Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems
Indoor positioning technology is becoming increasingly influential in indoor applications, akin to how Global Navigation Satellite Systems revolutionized outdoor navigation. One of the primary challenges in indoor settings is multipath propagation, where wireless signals encounter reflections, diffr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en |
| Published: |
2025
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/29009/2/Hybrid%20indoor%20positioning%20utilizing%20multipath-%20assisted%20fingerprint%20and%20geometric%20estimation%20for%20single%20base%20station%20systems.pdf http://eprints.utem.edu.my/id/eprint/29009/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1859055035512520704 |
|---|---|
| author | Manap, Zahariah |
| author_facet | Manap, Zahariah |
| author_sort | Manap, Zahariah |
| building | UTEM Library |
| collection | Institutional Repository |
| content_provider | Universiti Teknikal Malaysia Melaka |
| content_source | UTEM Institutional Repository |
| continent | Asia |
| country | Malaysia |
| description | Indoor positioning technology is becoming increasingly influential in indoor applications, akin to how Global Navigation Satellite Systems revolutionized outdoor navigation. One of the primary challenges in indoor settings is multipath propagation, where wireless signals encounter reflections, diffractions, and scattering, yielding less accurate position estimates. Rather than combating multipath errors, this thesis leverages them as they encapsulate how radio waves interact with the environment that stores the position-related information of the base station (BS) and the mobile station (MS). This thesis proposes a hybrid indoor positioning method for single base station systems by jointly utilizing the fingerprinting method and geometric multilateration. The proposed method leverages room geometry and takes advantage of the multipath signal propagation to construct multiple virtual base station system model. This is done based on the concept of mirror image to determine potential virtual base stations (VBSs) with respect to the reflection surfaces where the multipath rays bounce off before arriving at the MS. The methodological framework of the proposed method comprises two main phases which are fingerprinting phase and position estimation phase. In the fingerprinting phase, classifiers are trained in two stages, each to predict the MS regions and reflection surfaces, respectively. The key attributes that establish the classification learning sessions are the channel parameters extracted from the ray tracing generated multipath signals. The channel parameters such as received power, time of arrival, and angle of arrival are used as fingerprint features that act as predictors in both learning sessions. The performance of the selected trained classifiers is evaluated based on the accuracy, precision, sensitivity and the F1-score. The results show that Coarse K-Nearest Neighbours is the optimal classifier that predicts MS regions at an exceptionally high accuracy, while Support Vector Machine Kernel is the optimal classifier that perfectly predicts the reflection surfaces. In the position estimation phase, a novel Geometric Random Sample Consensus (Geometric-RANSAC) multilateration method is proposed by optimizing the MS position estimate over several potential position estimates calculated using regional 3D geometric equations. When compared to a simple fingerprinting method, the median error for the hybrid method is observed to be 4.06 cm, which is substantially lower than the 538 cm median error of the fingerprinting method. Furthermore, 95% of the MS’s positions are estimated with less than 7.86 cm of error, in contrast to the 773 cm 95th percentile error exhibited by the fingerprinting method. Notably, the proposed Geometric-RANSAC exceptionally outperforms various least squares methods by achieving a median distance error and the 95th percentile at 4.06 cm and 7.86 cm, respectively. Despite the reduced robustness of the Geometric-RANSAC algorithm when applied to scenarios with a limited number of VBSs, its exceptional accuracy compensates for this limitation. This study makes a significant contribution by introducing a hybrid method that leverages multipath signals to achieve centimeter-level accuracy in indoor positioning systems. |
| format | Thesis |
| id | my.utem.eprints-29009 |
| institution | Universiti Teknikal Malaysia Melaka |
| language | en |
| publishDate | 2025 |
| record_format | eprints |
| spelling | my.utem.eprints-290092026-02-27T00:16:04Z http://eprints.utem.edu.my/id/eprint/29009/ Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems Manap, Zahariah T Technology TK Electrical engineering. Electronics Nuclear engineering Indoor positioning technology is becoming increasingly influential in indoor applications, akin to how Global Navigation Satellite Systems revolutionized outdoor navigation. One of the primary challenges in indoor settings is multipath propagation, where wireless signals encounter reflections, diffractions, and scattering, yielding less accurate position estimates. Rather than combating multipath errors, this thesis leverages them as they encapsulate how radio waves interact with the environment that stores the position-related information of the base station (BS) and the mobile station (MS). This thesis proposes a hybrid indoor positioning method for single base station systems by jointly utilizing the fingerprinting method and geometric multilateration. The proposed method leverages room geometry and takes advantage of the multipath signal propagation to construct multiple virtual base station system model. This is done based on the concept of mirror image to determine potential virtual base stations (VBSs) with respect to the reflection surfaces where the multipath rays bounce off before arriving at the MS. The methodological framework of the proposed method comprises two main phases which are fingerprinting phase and position estimation phase. In the fingerprinting phase, classifiers are trained in two stages, each to predict the MS regions and reflection surfaces, respectively. The key attributes that establish the classification learning sessions are the channel parameters extracted from the ray tracing generated multipath signals. The channel parameters such as received power, time of arrival, and angle of arrival are used as fingerprint features that act as predictors in both learning sessions. The performance of the selected trained classifiers is evaluated based on the accuracy, precision, sensitivity and the F1-score. The results show that Coarse K-Nearest Neighbours is the optimal classifier that predicts MS regions at an exceptionally high accuracy, while Support Vector Machine Kernel is the optimal classifier that perfectly predicts the reflection surfaces. In the position estimation phase, a novel Geometric Random Sample Consensus (Geometric-RANSAC) multilateration method is proposed by optimizing the MS position estimate over several potential position estimates calculated using regional 3D geometric equations. When compared to a simple fingerprinting method, the median error for the hybrid method is observed to be 4.06 cm, which is substantially lower than the 538 cm median error of the fingerprinting method. Furthermore, 95% of the MS’s positions are estimated with less than 7.86 cm of error, in contrast to the 773 cm 95th percentile error exhibited by the fingerprinting method. Notably, the proposed Geometric-RANSAC exceptionally outperforms various least squares methods by achieving a median distance error and the 95th percentile at 4.06 cm and 7.86 cm, respectively. Despite the reduced robustness of the Geometric-RANSAC algorithm when applied to scenarios with a limited number of VBSs, its exceptional accuracy compensates for this limitation. This study makes a significant contribution by introducing a hybrid method that leverages multipath signals to achieve centimeter-level accuracy in indoor positioning systems. 2025 Thesis NonPeerReviewed text en cc_by_4 http://eprints.utem.edu.my/id/eprint/29009/2/Hybrid%20indoor%20positioning%20utilizing%20multipath-%20assisted%20fingerprint%20and%20geometric%20estimation%20for%20single%20base%20station%20systems.pdf Manap, Zahariah (2025) Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems. Doctoral thesis, Universiti Teknikal Malaysia Melaka. |
| spellingShingle | T Technology TK Electrical engineering. Electronics Nuclear engineering Manap, Zahariah Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title | Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title_full | Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title_fullStr | Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title_full_unstemmed | Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title_short | Hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| title_sort | hybrid indoor positioning utilizing multipath- assisted fingerprint and geometric estimation for single base station systems |
| topic | T Technology TK Electrical engineering. Electronics Nuclear engineering |
| url | http://eprints.utem.edu.my/id/eprint/29009/2/Hybrid%20indoor%20positioning%20utilizing%20multipath-%20assisted%20fingerprint%20and%20geometric%20estimation%20for%20single%20base%20station%20systems.pdf http://eprints.utem.edu.my/id/eprint/29009/ |
| url_provider | http://eprints.utem.edu.my/ |