The global navigation satellite systems (GNSS) have been widely used over the last decades in precise navigation, hydrographic survey, and maritime applications. There are different GNSS positioning techniques that can be used in maritime navigation, including differential, real-time kinematic (RTK), and precise point positioning (PPP) techniques. The maritime accuracy requirements for navigation and positioning are specified by the International Maritime Organization (IMO) standards (i.e., IMO resolution A.915 (IMO, 2002). Maritime navigation consists of a number of phases, including ocean, coastal, port approach and restricted waters, port, inland waterways, track control, and automatic docking. For maritime navigation, the positioning error should be estimated at 95% confidence level. The positioning accuracy requirements of each maritime navigation phase are given in Table 1.
IMO accuracy requirements for GNSS positioning (IMO, 2002)
Ocean | 10 | 25 | 10 | 10−5 |
Coastal | 10 | 25 | 10 | 10−5 |
Port approach | 10 | 25 | 10 | 10−5 |
Inland waterways | 10 | 25 | 10 | 10−5 |
Track control | 10 | 25 | 10 | 10−5 |
In port navigation | 1 | 2.5 | 10 | 10−5 |
Automatic docking | 0.1 | 0.25 | 10 | 10−5 |
The performance of the kinematic GNSS PPP solution in hydrographic and maritime applications has been investigated by a number of researchers (e.g., Alkan and Ocalan, 2013; El-Diasty and Elsobeiey, 2015; Berber and Wright, 2016; El-Diasty, 2016; Akpınar and Aykut, 2017; Alkan et al., 2017; Tegedor et al., 2017; Farah 2018; Specht et al., 2019; Yang et al., 2019; Erol 2020; Erol et al., 2020; Alkan et al., 2020; Nie et al., 2020; Tunalioglu et al., 2022). Yang et al. (2019) investigated the accuracy of the GPS/GLONASS/BeiDou RTK PPP solution for maritime applications. The accuracies of single constellation (i.e., GPS-only and BeiDou-only) and multi-constellation (i.e., GPS/BeiDou, GPS/GLONASS, GLONASS/BeiDou, and GPS/GLONASS/BeiDou) have been examined. The network RTK solution has been used as a reference. It was shown that the accuracy of the proposed real-time PPP solutions was within decimeter level. In addition, the accuracy of the GPS/BeiDou solution was superior to the GPS/GLONASS, GLONASS/BeiDou solutions. Erol (2020) examined the performance of multi-GNSS kinematic PPP solution using precise satellite orbit and clock products from different international GNSS service (IGS) analysis centers for marine applications. The carrier phase–based differential kinematic solution was used as a reference. It was found that the GPS/GLONASS/Galileo/BeiDou PPP solution using the German research center for geosciences (GFZ) products provided better accuracy than the other PPP solutions.
Presently, triple-frequency GPS/Galileo/BeiDou observations have become available, which in turn provide redundant measurements, various linear combinations, and fast ambiguity resolution (AR). Therefore, the PPP solution is enhanced particularly the convergence time. However, the satellite and receiver inter-frequency biases (IFB) are added as additional unknown parameters. IFB include phase and code biases. The phase IFB is the difference between the L1/L2 clock and the L1/L5 clock combinations, while the code IFB is the difference between differential code biases (DCB) estimated from P1–P2 and P1–P5 combinations. The ionosphere-free triple-frequency PPP solution can be executed using different models including two dual-frequency combinations (i.e., IF12 and IF13), a triple-frequency combination (i.e., IF123), between-satellite single-difference (BSSD) and semi-decoupled models.
The accuracy of the triple-frequency GNSS PPP has been examined in a number of studies (e.g., Cao et al., 2019; Pan et al., 2019; Geng et al., 2020; Li et al., 2020; Abd Rabbou et al., 2021; Guo et al., 2021; Naciri and Bisnath, 2021). Pan et al. (2019) evaluated the accuracy of three static triple-frequency GPS PPP solutions, which are two dual-frequency ionosphere-free combinations (i.e., L1/L2 and L2/L5) PPP, triple-frequency ionosphere-free PPP and triple-frequency uncombined PPP solutions. It was shown that comparable positioning accuracy was obtained from the three solutions. The convergence time of the triple-frequency ionosphere-free solution was better than the two solutions. Abd Rabbou et al. (2021) developed new triple-frequency GPS/Galileo PPP processing models, including undifferenced, BSSD, and semi-decoupled models. In addition, the traditional ionosphere-free dual-frequency PPP model has been used. The models developed by them have been validated for static and kinematic applications. It has been found that the positioning accuracy and convergence time of the triple-frequency PPP solutions have been improved compared to the undifferenced dual-frequency PPP solution. Furthermore, the performance of the triple-frequency semi-decoupled PPP model was superior to the other developed models.
