Investigations of Dempster-Shafer theory in the context of WLAN-based indoor localization
Kaseb Zadeh, Parinaz (2013)
2013
Master's Degree Programme in Information Technology
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2013-12-04
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201312121486
https://urn.fi/URN:NBN:fi:tty-201312121486
Tiivistelmä
Accurate user's locations and real-time location estimations in indoor environments, are important parameters to achieve reliable Location Based Services (LBSs).
Non-Bayesian frameworks are gaining more and more interest in order to improve the location accuracy indoors when WLAN positioning is used. The main objective of this thesis is to study the feasibility of Dempster Shafer non-Bayesian combining in the context of received signal strength (RSS)-based indoor WLAN localization.
The motivation of our work has been to look for new approaches in order to try to deal better with the incomplete or erroneous data measurements used in the training phase of any WLAN positioning algorithm. State-of-art studies show that the accuracy of mobile position estimation by WLAN localization algorithms with the Bayesian framework is not satisfactory. Thus, it makes sense to try to investigate non-Bayesian approaches and to see their usefulness in the context of WLAN localization. First, a comprehensive analysis of various DST combining rules with RSS-based positioning methods has been performed. Then, the idea has been implemented via MATLAB simulator and the outputs were compared to the Bayesian approaches. The comparison is in terms of root mean square errors, correct floor detection probabilities and error radius and we used real-field data measurements as test data. Typically, the current published research work based on non-Bayesian frameworks in the context of wireless localization is limited to fingerprinting methods. Both the fingerprinting and the path-loss model using the DST frameworks are carried out in this thesis.
The thesis results contain two parts. The first one examines the fingerprinting with various DST combination while the other one deals with the path-loss and DST combination.
The positioning accuracy estimated by Bayesian framework is compared to the DST and a high correlation between these two has been observed. As expected, the Bayesian framework results are slightly less accurate (on average) than the DST, because the DST fuse RSS from multiple access points with different beliefs or underlying uncertainty and allows the uncertainty to be a model parameter.
Non-Bayesian frameworks are gaining more and more interest in order to improve the location accuracy indoors when WLAN positioning is used. The main objective of this thesis is to study the feasibility of Dempster Shafer non-Bayesian combining in the context of received signal strength (RSS)-based indoor WLAN localization.
The motivation of our work has been to look for new approaches in order to try to deal better with the incomplete or erroneous data measurements used in the training phase of any WLAN positioning algorithm. State-of-art studies show that the accuracy of mobile position estimation by WLAN localization algorithms with the Bayesian framework is not satisfactory. Thus, it makes sense to try to investigate non-Bayesian approaches and to see their usefulness in the context of WLAN localization. First, a comprehensive analysis of various DST combining rules with RSS-based positioning methods has been performed. Then, the idea has been implemented via MATLAB simulator and the outputs were compared to the Bayesian approaches. The comparison is in terms of root mean square errors, correct floor detection probabilities and error radius and we used real-field data measurements as test data. Typically, the current published research work based on non-Bayesian frameworks in the context of wireless localization is limited to fingerprinting methods. Both the fingerprinting and the path-loss model using the DST frameworks are carried out in this thesis.
The thesis results contain two parts. The first one examines the fingerprinting with various DST combination while the other one deals with the path-loss and DST combination.
The positioning accuracy estimated by Bayesian framework is compared to the DST and a high correlation between these two has been observed. As expected, the Bayesian framework results are slightly less accurate (on average) than the DST, because the DST fuse RSS from multiple access points with different beliefs or underlying uncertainty and allows the uncertainty to be a model parameter.