A team of scientists from the Julius Maximilian University of Würzburg in Germany has successfully tested AI-based satellite attitude control for the first time. The experiment involved the InnoCube 3U nanosatellite.
Between 11:40 and 11:49 hours on October 30, the AI agent performed all orbital navigation maneuvers. He moves the satellite from the current home position to the predetermined target position using the reaction wheel. In subsequent tests, he “successfully and safely maneuvered the satellite into position. Scientists say they have taken a decisive step toward autonomy in space.”
Attitude controllers help stabilize satellites in orbit and prevent them from falling. They are also needed to point the spacecraft in the desired direction.
Usually, such devices operate on the basis of traditional fixed algorithms. The authors of the new project used deep reinforcement learning (DRL). In this case, the neural network will independently choose the optimal action strategy in the environment.
Traditional controllers often require time-consuming manual parameter settings. For engineers, this process can take months, even years. AI automates all processes and opens up the possibility of creating controllers that can independently adapt to changes.
The test showed that AI can not only work in simulation mode, but also perform accurate automatic operations in real conditions. In the future, scientists plan to work on improving the system, Phys.org reported.
Artificial intelligence is not only used in the space industry but also in other fields, especially in medicine. Thus, for the first time, AI created the genome of a virus capable of destroying bacteria. This development could help in the fight against superbugs – pathogens that have developed resistance to modern drugs.
