Proceedings of the 3rd Conference on AI in and for Space (SPAICE 2026)

Abstract

We are excited to release the preliminary proceedings of SPAICE 2026, the third edition of the European Space Agency’s conference on AI in and for Space, held at ESTEC in Noordwijk, the Netherlands, from 21 to 23 October 2026. When SPAICE began in 2024, the question in the room was still largely whether AI methods could earn a place in the space engineering toolbox. Two years later, the question has drastically changed to how: how to qualify a neural network for flight hardware, how to quantify what a learned model does not know, how to combine decades of physical modelling with data-driven methods instead of replacing it, and how can we trust neural models in the control of the spacecraft. The contributions in this volume reflect that shift. Many of them deal with topics such as FPGA deployment, quantisation robustness, fault tolerance under radiation, and models that run on qualified processors. These sit alongside physicsinformed and differentiable approaches to orbit determination, gravity inversion and guidance, and a growing body of work on trustworthiness and uncertainty. This year we received 147 submissions. After a double blind peer-review by our Scientific Committee, 79 made it to the program: 42 oral presentations, given in two parallel tracks, and 37 posters. The programme covers six themes: Onboard AI, Spaceflight Dynamics, Guidance, Navigation & Control, Operations, Earth Observation, and Astronomy. Talks range from continual adaptation for space robots and reinforcement learning for planetary landing, to transformer-based generative models of planetary systems, anomaly detection in spacecraft telemetry, and retrieval-augmented language models for mission operations. This breadth is intentional. SPAICE exists to put the people developing new learning methods in the same room as the people who fly them and the scientists who depend on their outputs.We hope this volume serves both as a record of the current state of AI for space and as a starting point for the collaborations it inspires.