Towards Resource Efficient Artificial Intelligence for In-Space Additive Manufacturing of Large Structures

In-space manufacturing has already achieved flight heritage through Additive Manufacturing (AM) technologies, which stand as key enablers for the next generation of space missions. Recent advances in large-format AM systems have unlocked the ability to produce structural elements of unprecedented size, but transferring this technology in space opens completely new challenges. AM platforms must be autonomous, self-adaptive, and robust to harsh environmental conditions. They must integrate intelligent sensing, monitoring, and control capabilities, enabling first-time-right production without human intervention. Resource efficient artificial intelligence (AI) is at the core of this transition, as it allows embedding intelligence into the system while meeting stringent power, memory, and computational limitations. This talk explores new solutions to tackle this issue. Starting from an analysis of possible scenarios, we investigate the suitability of AI-based monitoring methods suitable to detect – under severe resource limitations – critical defects that may possibly originate while the part is being produced. Grounding on experimental results and quantitative comparisons, the analysis highlights the potential of AI-based in-situ monitoring and quality inspection methods to enhance defect detection performance with minimal data volume and computational effort. We also highlight the critical role of new AI tools like generative AI to move from initial on-Earth developments to in-space adoption, focusing on technology development needs, open issues, and future challenges. The talk specifically addresses the convergence of in-situ and in-line sensing, data modelling, and efficient AI methods tailored to robotic extrusion of composite materials for large-scale structures. The study is the result of a de-risking activity funded by the European Space Agency, intended as a first step of a development roadmap ultimately targeting fully autonomous robotic AM systems for in-space manufacturing of large structures.