In this activity, a proof of concept of an attitude control system based on artificial intelligence, as an alternative to current state-of-the-art systems based on model approaches, was developed and tested in a simulator and in the engineering model of ESA's OPS-SAT.

Detecting anomalies from satellite telemetry is critical for the safe operation of that satellite. Although many approaches to autonomous onboard anomaly detection have already been proposed, most of them have so far only been tested on non-satellite or simulated data. This activity tackled this research gap and proposed an end-to-end machine learning-powered approach for detecting abnormal events in real-life OPS-SAT telemetry data, and deployed it on board a satellite.

ESA's OPS-SAT mission is a key element for operational validation in flight (or IOD/IOV) of the DeepCube service of Deep Learning at the edge that Agenium Space is preparing with the support of the GSTP programme (GSTP Make, ESA-CNES) and the R&D performed in the CORTEX project (Permanent Open Call EOEP-4, ESA/Phi-Lab, https://esacortexproject.agenium-space.com). The goal of this activity was to execute on board the inference of the simplified models defined in those projects.

The main objective of this study was to evaluate the merits of a dual-camera setup to acquire high-resolution images of identified events of iinterest, on board, through AI methods. A mission analysis of the approach for a single-satellite, dual camera setup as well as for a two-satellite (leading-trailing), single camera setup was performed. The dual-camera AI approach was demonstrated on commercially available high-performance embedded computers (Xilinx Kria SoM / Google Edge TPU SoM) through a state-of-the-art active fire detection use case.

While AI technologies have shown promise in various applications in the space industry, their adoption within spacecraft flight software has yet to happen in a significant manner. This is due to a number of factors, including the relatively low computing power available onboard traditional spacecraft, the long development process required in the industry and the lack of opportunities for in-orbit demonstrations. In ESA's OPS-SAT satellite, Airbus teams identified a unique opportunity to quickly test promising algorithms in orbit as the industry seeks to resolve the issues listed above.

Early space missions consisted of bespoke components communicating with facilities on Earth. Today this model has been updated with off-the-shelf components and reusable launch systems that lower mission cost into low-Earth orbit. Extending this model to develop a sustainable off-Earth economy requires bringing the concept of reconfigurable computing agents to deep space.

The main objective of the project was to select the right IMU for our Future GNSS Receiver Unit (GRU). Two IMU were selected from their datasheet : µ-UMI-IC-HP from Litef and STIM318 from Sensonor. These 2 IMU were submitted to hard characterization tests applicable to our Launchers (and
HAPS) application.

'Don't try this at home' (named after one of the AI ships from I.Banks Culture series) is a concept demonstrating the benefits of having a cognitive cloud computing infrastructure in space. Such an infrastructure will provide a comprehensive and scalable set of commercial services to EO sensors and space-based data providers. It enables several applications by efficiently collecting and fusing information on target objects from observations of different platforms at various wavelengths, resolutions, and times. As a result, it leads to timely, precise information.

The 'SaaSy ML' project will provide experimenters with on board Machine Learning (ML) functionalities that any app can subscribe to via a Software as a Service (SaaS) app hosted on the Satellite Experimental Processing Platform (SEPP). The ML features provided by the SaaS app will cover both training and prediction operations. The JSAT open-source java library for ML will be re-used thus making accessible over 100 training algorithms onboard a flying mission. Re-using the JSAT library will ensure the development to focus on the idea's novel aspect: a spacecraft SaaS app for in-orbit ML.