Lecture
From data to discoveries: machine learning and optimization in space
- Marcus Märtens (European Space Agency)
- Date
- Tuesday 25 July 2023
- Time
- Address
-
Newly emerging technologies from computer science and artificial intelligence have demonstrated high potential to transform our society and whole industries. This talk will investigate, how such technologies can be successfully leveraged for the concrete application domain of space, beginning with the collection and curation of datasets and the generation of scientific value through open competitions.
Following the recent success of several public challenges organized by the European Space Agency, we will show how neural networks are capable of estimating spacecraft poses, sharpen satellite imagery, steering spacecrafts and perform on-board classification of landcover types.
In a second part, multiobjective search techniques will be reviewed in the context of trajectory optimization, showcasing their
ability to explore large and complex decision spaces with interdependent objective functions. Lastly, we will show how symbolic regression techniques can construct interpretable models in the form of mathematical expressions, improving our understanding of the data we collect.