Machine Learning Advances Safety-Critical Aerospace Systems
Researchers from the Information Processing and Telecommunications Center (IPTC) at the Universidad Politécnica de Madrid (UPM), together with collaborators from the Instituto Universitario de Microgravedad “Ignacio Da Riva” (IDR/UPM) and the Ontology Engineering Group (OEG/UPM) have published a new book chapter exploring the use of Machine Learning techniques in safety-critical systems, with a particular focus on aerospace applications.
Published in the Springer book “Machine Learning Systems”, the chapter examines the challenges of deploying Artificial Intelligence in regulated systems where reliability, certification, and strict timing requirements are essential. While Machine Learning is increasingly adopted across engineering domains, its integration into safety-critical environments remains a significant challenge due to the rigorous verification and validation processes required by current certification standards.
The researchers analyzed European space software standards to identify the main barriers to the adoption of Machine Learning in certified systems. Building on this analysis, they developed and validated a Reinforcement Learning-based controller for the attitude control system of the UPMSat-2 satellite. The proposed solution was trained in simulation and subsequently deployed on a real-time embedded platform to assess its feasibility under realistic operating conditions.
The experimental results demonstrate that the neural network-based controller successfully stabilizes the satellite while meeting the timing and computational constraints required by safety-critical embedded systems. The work also shows that automatically generated embedded code can satisfy key software engineering requirements, including worst-case execution time and resource limitations, supporting the feasibility of integrating Machine Learning into future certified aerospace systems.
Beyond the satellite case study, the chapter contributes to ongoing research on trustworthy Artificial Intelligence by identifying practical guidelines for validating Machine Learning software under existing safety standards. The findings provide a foundation for the adoption of AI in aerospace and other safety-critical domains where dependability, transparency, and certification are fundamental requirements.
Bibliographic reference
Pérez-Muñoz, Á.G., López-García, G., García-Villoria, I., Alonso, A., Pérez, M.S., & Porras Hermoso, A. (2026). On the Use of Machine Learning Techniques in Safety-Critical Systems. In S. Liu, P. Reviriego, Z. Gao, & F. Lombardi (Eds.), Machine Learning Systems. Springer, Cham. https://doi.org/10.1007/978-3-032-17948-7_2
Ángel Grover Pérez Muñoz: GS / ORCID / LinkedIn
Alejandro Alonso Muñoz: GS / ORCID / LinkedIn
LinkedIn: https://www.linkedin.com/company/iptc-upm/
For more information: www.iptc.upm.es
Share this:
