Möt vår utbytesstudent från Tammerfors universitet

Läs hela intervjun på engelska:

Tell us about the project you are working on, what are you going to do?

I am an MSc student of IT in communication systems and networks. My current thesis project is about channel estimation and the application of Machine learning (ML) and Artificial intelligence (AI) in this area, focusing on 5G new radio indoor factory (InF.) positioning. We know that multipaths, i.e., delayed replicas of received signals in the user equipment (UE) such as a mobile phone highly degrade the positioning accuracy. The reason for this phenomenon is that obstacles in a wireless propagation environment such as reflections from the buildings in outdoor scenarios or scattering from mobile equipment in an indoor factory scenario constructively or deconstructivity distort the received signal which is used for positioning and localization purposes. In other words, non-line of sight signals reduces the accuracy of UE location in all radio access technologies (RATs) such as GNSS or 5G cellular networks. As a result, we are trying to deploy ML/AI for training a data-driven model for the identification and mitigation of non-line of sight (NLOS) 5G NR positioning reference signals (PRSs) to improve the UE positioning accuracy, below 1 meter of UE position uncertainty.

Why did you decide to come here to Mid Sweden University?

Prof. Mikael Gilund at Mid Sweden University introduced me to SCN research activities in 2022 via LinkedIn. Afterward, I was informed that there are a bunch of amazing research projects in this research group, and I asked him if there is a proper project for me to complete my MSc program. As I was working as a research assistant in the field of hybrid positioning in the first year of my MSc studies at Tampere universities, my home university, He and Prof. Aamir Mahmood kindly offered me that I can come here as an exchange student to further my previous studies in this area. As I liked to improve my academic network, get familiar with research projects at Mid Sweden University, collaborate with high rank researchers in Sundsvall, and experience living in a new city and atmosphere, I accepted this offer and started to work on my thesis project here.

What have you done so far in this project?

Currently, we have made a Python script for simulating 5G NR PRSs in the time and frequency domain, and we extract distinctive features that describe channel characteristics such as channel impulse response or power delay profile. The raw dataset on channel response is provided by a ray tracing software, called Winprop for InF. propagation scenarios and we overviewed the process of simulating InF. propagation environment with this software. As the main InF. propagation scenario is explained by 3GPP technical specifications (TSs) to enhance positioning accuracy by implementing ML-assisted models, we also overviewed 3GPP specification and LOS/NLOS signal identification results reported by companies such as Ericsson. These two blocks enable us to simulate the PRS signal as it is sensed by a UE. In addition, we implemented an LSTM/CNN ML model to classify LOS/NLOS signals at UE. we are going to find new features or NN layers to improve LOS/NLOS classification accuracy, while we are investigating different methods for performance analysis and comparing our channel propagation dataset with other tools such as Quadriga channel modelling software or other ML toolboxes. The next phase of our project is to implement a time-based and carrier-based positioning algorithm with different properties such as continuous phase measurement to compare the performance of direct 5G positioning with ML-assisted positioning.