La modélisation des transports à la SNCF

Organisés par le CERMICS et le LVMT, les séminaires Data Transitions visent à rendre accessible la recherche scientifique de pointe dans le domaine du numérique et de la donnée. Ils sont ouverts à tous les esprits curieux, juniors comme seniors, experts comme néophytes.

Le prochain séminaire "Working on real-world transportation modeling problems: examples at SNCF" sera animé par Clément Mantoux (ingénieur R&D SNCF), jeudi 11 décembre à 11h45 (salle F107, Bâtiment Coriolis).

Working on real-world transportation modeling problems: examples at SNCF

There is a rich literature on modeling transportation systems to further their understanding and optimize their capacity. This knowledge, however, only proves useful as far as the available data can get us to. In practice, hurdles such as data availability and quality or operational constraints often prevent us from achieving these goals. The complexity of railway operations is reflected in the data they produce, which is often difficult to model and visualize at scale. Careful project planning and algorithm design may help circumvent some of these issues, and leverage the versatility of modern machine learning and optimization techniques. This talk dives into examples of such projects developed at SNCF, which rely on large-scale data management systems to enhance passenger train delay prediction and travel time estimation for freight wagons.