Photo: Roman Klus
In his doctoral dissertation, Jaakko Pihlajasalo demonstrates that machine learning can effectively mitigate unmodeled errors and hardware distortions in satellite orbit predictions and 5G receiver signal processing. The study shows that convolutional neural networks and other learning-based methods enhance the accuracy of traditional physics-based models and compensate for various impairments when the training data properly represents the non-idealities. The findings highlight the practical added value of machine learning even in fields where computationally efficient physical models are already well established. The research paves the way for more accurate satellite positioning, more reliable wireless communication, and more cost-efficient receiver solutions.
The doctoral dissertation of M.Sc. (Tech) Jaakko Pihlajasalo in the field of applied mathematics titled Impairment Mitigation with Machine Learning for Satellite Ephemeris Extension and 5G Physical-Layer Communications will be publicly examined at the Faculty of Information Technology and Communication Sciences at Tampere University on 20 March 2026.
The Opponent will be Professor Laura Ruotsalainen from the University of Helsinki. The Custos will be Senior University Lecturer Simo Ali-Löytty from the Faculty of Information Technology and Communication Sciences.