The Design a TDCP-Smoothed GNSS/Odometer Integration Scheme with Vehicular-Motion Constraint and Robust Regression

Remote Sensing

Kai-Wei Chiang¹  Yu-Hua Li¹  Feng-Yu Chu²  Li-Ta Hsu³ 

[1] Department of Geomatics, National Cheng Kung University, Tainan 701, Taiwan;Department of Land Economics, National Chengchi University, Taipei 116, Taiwan;Interdisciplinary Division of Aeronautical and Aviation Engineering, The Hong Kong Polytechnic University, Kowloon, Hong Kong, China;

关键词: global navigation satellite system;    intelligent transport system;    multipath;    non-line-of-sight;    time-differenced carrier phase;    odometer;   

DOI  :  10.3390/rs12162550

来源: DOAJ

【 摘 要 】

Global navigation satellite system (GNSS) is widely regarded as the primary positioning solution for intelligent transport system (ITS) applications. However, its performance could degrade, due to signal outages and faulty-signal contamination, including multipath and non-line-of-sight reception. Considering the limitation of the performance and computation loads in mass-produced automotive products, this research investigates the methods for enhancing GNSS-based solutions without significantly increasing the cost for vehicular navigation system. In this study, the measurement technique of the odometer in modern vehicle designs is selected to integrate the GNSS information, without using an inertial navigation system. Three techniques are implemented to improve positioning accuracy; (a) Time-differenced carrier phase (TDCP) based filter: A state-augmented extended Kalman filter is designed to incorporate TDCP measurements for maximizing the effectiveness of phase-smoothing; (b) odometer-aided constraints: The aiding measurement from odometer utilizing forward speed with the lateral constraint enhances the state estimation; the information based on vehicular motion, comprising the zero-velocity constraint, fault detection and exclusion, and dead reckoning, maintains the stability of the positioning solution; (c) robust regression: A weighted-least-square based robust regression as a measurement-quality assessment is applied to adjust the weightings of the measurements adaptively. Experimental results in a GNSS-challenging environment indicate that, based on the single-point-positioning mode with an automotive-grade receiver, the combination of the proposed methods presented a root-mean-square error of 2.51 m, 3.63 m, 1.63 m, and 1.95 m for the horizontal, vertical, forward, and lateral directions, with improvements of 35.1%, 49.6%, 45.3%, and 21.1%, respectively. The statistical analysis exhibits 97.3% state estimation result in the horizontal direction for the percentage of epochs that had errors of less than 5 m, presenting that after the intervention of proposed methods, the positioning performance can fulfill the requirements for road level applications.

【 授权许可】

Unknown