• Kasper Ramsgaard-Jensen
4. semester, Mathematical Engineering, Master (Master Programme)
This master's thesis is written in collaboration with RTX A/S and addresses indoor radio localization in a DECT network using \textit{received signal strength} (RSS). In order to model the localization problem and the relationship between RSS and distance an extensive system specification has been employed. Based on the findings, an iterative moment matching variational message passing algorithm has been derived. It has proven advantageous to restrict the messages to the exponential family of probability distributions and employing moment matching for deriving simple message approximations. The performance of the algorithm has been tested with simulated and real data through Monte Carlo simulations. If the base stations are assumed to be sector antennas, the algorithm quickly converges to a final position in the vicinity of the true position. Those estimates which do not agree with the true position are caused by inferior RSS measurements and due to the fast convergence, other base stations may not contribute. Sorting the base stations with respect to the RSS does, however, yield mean error distances of approximately 5 meters in simulated environments and less than 5 meters using data from a measurement campaign. As the explored methods are able to infer latent distance information from RSS measurements the results of this project can impact the future direction of indoor localization based on RSS.
LanguageEnglish
Publication date2019
Number of pages123
External collaboratorRTX Telecom
Finn Hebsgaard Andreasen fa@rtx.dk
Information group
ID: 305271174