Get ready for a groundbreaking revelation! AI has taken control of a satellite's attitude in orbit, marking a historic first. This achievement, led by a research team at Julius-Maximilians-Universität Würzburg (JMU), is a giant leap towards autonomous space systems.
On October 30, 2025, between 11:40 and 11:49 a.m. CET, an AI agent developed at JMU executed a complete attitude maneuver on the 3U nanosatellite InnoCube. Using reaction wheels, the AI guided the satellite from its initial position to a specified target attitude, all under the command of artificial intelligence. But here's where it gets controversial...
The AI wasn't just a one-hit wonder. It had multiple chances to prove itself, and each time, it successfully and safely controlled the satellite to the desired attitude. This is a game-changer for space autonomy.
Meet the LeLaR research team: Dr. Kirill Djebko, Tom Baumann, Erik Dilger, Professor Frank Puppe, and Professor Sergio Montenegro. These experts have taken a giant leap towards space autonomy with their project, the In-Orbit Demonstrator for Learning Attitude Control (LeLaR).
The LeLaR project aimed to develop the next generation of autonomous attitude control systems. The focus was on designing, training, and testing an AI-based attitude controller aboard the InnoCube nanosatellite.
Attitude controllers are crucial for stabilizing satellites in orbit and preventing them from spinning out of control. They also help point spacecraft in the desired direction, aligning cameras, sensors, and antennas with specific targets.
What sets this work apart is the use of deep reinforcement learning (DRL), a branch of machine learning. The Würzburg controller was not built using traditional, fixed algorithms. Instead, the researchers employed a DRL approach, where a neural network autonomously learns the optimal control strategy in a simulated environment.
The key advantage of DRL is its speed and flexibility compared to classical control development. Traditional attitude controllers often require lengthy manual parameter tuning by engineers, sometimes taking months or even years. With DRL, this process is automated, and controllers can adapt to differences between expected and actual conditions without time-consuming manual recalibration.
One of the biggest challenges was bridging the Sim2Real gap, ensuring that the controller trained in simulation could operate on the real satellite in space.
"A truly decisive success," emphasizes Djebko. "We've proven that a satellite attitude controller trained using Deep Reinforcement Learning can operate successfully in orbit."
Baumann adds, "This successful test demonstrates that AI can perform precise, autonomous maneuvers under real conditions, not just in simulation."
By showcasing the reliability of AI in safety-critical space missions, the Würzburg team has increased acceptance and trust in AI for space applications. Puppe believes this will significantly impact aeronautics and space research, highlighting the importance of the simulation model.
Growing trust in this technology is crucial for future autonomous missions, especially in interplanetary or deep-space scenarios where human intervention is impossible due to vast distances or communication delays. AI-based approaches could be vital for spacecraft survival in these situations.
With this experiment, the Würzburg team has achieved a major goal in the LeLaR project. Erik Dilger says, "This success motivates us to extend the technology to new scenarios."
The test was conducted aboard InnoCube, a satellite developed in cooperation with Technische Universität Berlin (TU Berlin). InnoCube serves as a platform for innovative space technologies, allowing researchers to test new concepts directly in orbit.
One such innovation is the wireless satellite bus SKITH (Skip The Harness), which replaces conventional cabling with wireless data transmission, reducing mass and potential failure points.
The University of Würzburg has established itself as a pioneer in AI-driven space systems with this successful in-orbit test. The AI-based controller represents an important building block for future deep-space exploration. The LeLaR project's results could enable faster and more cost-effective development of new, complex AI-based controllers for various satellite platforms.
Djebko concludes, "We aim to build on this head start."
Montenegro adds, "We're witnessing the beginning of a new class of satellite control systems - intelligent, adaptive, and self-learning."
This achievement is a significant contribution to space autonomy, and the future looks bright for AI-driven space exploration.
