Previous Abstract Return to Session B5 Next Abstract

Session B5: Atmospheric Effects

Application of Machine Learning to Characterization of Multi-GNSS Ionospheric Scintillation
Yunxiang (Leo) Liu and Y. Jade Morton, Department of Aerospace Engineering Sciences, University of Colorado - Boulder; Yu (Joy) Jiao, Intel Corporation
Location: Cypress

Ionospheric scintillation, which is caused by the irregular structures in the ionospheric plasma, refers to the rapid fluctuation of amplitude and phase of radio-frequency signals, such as GNSS, propagating through Ionosphere [1]. Strong scintillation can severely impact signal acquisition and tracking in a GNSS receiver, resulting in performance degradation in accuracy and continuity. Therefore, a thorough understanding of Ionospheric scintillation on GNSS signals has drawn much attention in both the scientific fields and industry. Although previous works have explored the characteristics of scintillation on GNSS [2,3,4], only a limited amount of data is analyzed due to the time consuming nature of manually looking for scintillation events. Recently, machine learning techniques have been developed to automatically detect scintillations in GNSS measurements [5,6]. In this paper, we apply an improved implementation of machine learning algorithms to a large database of scintillation data collected in equatorial and high latitude areas. Statistical characteristics of both amplitude and phase scintillation on multi-frequency GPS and GLONASS signals are also presented.
The machine learning algorithm used in this study is a support vector machine [5]. In this improved version, the S4 index and sigma phi and the power spectrum density of the raw signal intensity and accumulated Doppler range (ADR) are used as input features to the support vector machine. Training data containing 140 hours data, which are manually labelled by visual inspection, are employed to build a robust prediction model for the support vector machine. This well-trained model is then applied to detect the scintillation events in the novel data. In this study, GNSS data collected at Alaska, Ascension Island, Chile, Greenland, Singapore, Hong Kong, and Peru are used for the detection of scintillation. Once the scintillation data are detected, we statistically analyze the properties of scintillation events to characterize their dependence on location, time, space, frequency bands, modulation, and solar and geomagnetic activities.
The machine learning algorithm has been implemented, validated, and tested. Its correct rates of detection are 95.3% and 91.4% for amplitude and phase scintillation, respectively. GNSS data collected from multiple sites in high latitudes and equatorial areas are being processed by the algorithm. The statistics of detected scintillation events will be derived based on the detected events and will be presented in the paper.
In this paper, more than 20TB data are automatically processed by a machine learning algorithm based on support vector machines. Statistical properties of the detected scintillation events are derived from the data to reveal important characteristics of scintillation activities in different regions of the world at different stage of the solar cycle. There results provide a comprehensive and solid view of scintillation events on GNSS signals. The unveiled characteristics of scintillation events on MCMF signals can be used to help GNSS receiver tracking algorithms design and to further understand space weather phenomena.
Reference:
[1] Yeh, Kung Chie, and Chao-Han Liu. "Radio wave scintillations in the ionosphere." Proceedings of the IEEE 70.4 (1982): 324-360.
[2] Jiao, Y., Morton, Y. T., Taylor, S., & Pelgrum, W. (2013). Characterization of high?latitude ionospheric scintillation of GPS signals. Radio Science, 48(6), 698-708.
[3] Jiao, Y., Xu, D., Morton, Y., & Rino, C. (2016). Equatorial scintillation amplitude fading characteristics across the GPS frequency bands. Navigation, 63(3), 267-281.
[4] Jiao, Y., & Morton, Y. T. (2015). Comparison of the effect of high?latitude and equatorial ionospheric scintillation on GPS signals during the maximum of solar cycle 24. Radio Science, 50(9), 886-903.
[5] Jiao, Y., Hall, J. J., & Morton, Y. T. (2017). Automatic Equatorial GPS Amplitude Scintillation Detection Using a Machine Learning Algorithm. IEEE Transactions on Aerospace and Electronic Systems, 53(1), 405-418.
[6] Jiao, Yu, John J. Hall, and Yu T. Morton. "Performance Evaluation of an Automatic GPS Ionospheric Phase Scintillation Detector Using a Machine?Learning Algorithm." Navigation 64.3 (2017): 391-402.



Previous Abstract Return to Session B5 Next Abstract