SONATE-2 – A technology demonstration mission for artificial intelligence

Tobias Greiner

JMU Würzburg

The increasing amount of data produced by sensors combined with low download rates is a challenge for nanosatellites. On-board data processing and classification can be used to reduce the amount of data to be downloaded. This "edge computing" approach is becoming increasingly common in various terrestrial applications and embedded devices. Recent missions have used artificial intelligence (AI) models pre-trained on the ground for on-board image analysis. The 6-U SONATE-2 technology demonstration mission extends this approach by introducing on-board training capabilities, enabling new applications such as on-board model generation and improvement. To achieve this, the platform will use an Nvidia Jetson module with integrated CPU and GPU for high flexibility and performance. The mission will verify the processing capabilities of the platform in space using multiple cameras in the visible and near-infrared spectrum. Images captured by these cameras will be analysed using conventional image processing and AI models. The experiments will cover a wide range of applications including lightning detection, image segmentation, object detection and anomaly detection.