Bringing artificial intelligence to orbit: paving the way for the standardisation of autonomous satellites and constellations to ensure sustainable orbital lifecycles and impactful Earth observations.
As the number of satellites in orbit continues to multiply, critical questions arise regarding their lifecycle and utility.
Our research aims to enhance satellite cooperation, improving their capacity to fulfill missions and strengthening our control over their orbital lifespan.
With the milestone of in-orbit machine learning training already achieved, supporting the generalization and adaptation of these models to the specific context of space should now be driven by robust standards.
Building on the group’s previous work, federated learning is a machine learning approach that enables the training of a network of devices in bandwidth-limited contexts, such as orbit. Supporting standards to generalize this approach would unlock new possibilities for collaborative satellite missions and accelerate the transition toward autonomous orbit management.
As satellites provide the foundation for environmental studies, enabling satellites to process large datasets autonomously on-board can significantly improve monitoring coverage. This shift could enable faster response times to critical planetary changes.
