International prize for an innovation in electromagnetic mapping based on artificial intelligence

This prize rewards the efforts made to advance research into the reconstruction of electromagnetic fields in order to assess levels of exposure to them. This work was carried out with the financial support of the Lille European Metropolis (MEL).

In this article, entitled ‘GLIP: Electromagnetic Field Exposure Map Completion by Deep Generative Networks’, the complex challenge of reconstructing electromagnetic field exposure maps was tackled using new artificial intelligence tools. This mapping is based on data collected in a scattered manner by sensors installed by the MEL in the Wazemmes and Euratechnologies districts of Lille. This innovative method solves an ill-conditioned inverse problem using only GANs (Generative Adversarial Networks). The proposed method does not require explicit training or complete reference maps. Nevertheless, by integrating local data from sensors as Local Image Prior ( LIP ) and relying on deep generative convolutional networks, this approach named GLIP (Generative Local Image Prior) overcomes the limitations of traditional methods and allows accurate estimates to be obtained, even with limited data.

Remarkably, the method demonstrates its effectiveness in accurately predicting maps with a minimum of input data, encompassing less than 1% of the 1km² area of the reference map. This means that the field is not too complex to predict, and that the method makes it possible to reduce sampling of this area considerably. Indeed, simultaneous measurement over the entire area of interest is impossible. In addition, it is no longer necessary to simulate reference maps, which saves time and expertise. Rather than relying on simplistic assumptions about signal propagation, GLIP skilfully estimates the complexity of propagation in an urban environment, taking due account of the characteristics of the buildings.

The potential applications of this technique go well beyond the framework of exposure to electromagnetic fields, enabling the reconstruction of non-accessible data sets of all types from a small database of measured data.


More informations :

https://pimrc2024.ieee-pimrc.org
Together with the International Conference on Communications (ICC) and the Global Communications Conference (GlobeCom), the symposium has become one of the main conferences of the IEEE Communications Society (ComSoc) in the field of wireless communications and networks.


Mohammed Mallik, Davy Gaillot and Laurent Clavier respectively from : the University of Lyon – INSA, the University of Lille – IEMN and IMT-NE – IEMN received the prestigious Best Paper Award at the latest IEEE Annual International Symposium on Personal, Indoor and Mobile Radio Communications (IEEE PIMRC 2024) held in Valencia, Spain from 2 to 5 September 2024.