Currently, a reliable real-time PPP solution can be obtained by using the IGS real-time service (RTS) products. RTS products include satellite orbit and clock corrections, code and phase biases, as well as vertical total electron contents (VTECs), which are available from a number of analysis centers. The Centre National d’Etudes Spatiales (CNES) provides real-time products for GPS, GLONASS, Galileo, and BeiDou satellite systems into two forms: the post processed and real-time streams (i.e., CLK 92 and CLK 93). A number of researchers have investigated the performance of the CNES real-time products (e.g., Wang et al., 2018; Katsigianni et al., 2019; Kazmierski et al., 2020; Nie et al., 2020; Wang et al., 2020; Zhao et al., 2020; Chen et al., 2021; Elmezayen and El-Rabbany 2021).
The objective of our study is to investigate the accuracy of kinematic triple-frequency GPS/Galileo/BeiDou PPP solution for maritime applications. Real-time CNES satellite orbit and clock products are used. GPS/Galileo/BeiDou measurements from a moving vessel are collected and then processed using two PPP models, which are the dual-frequency ionosphere-free PPP and the triple-frequency ionosphere-free PPP. The collected observations are also processed using different satellite combinations, namely GPS-only, GPS/Galileo, GPS/BeiDou, and GPS/Galileo/BeiDou. GPS differential solution is used as a reference. The estimated positioning accuracies are validated with respect to the minimum IMO positioning accuracy and integrity requirements for maritime applications.
The basic GNSS observation equation for pseudorange and carrier phase observations can be expressed as follows (Hofmann-Wellenhof et al., 2008):
The mathematical expression of the dual-frequency ionosphere-free PPP for pseudorange and carrier phase observables can be written as follows (Hofmann-Wellenhof et al., 2008):
In the dual-frequency ionosphere-free PPP model, DCB is lumped into the satellite clock errors and the receiver clock errors. After using the precise satellite clock products, the PPP model using GPS, Galileo and BeiDou observations can written as follows:
The unknown parameters for the dual-frequency ionosphere-free PPP solution can take the following mathematical expression:
The triple-frequency ionosphere-free PPP model can be formed using two dual-frequency ionosphere-free linear combinations (i.e., IF12 and IF13) of pseudorange and carrier phase observations as follows (Guo et al., 2016):
It can be seen that both receiver and satellite clock parameters are estimated for L1/L2 ionosphere-free combination. In addition, DCB are not lumped into receiver and satellite clock parameters. This is because DCBs are different on each ionosphere-free combination. Therefore, an IFB parameter, known as inter-frequency clock bias (IFCB), is introduced on IF13 combination. For code observations, the IFB parameter is the difference between ionosphere-free receiver DCB12 and DCB13 (
As a result, the vector of the parameter estimates can be expressed mathematically as follows:
To assess the performance of the kinematic PPP solution for maritime applications, a kinematic trajectory was observed in Bosporus, Istanbul, Turkey. The vessel trajectory is shown in Figure 1. A multi-frequency multi-GNSS Leica GS15 receiver had been located onboard the vessel to collect the observations with 1-s interval. The duration of the kinematic data set was about 11,298 s (about 3.18 h). The base station PALA of the Istanbul water and sewerage administration (ISKI) continuously operating reference stations network has been used as a reference station. Static GPS observations at the reference station have been simultaneously collected and processed using the differential solution, which has been used as the reference solution.
Figure 2 illustrates the number of tracked satellites and their position dilution of precision (PDOP) values. It can be seen that the combined GPS/Galileo/BeiDou (GEC) system provides more tracked satellites and improves the PDOP values. In addition, the PDOP values of the combined GPS/Galileo (GE) system are better than those of the GPS/BeiDou (GC) and GPS (G) systems.
To process the collected observations, Net_Diff GNSS software (Zhang et al., 2020) has been used. Net_Diff software can process multi-frequency multi-GNSS observations using different solutions, including single point positioning (SPP), PPP, PPP AR, differential, RTK, and RTK-PPP. For the orbit file, the SP3, broadcast, and state space representation (SSR) corrections formats can be used as well as for the clock file. In addition, the standard data formats including RINEX, EOP, ATX, DCB, and BSX can be imported. For data processing, the software applies different tropospheric delay correction, mapping functions, cycle slip, data smoothing, data combination, and parameter estimation models. It also provides various processing outputs, including coordinate comparison, cycle slip, ionosphere, troposphere, observation minus correction, and positioning residuals. In addition, several plots (e.g., coordinate plots, satellite number, satellite sky view, and satellite visibility) are provided.
Our kinematic trajectory data set has been processed using dual-frequency ionosphere-free, and triple-frequency ionosphere-free kinematic PPP solutions. To validate the accuracy of the obtained kinematic PPP solutions, the GPS observations from both the onboard vessel and the reference station have been processed using the post processed kinematic differential solution available in the Net_Diff software. The real-time CNES products (CNES, 2023) have been used to account for the satellite orbit and clock errors for PPP solution, while the final IGS products (IGS, 2023) have been used for the differential solution. The processing parameters of each solution are summarized in Table 2.
