Por Luís Simões, Marília Simões e Nikki Grens (ML Analytics).

Este seminário apresentará uma mostra de aplicações de IA/Machine Learning no setor espacial, através do trabalho aplicado da ML Analytics.

Abstract: This seminar will present a showcase of AI/Machine Learning applications in the space sector, through the lens of ML Analytics’ applied work. Over the past decade, team members have contributed to a diverse portfolio of ML projects in space - including autonomous landing site selection, spacecraft trajectory optimization, collision-risk assessment, operations scheduling and distributed control - each illustrating how data-driven methods can augment traditional aerospace engineering and scientific workflows. A central case study will be the European Space Agency’s Ariel mission, dedicated to characterizing the atmospheres of 1,000 exoplanets. After winning the 2021 Ariel Machine Learning Data Challenge, ML Analytics has been creating a pipeline for integration of ML models into the mission’s operations, helping extract subtle atmospheric signatures from highly noisy observational data. The seminar will also discuss the practical engineering required to translate academic ML approaches into robust, operational-grade components. Topics include verification and validation practices, model interpretability, stress-testing for failure modes, and the design of “safety cages” for operational range bounding.