial intelligence. Using reaction wheels, the AI brought the satellite from its current initial orientation to a predetermined target orientation. The AI then had several further opportunities to demonstrate its capabilities: in subsequent tests, it also successfully and safely maneuvered the satellite into the desired orientation. The LeLaR research team—Dr. Kirill Djebko, Tom Baumann, Erik Dilger, Professor Frank Puppe and Professor Sergio Montenegro—thus took a decisive step toward space autonomy.

LeLaR Project

The goal of the In-Orbit Demonstrator for Learning Attitude Control (LeLaR) is to develop the next generation of autonomous attitude control systems. The main focus is on the development, training, and in-orbit testing of an artificial intelligence-based attitude controller aboard the InnoCube nanosatellite.

Attitude controllers stabilize satellites in orbit and prevent them from spinning chaotically. They are also used to point the spacecraft in the desired direction. For example, to align cameras, sensors, or antennas toward a specific target.

What makes this work special is that the Würzburg controller was not created using traditional, fixed algorithms. Researchers applied deep reinforcement learning (DRL), a branch of machine learning in which a neural network autonomously learns the optimal control strategy in a simulated environment.

The key advantage of the DRL approach is its speed and flexibility compared to classical control system development. Traditional orientation controllers often require lengthy manual parameter tuning by engineers—sometimes taking months or even years. The DRL method automates this process. In addition, it opens up the possibility of creating controllers that automatically adapt to differences between expected and actual conditions, eliminating the need for lengthy manual readjustment.

Bridging the gap between simulation and reality

Before implementation, the AI controller was trained on Earth in a highly detailed simulation and then uploaded to the flight model of the satellite in orbit. One of the biggest challenges was overcoming the so-called simulation-to-reality gap (Sim2Real) — ensuring that the controller trained in simulation would also work effectively on a real satellite in space.

“This is a real breakthrough,” emphasizes Djebko from JMU. “We have achieved the world’s first practical confirmation that a satellite orientation controller trained using deep reinforcement learning can operate successfully in orbit.”

Trust in AI in space applications

By successfully demonstrating an AI-based controller in orbit, the Würzburg team has shown that artificial intelligence can be reliably used in space missions that are critical to safety. Puppe is convinced that “this will significantly increase the acceptance of AI methods in aeronautics and space research,” pointing to the important role of simulation models. Growing confidence in this technology is a crucial step toward future autonomous missions, such as interplanetary or deep space missions, where human intervention is impossible due to vast distances or communication delays. An AI-based approach could thus become vital for the survival of spacecraft.

Prospects for the application of AI for satellite platforms

This successful test in orbit establishes the University of Würzburg as a pioneer in the field of space systems controlled by artificial intelligence. The demonstrated AI-based controller is an important building block for future space exploration. The results of the LeLaR project could enable faster and more cost-effective development of new, complex AI-based controllers for a wide range of satellite platforms.

“The next goal is to build on this initial success,” says Djebko. “This is a big step toward complete autonomy in space,” adds Montenegro. We are at the beginning of a new category of satellite control systems: intelligent, adaptive, and capable of self-learning.