Hammes, Ulrich Richard
Robust Positioning Algorithms for Wireless Networks.
Technische Universität, Darmstadt
[Ph.D. Thesis], (2010)
Available under Creative Commons Attribution Non-commercial No Derivatives, 2.5.
Download (1MB) | Preview
|Item Type:||Ph.D. Thesis|
|Title:||Robust Positioning Algorithms for Wireless Networks|
In this thesis, we consider the problem of finding the geographic position of a transmitter device (e.g. mobile phone), denoted as user equipment (UE), based on signal parameter estimates such as angle-of-arrival or time-of-arrival that are provided by surrounded sensors or base stations. If line-of-sight (LOS) channels between the UE and the base stations exists, high positioning accuracy can be obtained using trilateration or triangulation techniques. However, this assumption is ideal and not often encountered in practice. Especially in urban areas and hilly terrain, reflections at obstacles such as buildings and trees occur which force the signals of the UE to arrive at the base station via an indirect path. This phenomenon, called non-line-of-sight (NLOS) propagation, leads to erroneous signal parameter estimates that can strongly differ from the true ones and are thus modeled as outliers here. These NLOS errors result in large positioning errors when using standard techniques such as least-squares estimation and extended Kalman filtering. Thus, positioning algorithms that are robust against deviations from the LOS assumption are required. Since the statistics of the errors due to NLOS propagation are unknown in general we develop estimators that determine the NLOS error statistics from the observations non-parametrically. This estimate is then used in a parametric model to obtain the position estimate of the UE based on the maximum likelihood principle. The approach is termed semi-parametric since non-parametric pdf estimation is used for position estimation within a parametric signal model. A significant improvement in positioning accuracy with respect to conventional techniques is achieved in NLOS environments. For LOS environment, where Gaussian sensor noise is predominant, the proposed approach performs similar to a least-squares estimator. This approach is further extended to the case when the UE is moving over time. For this purpose, the framework of an extended Kalman filter (EKF) is used where the EKF equations are rewritten into a linear regression model at each time step and the semi-parametric estimator is used to solve for the state vector, containing position and velocity of the UE. Furthermore, a multiple model tracking algorithm is proposed that combines the advantages of robust EKFs and the standard EKF to achieve high accuracy in both LOS and NLOS environments. Finally, a different approach for positioning of a moving UE in NLOS environments is developed. It is based on a joint outlier detection and tracking algorithm where the errors due to NLOS effects are detected and discarded and the remaining measurements are used for updating the position estimate. Since we do not know which of them yields highest precision the remaining measurements are weighted with different probabilities to obtain the state estimate at each time step. The developed tracking algorithms outperform various robust competing estimators found in the literature while no knowledge of the NLOS error statistics is required.
|Place of Publication:||Darmstadt|
|Classification DDC:||600 Technik, Medizin, angewandte Wissenschaften > 620 Ingenieurwissenschaften|
|Divisions:||18 Department of Electrical Engineering and Information Technology > Institute for Telecommunications > Signal Processing|
|Date Deposited:||10 Feb 2010 14:47|
|Last Modified:||07 Dec 2012 11:56|
|Referees:||Zoubir, Prof. Dr.- Abdelhak M. and Gustafsson, Prof. Dr. Fredrik|
|Refereed:||10 December 2009|