Application of Machine Learning techniques for indoor localization of connected devices in the context of future 5G. ; Application des techniques de « Machine Learning » à la géolocalisation « indoor » des objets connectés dans le contexte de la future 5G

With the introduction of the new generation of '5G' mobile standard, it will be possible to propose several services which can be divided into 3 groups: the enhanced mobile broadband (eMB), the ultra-reliable low-latency communications (uRLLC) and the massive machine type communications (mMTC). Then, the indoor localization will be different depending on the framework which is unavoidable to choose properly the well-fitted solution. This thesis consists in studying indoor localization solutions for mMTC expecting high battery life. Moreover and with the exponential increasing number of low-energy connected devices around the world, the ambiant connectivity becomes a major requirement which has to be taken into account in our approach. Hence, the thesis deals with a plethoria of costraints such as a high probability to have a target communicating with only one access point, an unknown antenna elements geometry, a short bandwidth and connected devices respecting low-energy communications. Furthermore, the signals in indoor environments suffers from the multipath, the modifications of studied area topology and moving people. Finally and to deal with the conext, the thesis will study the indoor localization based on fingerprinting and channel state information (CSI) with the support of Machine Learning techniques. ; L'introduction de la nouvelle génération de standard mobile '5G' va permettre l'expansion de divers services qui peuvent diviser en 3 catégories: la connectivité mobile améliorée, les communication ultra-stable de faible latence et les communications massives des machines. De ces 3 catégories, la géolocalisation des objets ne sera pas la même et il est nécessaire de connaître le contexte technique pour définir le choix des solutions de géolocalisation. De ce fait, cette thèse propose des solutions de localisation en intérieur pour les objets massivement connectés et à faible consommation ce qui s'inscrit donc dans le contexte 5G. De plus, et avec la multiplication exponentiel des objets connectés à ...