Previous Abstract Return to Session P6

Session P6: Present and Emerging Applications and Techniques for Time and Frequency using GNSS/RNSS/LEO and Optics

Machine Learning-based Characterization of GPS Satellite Oscillator Anomaly
Yunxiang Liu, and Y. Jade Morton, University of Colorado Boulder
Location: Beacon A
Date/Time: Thursday, Jan. 27, 2:12 p.m.

Objective
The objective of this paper is to present characteristics of the GPS satellite oscillator anomalies. The anomalies are detected by applying a machine learning-based algorithm on data collected from the SeNSe Lab global GNSS monitoring network from 2013 to 2018.
Introduction
GNSS signal-in-space (SIS) quality is fundamental to the operation and accuracy of the PNT solutions provided by GNSS. The stability of the on-board oscillators is of paramount importance as they provide the signal time-of-transmission for satellite–receiver range measurements. Satellite oscillator anomalies are manifested as SIS carrier phase disturbances, which may lead to degradation of precision, service discontinuity, loss of correction service coverage, or even outage, ultimately impacting the accuracy and integrity of GNSS applications [1]–[4]. Therefore, it is important to monitor and provide a timely warning of anomaly occurrence.
Satellite oscillator anomaly has been investigated in several past studies (e.g., [4]–[7]). The authors in [5] observed pulses of rapid phase variations at GPS L1 caused by satellite oscillator anomaly. In [6], a modern GPS block IIF satellite (PRN 1) showed sequences of large oscillator anomalies (maximum phase deviation larger than 1 cycle) on October 26th, 2012. In [4], frequent micro satellite oscillator anomalies (maximum phase deviation around 0.15 cycles) were observed from multiple GPS satellites and more anomalies were observed on block IIF satellites than on block IIRM satellites. In addition, large number of micro satellite oscillator anomaly events from block IIF satellites were observed in [7].
Methodology
In this paper, we apply a machine learning-based satellite oscillator anomaly detection method named Random Forest [7] to automatically detect the GPS satellite oscillator anomaly on the data collected by the SeNSe lab global GNSS monitoring network from 2013 to 2018. Comprehensive characterization of the detected GPS satellite oscillator anomaly is conducted. Patterns of the occurrence of satellite oscillator anomaly for each GPS satellite will be investigated. The statistics of the frequency, magnitude and duration, time between events of the satellite oscillator anomaly are discussed.
Actual or Anticipated Results
In this work, we apply the Random Forest-based detection method to a database collected by SeNSe Lab global GNSS monitoring network in 2013-2018, where Septentrio PolaRx5S receivers are deployed to obtain 100 Hz phase measurements [8]. Station locations include Poker Flat and Gakona in Alaska, Greenland, South Korea, Hong Kong, Singapore, Jicamarca in Peru, and Tololo and La Serena in Chile. All GPS block IIRM and block IIF satellites are processed by using the Random Forest with dual-frequency signals.
In total, 78751 satellite oscillator anomaly events are detected by the Random Forest detector from these stations. On average, 30.8 events are detected by a station on each day. In this work, an elevation mask of 0^° is applied. This low elevation mask is supported by the accurate detection performance: 205 detected events with elevation angles between 0^° and 5^° were randomly selected and manually inspected; among them, 200 events are genuine oscillator anomalies, corresponding to a precision of 97.5%. This result demonstrates the robustness and accuracy of the detection method on measurements with low elevation angles.
