Year: 2025
Type: dissertation
Abstract: Localization is a key enabler of solutions and services in modern systems, with application areas ranging from industrial internet-of-things (i-iot) to emergency services.In addition to positional information, reporting the orientation of user equipment (ue) can improve high-frequency communication and elevate next generation use cases such as augmented reality and context-aware services.In indoor localization, global navigation satellite systems (gnss) lacks coverage, and geometry-based approaches suffer in cluttered environments with non-lineof-sight (nlos) and multi-path signal propagation.To overcome the challenges posed by the indoor localization problem, fingerprinting-based approaches can relate a device's ground truth location to the signature of the received signal between the device and base station (bs).This study investigates orientation estimation of ue in an indoor scenario.In particular, artificial neural networks (ann) are employed to map from channel impulse responses (cir) to ground truth orientation.Multiple ann variants are implemented and compared, performing both orientation estimation and joint position and orientation estimation.Furthermore, the effect of replicating the cir-based input features using a simple auto-encoder is studied.iii I would like to thank my supervisor from the company, Yuxin Zhao, for her support and guidance throughout the thesis.Additionally, I would also like to thank my supervisor, Yuhang Deng, for his invaluable insights and assistance
Source: DiVA (Linkoping University)
Author: Henriksson, Lorentz
Institution: Linköping University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7073934399
