France Center

The France center draws on strong expertise in modeling science, both in mathematics and computer science, to train methodological frameworks in mathematics (optimal control, variable aggregation) and computer science (multi-scale modeling, participatory modeling-simulation). Our challenge is to produce a set of conceptual and software building blocks to promote the modeling and simulation of real phenomena in a participatory and transdisciplinary way.
In particular, we focus on the fundamental aspects of modelling, embracing the following fields (without being exhaustive) under the umbrella of multi-agent systems:
- Model coupling and co-modeling,
- Digital twin and on-line simulation, (Sok, C et al.)
- Distributed large-scale simulation on high-performance computing platforms (clusters) (Grosjean et al. 2025),
- and Participatory simulation for awareness and decision support (Serine K 2023).

In terms of applications, we respond to sustainable development issues of general interest in France and internationally, notably in the UMMISCO network and its partners such as :
- Covid 19 crisis management and, more generally, the dynamics of epidemic evolution (Cazelles et al. 2021b),
- urban redevelopment to cope with flooding and global change (Littosim project),
- biotic and abiotic soil modeling for sustainable agriculture (Monga et al. 2024),
- urban dynamics modeling and tangible interfaces (Marrakair)

Lucas Grosjean, Alexis Drogoul, Bénédicte Herrmann, Nghi Quang Huynh, Christophe Lang, et al. Distribution Model: Separation of Concerns to Facilitate the Distribution of Agent-Based Models. PAAMS 2025, Jun 2025, Lille, France. ⟨hal-05208954⟩
Sok, C., Herzog, R., & Grignard, A. Towards an Energy Twin: Simulating Global Warming Potential in Hamburg's Building Stock.
Serigne Kosso Sene. From model to decision aid through modelling: application to deficit irrigation of plants in cities, context France and Senegal. Modeling and simulation. Sorbonne University, 2023. French. ⟨NNT: 2023SORUS579⟩. ⟨tel-04496124⟩

The France center also draws on strong methodological expertise in machine learning in general and deep learning in particular. Our challenge is to develop new AI approaches capable of processing and interpreting a wide variety of data to answer various methodological research questions. Our expertise extends to a diverse range of advanced techniques such as supervised learning (Prifti et al. 2020), unsupervised learning (Affeldt et al. 2016) or self-supervised learning and raises first-order questioning on:
architectures (e.g. extending transformer architecture to multi-instance data),
performance (e.g. increase in data size), model robustness (e.g. resistance to attacks on workbooks),
model interpretability (e.g. sparse model learning (Prifti et al. 2020) or occlusion approaches),
Our research is applied in a variety of critical areas, including
healthcare, notably for the analysis of tabular and textual data (DeepIntegrOmics or DeepECG4U projects, Prifti et al. 2021),
biodiversity (eCOL+, MetaPLantcode (Ariouat et al. 2023) or AIME), and
distributed scientific computing.

