• Citer

Stochastic Geometry Analysis of Electromagnetic Field Exposure in Cellular Networks

par Gontier, Quentin
Président du jury Quitin, François
Promoteur De Doncker, Philippe
Co-Promoteur Oestges, Claude
Publication Non publié, 2025-02-24

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Each new generation of wireless networks reignites discussions surrounding electromagnetic field exposure (EMFE) levels. With every technological advancement introduced by successive network generations, fresh studies become essential. The fifth generation of networks (5G) is notable for increasing the variability of EMFE levels due to the implementation of dynamic beamforming, the anticipated proliferation of connected devices, and the deployment of low-power small cells. To guide deployment effectively, a network-level analysis is required to provide network planners with actionable insights into design parameters, such as base station density. However, existing assessment methods, including in-situ measurements and deterministic models, fail to account for network randomness or the effects of large-scale infrastructure changes and emerging technologies. Stochastic geometry overcomes these limitations by modeling base stations and user equipments as random point processes, facilitating the statistical characterization of EMFE through analytical and numerical approaches.The research begins by investigating the application of stochastic geometry to wireless networks, highlighting its capacity to derive performance metrics across diverse spatial configurations. Calibration of the homogeneous Poisson point process with empirical data validates the model and evaluates the impact of network densification strategies on EMFE. More advanced models, such as the β-Ginibre point process and the inhomogeneous Poisson point process, enhance the analysis by capturing spatial regularity and variations in density. For instance, the β-GPP improves EMFE predictions by up to 7.3% compared to the H-PPP, while I-PPP analyses identify EMFE hotspots in high-density regions, emphasizing the importance of localized evaluations.This thesis also introduces the multi-cosine antenna gain model for dynamic beamforming networks, offering significant improvements in the accuracy of EMFE and SINR predictions. A novel metric is presented to jointly optimize coverage for active users and EMFE for inactive users. The meta distribution of EMFE further provides detailed insights into spatial and temporal variations, enabling the definition of reduced-exposure areas and ensuring compliance with stringent EMFE regulations.For uplink scenarios, the analysis incorporates a new model that combines clustered user equipment distributions with truncated power control to evaluate uplink and downlink EMFE. The findings confirm that uplink EMFE is typically lower than downlink EMFE but highlight the potential for significant peaks under specific conditions, warranting regulatory attention to instantaneous exposure.By addressing key research questions with both mathematical and empirical rigor, this thesis establishes a foundation for using stochastic geometry in the design of future wireless networks, balancing performance, user safety, and regulatory compliance.