Probabilistic graphical models and Bayesian inference for RIS-aided indoor positioning
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
Auditorium IT116, Linnanmaa campus
Topic of the dissertation
Probabilistic graphical models and Bayesian inference for RIS-aided indoor positioning
Doctoral candidate
Master of Science Leonardo Tercas
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, Centre for Wireless Communications (CWC)
Subject of study
Communications Engineering
Opponent
Professor Simo Särkkä, Aalto University
Custos
Professor Markku Juntti, University of Oulu
Indoor positioning with intelligent surfaces and probabilistic models
Indoor positioning is becoming increasingly important for modern wireless systems, as many applications require precise device localization. Reconfigurable intelligent surface (RIS) technology can enhance positioning in addition to communications.
This thesis develops a framework for position estimation that provides accurate position and rotation estimates in systems incorporating RIS technology. The estimator uses probabilistic graphical models and employs Markov Chain Monte Carlo-based Bayesian inference to estimate the target coordinates. This approach incorporates prior knowledge about the system and enables estimation without requiring the simplification of complex mathematical equations.
First, the estimation performance for different numbers of measurements, antennas, and base stations (BSs) is analyzed to determine how these factors affect positioning accuracy. The combination of distance and angle measurements proves to be an effective approach for estimating the target position, although increasing the number of observations does not always improve performance.
The framework is then extended to include RIS, which provides additional geometric information about the target. In particular, RIS is shown to be crucial for accurately estimating device orientation, while also achieving very high precision in position estimation.
Finally, the approach is adapted to dynamic scenarios where the target moves over time. By incorporating an outlier detection mechanism, the method remains robust, achieving highly accurate estimates for both position and orientation.
This thesis develops a framework for position estimation that provides accurate position and rotation estimates in systems incorporating RIS technology. The estimator uses probabilistic graphical models and employs Markov Chain Monte Carlo-based Bayesian inference to estimate the target coordinates. This approach incorporates prior knowledge about the system and enables estimation without requiring the simplification of complex mathematical equations.
First, the estimation performance for different numbers of measurements, antennas, and base stations (BSs) is analyzed to determine how these factors affect positioning accuracy. The combination of distance and angle measurements proves to be an effective approach for estimating the target position, although increasing the number of observations does not always improve performance.
The framework is then extended to include RIS, which provides additional geometric information about the target. In particular, RIS is shown to be crucial for accurately estimating device orientation, while also achieving very high precision in position estimation.
Finally, the approach is adapted to dynamic scenarios where the target moves over time. By incorporating an outlier detection mechanism, the method remains robust, achieving highly accurate estimates for both position and orientation.
Created 17.3.2026 | Updated 18.3.2026