Air pollution in urban areas constitutes a major public health concern, with at least 238,000 premature deaths per year in Europe due to fine particles PM 2.5 . At the city or district level, it is necessary to better estimate and predict the levels of pollution citizens are exposed to. Current measurement resources consist of a small number of fixed stations, whose measurements are assimilated into a simplified model of the city (street network). In recent years, the miniaturization of sensors has progressed considerably, and it has become realistic to design estimation approaches based on a large number of sensors scattered in the environment (crowdsensing), for example, the estimation of pollution sources in an urban environment.
The question is: where should the sensors be placed in a city district to maximize the estimation quality?
Previous work has considered the assimilation of measurements from a network of low-cost fixed sensors for estimating air pollution sources in urban environments [1,2]. The geometry of buildings results in a very heterogeneous domain, with complex boundary conditions [3]. We have developed a reduced finite-difference approach supported by radial basis functions for an advection-diffusion partial differential equation (PDE) model of atmospheric dispersion.
Various sensor placement approaches exist in the literature. For the most part, they are methods developed specifically for a given application, exploiting the spatial concentration gradient and knowledge of the local wind field. For infinite-dimensional systems described by partial differential equation (PDE) models, promising approaches have been proposed [4,5,6].
The aim of the internship is to develop a generic approach to optimal sensor placement in the context of atmospheric dispersion of pollutants in urban areas. This approach will be developed on simplified 2D/3D case studies, and tested on a city district in Grenoble and/or Paris [3].
[1] R. Lopez-Ferber, S. Leirens, D. Georges, “Source Estimation: Variational Method versus Machine Learning Applied to Urban Air Pollution”, IFAC Workshop on Control for Smart Cities, CSC 2022
[2] R. Lopez-Ferber, D. Georges, S. Leirens, “Fast Estimation of Pollution Sources in Urban Areas Using a 3D LS-RBF-FD Approach”, submitted to the European Control Conference 2024
[3] M. Mendil, S. Leirens, P. Armand, C. Duchenne, “Hazardous atmospheric dispersion in urban areas: A Deep Learning approach for emergency pollution forecast”, Environmental Modelling & Software, Volume 152, 2022
[4] D. Georges, “Optimal Location of Mobile Sensors for Environmental Monitoring”, European Control Conference (ECC), July 17-19, 2013, Zürich, Switzerland
[5] D. Georges, “Optimal Location of a Mobile Sensor Continuum for Environmental Monitoring”, 1st IFAC Workshop on Control of Systems Governed by Partial Differential Equations, September 25-27, 2013, Paris, France
[6] VT Nguyen, D. Georges, G. Besançon, “Optimal sensor location for overland flow network monitoring”, 3rd Conference on Control and Fault-Tolerant Systems, Barcelona, Spain, Sept. 7-9, 2016
Profile:
- Applied mathematics profile
- with a taste for physical models and numerical methods
Skills:
- PDE models, optimal control, optimization
- Python programming
- Autonomy and adaptability
- Good writing skills