AI controller pulls off first real world satellite maneuver in orbit

A research team from Julius Maximilians Universität Würzburg has achieved a world first by successfully running an AI-based attitude controller directly in orbit aboard the InnoCube nanosatellite. Using Deep Reinforcement Learning (DRL), the AI autonomously executed precise satellite orientation maneuvers during a short orbital pass, repeatedly hitting target orientations in follow-up tests. This demonstrated that the AI controller could operate effectively under real space conditions, overcoming the common challenge of transferring behavior learned in simulation (Sim2Real gap) to the actual environment. The LeLaR project, which developed this controller, aims to create adaptive spacecraft control systems that require no human tuning and can adjust to unexpected conditions. Unlike traditional controllers that need months of expert calibration, the DRL-based system learns through extensive simulation training and adapts autonomously in orbit. This milestone, supported by the German Federal Ministry for Economic Affairs and Energy and managed by the German Space Agency (DLR), marks a significant advance toward fully autonomous space missions, especially those requiring