Simulation-Based Neural Network Models for Improved Bluetooth Ranging and Localization
Author
Term
4. semester
Education
Publication year
2024
Submitted on
2024-05-31
Pages
41
Abstract
This thesis investigates the application of machine learning techniques to enhance the accuracy of Bluetooth-based ranging using channel sounding. Channel responses were simulated for multiple environments using the Sionna ray tracer, generating simulated Bluetooth channel sounding measurements to construct datasets. The study evaluated multiple representations for training neural networks on this simulation dataset and explored the transferability of simulation-trained models to real measurement data. The generated data demonstrated sufficient accuracy to train models that generalize effectively to real-world scenarios. Notably, the simulation-trained models achieved ranging accuracy comparable to or exceeding that of the MUSIC algorithm.
Documents