JMU Demonstrates World-First Autonomous Satellite Orientation Using Artificial Intelligence

Researchers at Julius-Maximilians-Universität Würzburg (JMU) have achieved a significant milestone by successfully demonstrating the autonomous control of satellite orientation using artificial intelligence (AI) directly in orbit. This innovation promises substantial improvements in the safety and operational efficiency of future spacecraft. Precise orientation adjustment is a critical prerequisite for satellites to ensure instruments are perfectly aligned, manage thermal loads from solar radiation, and execute necessary positional maneuvers.

Traditionally, these vital tasks have relied heavily on human intervention or predetermined software routines, a methodology often proving slow and insufficiently adaptable to unforeseen anomalies in space. The JMU team, led by Professor Sergio Montenegro, developed an AI-based attitude controller for the 3U nanosatellite designated InnoCube. During a satellite pass on October 30, 2025, the AI agent independently executed a full orientation maneuver, successfully shifting the satellite to a specified target setting by utilizing reaction wheels.

This successful test forms an integral part of the larger In-Orbit Demonstrator for Learning Attitude Control (LeLaR) project, which is mandated to engineer a state-of-the-art autonomous attitude control system. Key contributors to the LeLaR team include Dr. Kirill Djebko, Tom Baumann, Erik Dilger, Professor Frank Puppe, and Professor Sergio Montenegro. The approach employed utilizes Deep Reinforcement Learning (DRL), where a neural network autonomously learns the optimal control strategy within a simulated environment, distinguishing it from conventional fixed algorithms.

The LeLaR project has secured substantial funding, amounting to approximately €430,000, from the German Federal Ministry for Economic Affairs and Energy. Professor Montenegro affirmed that this achievement inaugurates a new phase for satellite control systems, defining them as intelligent, adaptive, and capable of self-learning. Successful implementation has the potential to foster significantly more efficient missions by minimizing the requirement for human intervention and enabling quicker responses to changing space environment conditions.

This success further establishes the University of Würzburg as a pioneer in the domain of AI-controlled space systems. The achievement follows JMU's prior work in space autonomy, notably the SONATE-2 mission. Launched on March 4, 2024, SONATE-2 was designed to train an onboard AI system to independently identify and photograph anomalies on the Earth's surface, representing a vital progression toward enhanced autonomy for small satellites. In SONATE-2, the team successfully trained a neural network in orbit, which reliably recognized non-desert features in the Sahara region, such as the Nile River and its surrounding green areas.

JMU is recognized as a prominent center for small satellite technology in Germany, having built 21 satellites, five of which are currently operational. The development of this autonomous control system is highly relevant, as AI methods may prove crucial for the future viability of spacecraft, particularly when facing unanticipated scenarios. This demonstration represents a concrete exhibition of German expertise in integrating advanced computer science with space technology, a discipline central to the university's Master of Satellite Technology program.