Abstract

This master project presents a systematic approach to generating position-labeled Channel Impusle Response (CIR) datasets for radio localization using deterministic Ray-Tracing (RT) techniques. The approach integrates advanced ray tracing simulation tools, particularly NVIDIA’s Sionna RT and Blender, to model electromagnetic wave propagation, including phenomena such as reflection, diffraction, and scattering. By constructing detailed scene environments and simulating the multipath propagation of radio waves, the project aims to produce position-labelled CIR datasets that accurately reflect real-world propagation conditions. An outdoor scenario of the NEOS building in Bremen, Germany was modelled to demonstrate the methodology. The resulting dataset shows the position-dependent variations in CIR. This is essential for improving the accuracy of machine learning–based localization systems. The generated dataset is available in Zenodo and provides a resource for future research in radio localization.