An Adaptive Stochastic Model Based on Scene Recognition and Satellite Classification with Low-Cost GNSS Devices

Yixin Wang, Chuang Qian, Hui Liu, Yeqing Bao, and Zhiwei Yang

Abstract: Low-cost Global Navigation Satellite System (GNSS) devices are with the advantages of inexpensive price, which have been widely used in public services. However, the performance of the hardware and antennas used in low-cost devices is limited, which result in that the satellite observations are susceptible to the surrounding environment, and the probability of multipath, Non-Line-Of-Sight (NLOS) signal observations occurring increases significantly. Therefore, the realization of high accuracy localization by low-cost receivers requires weakening or eliminating the effects of these anomalous observations, which is highly dependent on accurate stochastic models of observations in complex environments. In this paper, we propose an adaptive stochastic model refinement algorithm based on the GNSS environment scenes and satellite quality information with low energy consumption. We propose a scene recognition algorithm to classify typical urban scenes based on various features of GNSS raw observations. Based on the scene recognition results, we first obtain the possible signal occlusion area based on the assumption that the heading is the same as the direction of the road, then the received satellite signals are categorized into the LOS signals, the low-probability NLOS/multipath signals, and the high-probability NLOS/multipath signals according to the occlusion area, C/N0 and elevation angle. For LOS signals, the traditional stochastic model of C/N0 and elevation angle remains unchanged, and for the latter two signals, the adaptive coefficients based on azimuthal calculation and C/N0 are introduced to refine the concordance between the stochastic model and the actual observation noise. To evaluate the proposed method, experiments on two-wheeled motorbike, which is equipped with Allystar HD8040 low-cost devices are conducted under typical urban environments (open sky, single-sided building and urban canyon scenes). Compared to the C/N0 based stochastic modeling method, our proposed method could improve the horizontal positioning accuracy from 6-20m to 1-2m under complex urban environments.
Published in: Proceedings of the ION 2024 Pacific PNT Meeting
April 15 - 18, 2024
Hilton Waikiki Beach
Honolulu, Hawaii
Pages: 329 - 343
Cite this article: Wang, Yixin, Qian, Chuang, Liu, Hui, Bao, Yeqing, Yang, Zhiwei, "An Adaptive Stochastic Model Based on Scene Recognition and Satellite Classification with Low-Cost GNSS Devices," Proceedings of the ION 2024 Pacific PNT Meeting, Honolulu, Hawaii, April 2024, pp. 329-343.
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In