Accurate Positioning in Urban Canyons with Multi-frequency Satellite Navigation
by Simon Ollander
Date of Examination:2020-12-07
Date of issue:2021-02-19
Advisor:Prof. Dr. Marcus Baum
Referee:Prof. Dr. Marcus Baum
Referee:Prof. Dr. Steffen, Schön
Files in this item
Name:Thesis_Ollander_Accurate_Positioning_in_Urba...pdf
Size:19.9Mb
Format:PDF
Description:Doctoral thesis
Abstract
English
Global Navigation Satellite System (GNSS) are crucial for fast and precise positioning in outdoor areas worldwide. Their applications include personal navigation in vehicles or with smartphones, autonomous driving and Internet of Things (IoT). However, in urban areas, tall buildings block and reflect the satellite signals, causing large and unpredictable positioning errors. To improve positioning in urban areas, there is a need for efficient signal processing methods, ideally ones that can operate with low-cost hardware, so that the solutions can be applied on a large scale. Error Characterization: To get an understanding of the measurement errors in urban areas, they have been characterized for two carrier frequencies using a ray tracing simulation. The results show that the pseudorange errors follow a highly non-Gaussian distribution, which highlights the challenge of navigation in urban canyons. Furthermore, the pseudorange errors on two frequencies are highly correlated in case of the Non-Line-of-Sight (NLOS) reception mode, which is not the case for Multipath (MP), where interference causes irregularity in the error. In general, the largest errors are caused by NLOS reception. Multipath Detection: To detect satellites with faulty measurements so that an optimal subset of satellites can be selected, two methods have been developed, both based on dual-frequency correlator output. The first one is closely related to Signal-to-Noise Ratio (SNR)-based methods by averaging 10 correlator points, while the second method scales better with more correlator points by computing the Pearson linear correlation. Both methods show good performance in classification of GNSS signals compared with the state of the art, and a preliminary validation from a test drive with real satellites shows promising results. Positioning can thus be improved simply by selecting the most error-free measurements. Multipath Mitigation: In order to correct measurements from the specific case of MP reception, a triple-frequency signal power-based method has been developed. For this, measurement noise on the signal power of a dual-frequency receiver has been characterized, and a corresponding Maximum Likelihood (ML) estimation problem has been derived. The method shows promising results in MP delay estimation compared with state of the art methods, given that the hardware can deliver precise signal power measurements. Using this method, navigation in areas where less than four satellites are available can be improved. Positioning: To produce a complete positioning solution, the previously explained dual-frequency satellite selection procedure has been integrated into an urban area simulation, where multiple vehicles are communicating in a collaborative positioning scenario. By combining dual-frequency satellite selection with collaborative positioning, the faulty measurements can be excluded, while more measurements become available through Vehicle to Vehicle Communication (V2V). According to the simulation, this combination can reduce the positioning error in a moderate urban area to 2.5 m.
Keywords: GNSS; Satellite Navigation; GPS; Navigation; Positioning; Multi-frequency; Multipath; Urban Canyon; NLOS; Reflection; Dual-frequency; Triple-frequency; Detection; Mitigation; Simulation; Raytracing; Accurate Positioning