Our many modeling experiences, particularly in the South, show that there is a cruel lack of data available to respond to societal challenges through modeling. While the model is a formidable tool for producing scientific data, it is also a major consumer.
This is why the France center has invested heavily in the development of low-cost environmental measurement instruments and in the design of pipelines for automated integration of multi-source and multi-scale data (point measurements on the ground, satellite measurements, drones), from measurement by sensors to real-time assimilation into the simulation model or even the digital twin.
This work has led us to develop research in the fields of:
-embedded AI (Elouiaazzani et al.)
-data assimilation (Bassirou et al.)
-multi-source data fusion (Niane 2023)
-sensor networks ( Martiny et al.)
-data qualification through continuous measurement calibration
This work is contributing to the development of low-cost, citizen-oriented sensors, and has applications in a number of topical areas:
-Air quality through the development of the QameleO sensor
-Deficit watering for optimized water use through the development of the WAOU sensor, which measures water tension in the soil (the effect of suction exerted by plants).
-Acoustic monitoring of marine protected areas through the design and production of appropriate instruments
Vehicle counting using on-board AI (I-Maroc Project)
In addition to enabling scientifically reliable measurement at low cost, these new scientific instruments (licensed under Cern OHL V3, GPL3 and Creative common) have a societal dimension: they are designed to be built in the South by fablabs and to be easily repairable.
For example, just as QaméléO was conceived at Cofab-in-Bondy, it is now being built and deployed by FabLab partners such as Blolab and Cofab-in-Dakar in several countries, including Benin, Congo, Ghana, Madagascar, Côte d'Ivoire and Senegal. This work ensures that research is carried out in the field, in close collaboration with local players, and financed by various sources (POPSU, ANR, local authorities, FID, BPIFrance...), or via industrial partnerships with Diginove, Group Tera or Urbasense.
This work has resulted in scientific publications, patents and, above all, a direct impact on :
-Research in the South, notably through participation in international research groups such as IRN Reallity,
-The Company by installing sensors in different areas, producing new data and, above all, raising public awareness of these new environmental data.
Hamza Elouiaazzani, Nicolas Marilleau, Tri Nguyen-Huu, Ahmed Laatabi and Mohamed Ait Babram “A Real-Time Vision-Based Vehicle Counting Smart Sensor on the Jetson Orin Nano” ANT’2026
Ngom, Bassirou, Moussa Diallo, Madoune Robert Seyc, Mamadou Simina Drame, Christophe Cambier, and Nicolas Marilleau. «PM10 Data Assimilation on Real-time Agent-based Simulation using Machine Learning Models: case of Dakar Urban Air Pollution Study». 2021 IEEE/ACM 25th International Symposium on Distributed Simulation and Real Time Applications (DS-RT), September 27, 2021, 1-4. https://doi.org/10.1109/DS-RT52167.2021.9576143.
Papa Massar Niane. Modeling bacterial meningitis in the Environment-Climate-Society interface using a multi-agent approach: case study in Senegal. Modélisation et simulation. Sorbonne Université; Université Cheikh Anta Diop (Dakar, Senegal ; 1957-..), 2023.
Martiny et al, «Quality of Air Module for Environmental Learning Engineering and Observation Network (QameleON-Dijon): a dense network of air quality measurements in Dijon».

The link between research and society is at the heart of the concerns of Centre France researchers. This is reflected in the development of awareness-raising and decision-making tools for society, government bodies and industry.
The “complex system” model representing a phenomenon (as described in the first UMMISCO theme) cannot, unfortunately, be put directly into the hands of populations and decision-makers. It has to be encapsulated in another (end-user oriented) model, ensuring the post-processing and formatting of the data resulting from the simulations of a “complex system” model.
We are developing methodological and technological frameworks for..:
decision support (Serigne Kosso Sene 2023)
participatory science (Laatabi et al. 2022)
awareness (Becu et al. 2019)
scientific mediation (Minh Duc et al. 2020)
This research is being applied in various fields, including
urban dynamics (Projet I-Maroc, Marrack'Air, HoanKiem'Air or Dij'Air)
irrigation deficit (Anr labcom Waqatali project)
flooding in Madagascar (Didem project)
..
This research is carried out in partnership with local authorities and/or entrepreneurs.

Serigne Kosso Sene. Du modèle à l'aide à la décision par la modélisation : application à l'irrigation déficitaire des végétaux en ville, contexte France et Sénégal. Modeling and simulation. Sorbonne University, 2023. French.
Minh Duc, Pham, Kevin Chapuis, Alexis Drogoul, et al. «HoanKiemAir: simulating impacts of urban management practices on traffic and air pollution using a tangible agent-based model». 2020 RIVF International Conference on Computing and Communication Technologies (RIVF), October 2020, 1-7. https://doi.org/10.1109/RIVF48685.2020.9140787.
Laatabi, Ahmed, Nicolas Becu, Nicolas Marilleau, et al. «LittoSIM-GEN: A Generic Platform of Coastal Flooding Management for Participatory Simulation». LittoSIM-GEN: A Generic Platform of Coastal Flooding Management for Participatory Simulation". Environmental Modelling & Software 149 (March 2022): 105319. https://doi.org/10.1016/j.envsoft.2022.105319.
Becu, Nicolas, Marion Amalric, Brice Anselme, et al. «Participatory simulations with decision makers on coastal flooding prevention: what did they learn? Participatory simulations with decision makers on coastal flooding prevention: what did they learn?» International Simulation and Gaming Association Conference (Warsaw, Poland), August 2019. https://hal.archives-ouvertes.fr/hal-02431600.