Congratulations Yuanxin!
On the 12th of June, Yuanxin Pan successfully defended his Ph.D. thesis entitled "Characterization of Reflected and Low-cost GNSS Signals with Machine Learning".
Yuanxin is the third doctoral student to join the Space Geodesy group in 2021, one year after the group was established. His doctoral research was supervised by Prof. Benedikt Soja and Prof. Gregor Möller.
During his Ph.D., Yuanxin focused on the application of machine learning (ML) techniques for GNSS data processing, particularly in two key areas: multipath modeling for geodetic stations and non-line-of-sight (NLOS) signal classification in urban environments. His work demonstrated that the ML-based approaches can outperform traditional methods in mitigating multipath errors. Notably, he developed long-term multipath models that remain valid over several years — an achievement with practical benefit for operational data processing of geodetic GNSS stations. For NLOS signal classification, Yuanxin showed that transfer learning can effectively enhance the generalizability of ML models across different datasets, which is important for real-world applications. He was also one of the major contributors to the ESA-funded CAMALIOT project, where he developed an ML-based pipeline for automatically selecting crowdsourced smartphone GNSS data and a dedicated method for estimating high-precision zenith total delays (ZTD) from these low-cost observations. Overall, Yuanxin’s work exemplifies the potential of ML to advance GNSS data analysis and has made a valuable impact on this field.
Following his successful defense, a celebratory apéro was held with the examination committee, colleagues, and friends to mark this important milestone.
