Jin, Di (2021)
Wireless Network Localization via Cooperation.
Technische Universität Darmstadt
doi: 10.26083/tuprints-00019654
Dissertation, Erstveröffentlichung, Verlagsversion
Kurzbeschreibung (Abstract)
This dissertation details two classes of cooperative localization methods for wireless networks in mixed line-of-sight and non-line-of-sight (LOS/NLOS) environments. The classes of methods depend on the amount of prior knowledge available. The methods used for both classes are based on the assumptions in practical localization environments that neither NLOS identification nor experimental campaigns are affordable. Two major contributions are, first, in methods that provide satisfactory localization accuracy whilst relaxing the requirement on statistical knowledge about the measurement model. Second, in methods that provide significantly improved localization performance without the requirement of good initialization.
In the first half of the dissertation, cooperative localization using received signal strength (RSS) measurements in homogeneous mixed LOS/NLOS environments is considered for the case where the key model parameter, the path loss exponent, is unknown. The approach taken is to model the positions and the path loss exponent as random variables and to utilize a Bayesian framework. The goal is to infer the marginal posterior distribution of each unknown parameter, from which a position estimate, as well as its uncertainty information, can be obtained. This is achieved by using message passing methods in which two sets of functions, referred to as messages and beliefs, are iteratively updated. By combining variable discretization and Monte-Carlo-based numerical approximation schemes, two sets of functions are obtained. Such a numerical strategy allows the message updating rule to be implemented approximately while keeping the computational complexity affordable. Additionally, for networks with low-end sensors that only provide quantized RSS measurements, message passing algorithms and their parametric variants of low complexity are derived.
The second part of the thesis considers the more general case where statistical knowledge of the LOS/NLOS measurement errors is completely unknown, and range measurements, which are believed to be more accurate but quite sensitive to NLOS propagation, are available. The bias associated with each range measurement is modeled as an unknown parameter, and it is shown that bias parameters possess a sparsity property in LOS-heavy scenarios. This sparsity is exploited by introducing a sparsity-promoting term in the conventional cost functions, giving rise to a generic sparsity-promoting regularized formulation. By bounding the cost function, an alternative generic bound-constrained regularized formulation is developed. To ensure global optimality, the cost functions in these two generic formulations are specified so that they can be conveniently solved as two semi-definite programs (SDPs). It is theoretically shown, for certain conditions, that these two SDPs are equivalent in the sense that they share the same optimal solution. A major challenge of these two SDPs lies in the selection of an appropriate regularization parameter. An efficient data-driven strategy is developed to determine the regularization parameter and this is based on the special structure of the bound-constrained regularized SDP. Finally, numerical results, based on both synthetic and experimental data, are detailed. It is shown that the devised SDP approach provides overall good localization performance.
Typ des Eintrags: | Dissertation | ||||
---|---|---|---|---|---|
Erschienen: | 2021 | ||||
Autor(en): | Jin, Di | ||||
Art des Eintrags: | Erstveröffentlichung | ||||
Titel: | Wireless Network Localization via Cooperation | ||||
Sprache: | Englisch | ||||
Referenten: | Zoubir, Prof. Dr. Abdelhak M. ; So, Prof. Dr. Hing Cheung ; Yin, Prof. Dr. Feng | ||||
Publikationsjahr: | 2021 | ||||
Ort: | Darmstadt | ||||
Kollation: | IX, 127 Seiten | ||||
Datum der mündlichen Prüfung: | 17 Dezember 2020 | ||||
DOI: | 10.26083/tuprints-00019654 | ||||
URL / URN: | https://tuprints.ulb.tu-darmstadt.de/19654 | ||||
Kurzbeschreibung (Abstract): | This dissertation details two classes of cooperative localization methods for wireless networks in mixed line-of-sight and non-line-of-sight (LOS/NLOS) environments. The classes of methods depend on the amount of prior knowledge available. The methods used for both classes are based on the assumptions in practical localization environments that neither NLOS identification nor experimental campaigns are affordable. Two major contributions are, first, in methods that provide satisfactory localization accuracy whilst relaxing the requirement on statistical knowledge about the measurement model. Second, in methods that provide significantly improved localization performance without the requirement of good initialization. In the first half of the dissertation, cooperative localization using received signal strength (RSS) measurements in homogeneous mixed LOS/NLOS environments is considered for the case where the key model parameter, the path loss exponent, is unknown. The approach taken is to model the positions and the path loss exponent as random variables and to utilize a Bayesian framework. The goal is to infer the marginal posterior distribution of each unknown parameter, from which a position estimate, as well as its uncertainty information, can be obtained. This is achieved by using message passing methods in which two sets of functions, referred to as messages and beliefs, are iteratively updated. By combining variable discretization and Monte-Carlo-based numerical approximation schemes, two sets of functions are obtained. Such a numerical strategy allows the message updating rule to be implemented approximately while keeping the computational complexity affordable. Additionally, for networks with low-end sensors that only provide quantized RSS measurements, message passing algorithms and their parametric variants of low complexity are derived. The second part of the thesis considers the more general case where statistical knowledge of the LOS/NLOS measurement errors is completely unknown, and range measurements, which are believed to be more accurate but quite sensitive to NLOS propagation, are available. The bias associated with each range measurement is modeled as an unknown parameter, and it is shown that bias parameters possess a sparsity property in LOS-heavy scenarios. This sparsity is exploited by introducing a sparsity-promoting term in the conventional cost functions, giving rise to a generic sparsity-promoting regularized formulation. By bounding the cost function, an alternative generic bound-constrained regularized formulation is developed. To ensure global optimality, the cost functions in these two generic formulations are specified so that they can be conveniently solved as two semi-definite programs (SDPs). It is theoretically shown, for certain conditions, that these two SDPs are equivalent in the sense that they share the same optimal solution. A major challenge of these two SDPs lies in the selection of an appropriate regularization parameter. An efficient data-driven strategy is developed to determine the regularization parameter and this is based on the special structure of the bound-constrained regularized SDP. Finally, numerical results, based on both synthetic and experimental data, are detailed. It is shown that the devised SDP approach provides overall good localization performance. |
||||
Alternatives oder übersetztes Abstract: |
|
||||
Status: | Verlagsversion | ||||
URN: | urn:nbn:de:tuda-tuprints-196540 | ||||
Sachgruppe der Dewey Dezimalklassifikatin (DDC): | 600 Technik, Medizin, angewandte Wissenschaften > 620 Ingenieurwissenschaften und Maschinenbau | ||||
Fachbereich(e)/-gebiet(e): | 18 Fachbereich Elektrotechnik und Informationstechnik 18 Fachbereich Elektrotechnik und Informationstechnik > Institut für Nachrichtentechnik 18 Fachbereich Elektrotechnik und Informationstechnik > Institut für Nachrichtentechnik > Signalverarbeitung |
||||
Hinterlegungsdatum: | 01 Okt 2021 11:17 | ||||
Letzte Änderung: | 04 Okt 2021 07:17 | ||||
PPN: | |||||
Referenten: | Zoubir, Prof. Dr. Abdelhak M. ; So, Prof. Dr. Hing Cheung ; Yin, Prof. Dr. Feng | ||||
Datum der mündlichen Prüfung / Verteidigung / mdl. Prüfung: | 17 Dezember 2020 | ||||
Export: | |||||
Suche nach Titel in: | TUfind oder in Google |
Frage zum Eintrag |
Optionen (nur für Redakteure)
Redaktionelle Details anzeigen |