Deep learning on graphs for mobility-flows prediction and air quality in urban area

The work targeted in this thesis is part of the current global context of reducing the energy footprint of mobility through innovative techniques. Within IFPEN, work is already being carried out on eco-driving and eco-routing, but also on mobility modeling via hybridization approaches between classical physical models and new mobility data (floating mobile data, floating car data, census data, survey data on people's mobility, telephone data, etc.). This thesis topic is in line with these activities, wishing to model and predict mobility flows within an urban network from multi-source mobility data to set up the necessary tools to evaluate, or even design, public policies impacting air quality.
The thesis work will be based on disciplines such as graph theory, artificial intelligence (neural networks) and deep learning to address the estimation of mobility flows in an urban network. IFPEN already has in its possession dynamic multi-modal mobility data via the application Geco air. The objective of the thesis will be to set up a methodology to merge dynamic mobility data with "static" survey and urban network topology data, which will be used as inputs for a "graph neural network" to predict mobility flows on the territory. A major challenge of the thesis will be to ensure the spatial extrapolation capabilities of the method as well as the consideration of physical constraints on the prediction of flows inspired by physical mobility models.

Keywords: mobility flow, graph theory, artificial intelligence, neural network, deep learning

• Academic supervisor    NAJMAN Laurent, Université Gustave Eiffel / A3SI ESIEE Paris – laurent.najman@esiee.fr 
• Doctoral School    MSTIC : Mathématiques et Sciences et Technologies de l'Information et de la Communication
• IFPEN supervisor    CHATAIGNON Aurélie, IFPEN, Sciences et Technologies du Numérique, aurelie.chataignon@ifpen.fr 
• PhD location    IFP Energies nouvelles, Rueil-Malmaison, France  ESIEE, Noisy-Le-Grand, France
• Duration and start date    3 years, starting in November 2023
• Employer    IFP Energies nouvelles, Rueil-Malmaison, France
• Academic requirements    University Master degree in applied mathematics, computer science, data science
• Language requirements    Fluency in French or English, willingness to learn French
• Other requirements    Deep learning, graph theory, programming skills (Matlab/Python)