Fusion of model-driven and data-driven models for EMF exposure prediction

Organisation/Company

Télécom Paris

Research Field

Computer science » Database management

Researcher Profile

First Stage Researcher (R1)

Positions

PhD Positions

Country

France

Application Deadline

17 Feb 2025 - 00:00 (Europe/Paris)

Type of Contract

Temporary

Job Status

Full-time

Offer Starting Date

1 Apr 2025

Is the job funded through the EU Research Framework Programme?

Not funded by a EU programme

Is the Job related to staff position within a Research Infrastructure?

Yes

RÉSUMÉ EN 30 LIGNES :

Subject: Fusion of model-driven and data-driven models for EMF exposure prediction

Scientific objectives: We aim to develop a novel framework for predicting spatial mapping of electromagnetic field (EMF) exposure by integrating data-driven models (leveraging machine learning techniques) with model-driven approaches (based on physical principles of EMF propagation). By combining the strengths of these two paradigms, the thesis aims to enhance the precision and interpretability of EMF exposure mapping. It will provide reliable exposure prediction for public health, and the future wireless communication infrastructure.

State of the art: The data-driven approach in EMF exposure prediction, as demonstrated in [1,2], treats the estimation problem as a black-box mapping, relying solely on empirical data. The neural networks act as general models. In contrast, Physics-Informed Neural Networks (PINNs) have garnered increasing attention for their ability to embed physical laws directly into neural networks. For instance, [3] illustrates the application of PINNs in path loss estimation, effectively improving prediction accuracy and interpretability.

Innovative nature: It fuses data-driven and model-driven approaches for EMF exposure mapping prediction given the fact that most of work in AI-assisted EMF prediction is data-driven. The physical nature of EMF propagation is not exploited enough. This thesis will leverage the strengths of both paradigms to overcome the limitations of traditional methods. 

Approach + Expected results: In this thesis, the PhD candidate will begin by reviewing classical and widely-used models for both data-driven and model-driven approaches to EMF exposure prediction. Following this, the candidate will examine existing PINN models [4], with a particular emphasis on identifying suitable physical models relevant to EMF exposure propagation. Based on these insights, the candidate will investigate and propose novel fusion models that integrate data-driven and model-driven methodologies for exposure prediction. The proposed models will first be tested on simulated datasets, such as using test functions or ray-tracing simulators, then on real-world data.

Perspectives: The outcome of this thesis can support epidemiological studies exploring the relationship between EMF exposure and potential health outcomes. On the other hand, it can be extended to predict RF sensing for emerging wireless technologies, including 6G and IoT networks.

industrial and societal impact: This thesis will contribute to the environmental study of RF EMF exposure and its impact on human health by improving the accuracy and explainability of EMF exposure mapping in real-world environments. It aligns with key strategic themes from IMT, including health prevention and the integration of digital data and AI technologies.

Ref:

[1] Wang, S., & Wiart, J. (2020). Sensor-aided EMF exposure assessments in an urban environment using artificial neural networks. International Journal of Environmental Research and Public Health, 17(9), 3052.

[2] Chikha, W. B., Wang, S., & Wiart, J. (2023). An extrapolation approach for RF-EMF exposure prediction in an urban area using artificial neural network. IEEE Access, 11, 52686-52694.

[3] Jiang, F., Li, T., Lv, X., Rui, H., & Jin, D. (2024). Physics-informed neural networks for path loss estimation by solving electromagnetic integral equations. IEEE Transactions on Wireless Communications.

[4] Raissi, M., Perdikaris, P., & Karniadakis, G. E. (2019). Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational physics, 378, 686-707.

E-mail

shanshan.wang@telecom-paris.fr

Research Field

Computer science » Database management

Education Level

Master Degree or equivalent

Number of offers available

1

Company/Institute

Télécom Paris

Country

France

Geofield

City

PALAISEAU

Website

https://www.telecom-paris.fr/

Street

19 Place Marguerite Perey

• X (formerly Twitter)
• Facebook
• LinkedIn
• Whatsapp
• 
  More share options
  • E-mail
  • Pocket
  • Viadeo
  • Gmail
  • Weibo
  • Blogger
  • Qzone
  • YahooMail