The detected events can be further verified by simultaneous observations from multiple stations. It turned out that there are 11228 satellite oscillator anomalies observed by two stations, corresponding to 22456 detected events. In addition, 261 anomalies and 13 anomalies are observed by 3 and 4 stations, respectively. In total, 23291 out of 78751 (~30%) detected events are verified by multiple stations. Surprisingly, no verified anomalies are found from block IIRM satellites. The remaining 55460 unverified events may be either non-satellite oscillator anomalies or genuine satellite oscillator anomalies that are not observed by other stations due to lack of coverage. Further investigation shows that 45863 out of 55460 unverified events are not observable by another station due to lack of data.
In this paper, comprehensive characterization of the satellite oscillator anomaly will be conducted by using the 23291 verified events, including investigation of distribution of anomaly duration, distribution of time between events, etc.
Significance of the work/Describe what is new and/or innovative about your presentation.
The application of satellite oscillator anomaly detection method in [7] is capable of detecting a large number of micro oscillator anomaly events that were never observed in the past. To the best of our knowledge, this will be the first work to show a characterization of GPS satellite oscillator anomaly events.
The characterization of the anomaly occurrence will help to improve monitoring systems’ performances [9] and enable the investigation of the cause of these events, where there is a lack of understanding due to the availability of anomaly data [10].
The characterization also provides guidance on future navigation satellite system and satellite oscillator design in order to avoid these glitches, which is of great interests to the GPS satellite manufacturer (e.g. GPS III: Lockheed Martin [11]).
Reference
[1] N. Vary, “DR#110: PRN4 Carrier Phase Anomalies Cause WAAS SV Alerts.” WAAS Technical Memorandum, Federal Aviation Administration WJHTC, Atlantic City International Airport, NJ, Oct. 17, 2012. [Online]. Available: http://www.nstb.tc.faa.gov/DisplayDiscrepancyReport.htm
[2] Y. Liu and Y. Morton, “Automatic Detection of Ionospheric Scintillation like GNSS Oscillator Anomaly Using a Machine Learning Algorithm,” in Proceedings of the 32nd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2019), Miami, Florida, Sep. 2019, pp. 3390–3400.
[3] Y. Liu and J. Morton, “Machine learning based Automatic Detection of Ionospheric Scintillation-like GNSS Oscillator Anomaly Using Dual Frequency Signals,” in Proceedings of the 51st Annual Precise Time and Time Interval Systems and Applications Meeting, San Diego, California, Jan. 2020, pp. 366–373. doi: 10.33012/2020.17311.
[4] Y. Liu and J. Morton, “Automatic Detection of Ionospheric Scintillation-like GNSS Satellite Oscillator Anomaly Using a Machine Learning Algorithm,” Journal of Institute of Navigation, vol. 67, no. 3, pp. 651–662, 2020, doi: https://doi.org/10.1002/navi.385.
[5] C. J. Benton and C. N. Mitchell, “GPS Satellite Oscillator Faults Mimicking Ionospheric Phase Scintillation,” GPS Solutions, vol. 16, no. 4, pp. 477–482, Oct. 2012, doi: 10.1007/s10291-011-0247-3.
[6] C. J. Benton and C. N. Mitchell, “Further observations of GPS satellite oscillator anomalies mimicking ionospheric phase scintillation,” GPS Solutions, vol. 18, no. 3, pp. 387–391, Jul. 2014, doi: 10.1007/s10291-013-0338-4.
[7] Y. Liu and Y. J. Morton, “Improved Automatic Detection of GPS Satellite Oscillator Anomaly using A Machine Learning Algorithm,” in Proceedings of the 33rd International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2020), Sep. 2020, pp. 3567–3580.
[8] Y. Jiao, “Low-Latitude Ionospheric Scintillation Signal Simulation, Characterization, and Detection on GPS Signals,” Doctoral Dissertation, Colorado State University, 2017.
[9] L. Heng, “Safe Satellite Navigation With Multiple Constellations: Global Monitoring Of GPS And GLONASS Signal-In-Space Anomalies,” Doctoral Dissertation, Stanford University.
[10] H. S. Cobb et al., “Observed GPS Signal Continuity Interruptions,” in Proceedings of the 8th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1995), Palm Springs, CA, 1995, vol. 8, pp. 793–795.
[11] “Lockheed Martin - GPS III Satellites.” https://www.lockheedmartin.com/en-us/products/gps.html?gclid=CjwKCAiA25v_BRBNEiwAZb4-ZdaLrnDBXuuvqP-PNQYJvh9xfT4Ge3deCyjz-xrLcDXjuwg3LTl3jBoCzzEQAvD_BwE



Previous Abstract Return to Session P6