Processing parameters for both PPP and differential solutions
GNSS system | G, E, C | G, E, C | G |
Observations | Code and phase | Code and phase | Code and phase |
Mathematical model | Undifferenced ionosphere-free | Undifferenced ionosphere-free | Differenced |
Frequency |
G: L1/L2 E: E1/E5a C: B1/B2 |
G: L1/L2/L5 E: E1/E5a /E5b C: B1/B2/B3 |
L1/L2 |
Sampling rate | 1 Hz | 1 Hz | 1 Hz |
Elevation mask | 10° | 10° | 10° |
Orbit and clock | CNES (RT) | CNES (RT) | IGS (final) |
Tropospheric model | Saastamoinen model | ||
Mapping function | Vienna mapping function 1 (VMF-1) | ||
Parameter estimation | Kalman filter |
To investigate the accuracy of the developed kinematic PPP models using CNES real-time satellite orbits and clock products, different GNSS combinations are used, including GPS (G), GPS/Galileo (GE), GPS/BeiDou (GC), and GPS/Galileo/BeiDou (GEC). Furthermore, the aforementioned GNSS combinations are processed using two models namely, undifferenced dual-frequency ionosphere-free PPP (i.e., G-DF, GE-DF, GC-DF and GEC-DF) and undifferenced triple-frequency ionosphere-free PPP (i.e., G-TF, GE-TF, GC-TF and GEC-TF). Then, the estimated coordinates from all PPP solutions are compared with those obtained through the differential solution.
Figures 3 and 4 show the positioning errors in longitude, latitude, and altitude components, respectively, for the dual-frequency ionosphere-free and the triple-frequency ionosphere-free kinematic PPP solutions. As can be seen in the figures, the kinematic PPP solutions show very less decimeter-level positioning accuracy. Improvement in accuracy of the kinematic PPP solutions in the three components is obtained by adding Galileo and BeiDou observations to the GPS-only solution. In addition, the latitude positioning accuracy is superior to the other positioning components. This can be attributed to the GNSS satellite configuration. Furthermore, the accuracy of the triple-frequency ionosphere-free PPP solution is better than the dual-frequency PPP in longitude, latitude, and height components. It is also shown that the triple-frequency ionosphere-free PPP model converges faster than the dual-frequency PPP model. This is because the triple-frequency ionosphere-free PPP model provides redundant observations and thus more linear combinations than the traditional dual-frequency solution.
The PPP positioning accuracy is also investigated through computation of the differences with respect to the differential solution in longitude, latitude, and altitude components. Figures 5 and 6 illustrate the distribution of the differences for the dual-frequency PPP and triple-frequency PPP solutions, respectively. It can be noticed that the differences are in ±10 cm level for both longitude and latitude components, while it is in ±20 cm level for the height component. Moreover, the distribution of the height component for the triple-frequency PPP solutions is better than that of the dual-frequency PPP solutions. Furthermore, distribution of latitude differences is better than that of the longitude component. This may be due to the satellite distribution geometry.
To evaluate the precision of the PPP solutions, the standard deviation (STD) of the differences in longitude, latitude, and altitude components is computed (Figure 7). It can be seen that the maximum STD value is less than 0.12 m in longitude, latitude, and altitude components. In addition, the precision of the triple-frequency ionosphere-free PPP solutions is enhanced compared to that of the dual-frequency ionosphere-free PPP solutions.
To further assess the precision of triple-frequency PPP solutions, STD of the two-dimensional (2D) and three-dimensional (3D) positioning differences are computed and it is presented in Figure 8. It is clear that the 2D STD values are less than 0.12 m, while the 3D STDs are less than 0.16 m for the proposed PPP scenarios. Moreover, adding the third frequency to the PPP solution obviously improves the 2D or 3D positioning precision compared to the dual-frequency solution. This can be attributed to observations and linear combination redundancy. For the GPS-only solution, as an example, 2D STD is enhanced from 0.12 to 0.09 m, while 3D STD is improved from 0.16 to 0.14 m for the G-TF and G-DF solutions, respectively. Additionally, 2D and 3D STD of the GE-TF solution is reduced from 0.09 to 0.06 m and from 0.13 to 0.10 m, respectively, compared to the GE-DF solution. Also, the GC-TF model improves the 2D and 3D positioning precision from 0.10 to 0.08 m and from 0.14 to 0.12 m, respectively, with respect to the GC-DF model. Improvement in positioning precision is obtained from the GEC-TF processing model as well in comparison to the GEC-DF processing model by about 0.03 and 0.04 m in the 2D and 3D directions, respectively.
Furthermore, the statistical parameters for the 2D position including the mean and root mean square error (RMSE) are computed (Table 3) to assess the accuracy of the developed PPP solutions. As can be seen in Table 3, the 2D positioning accuracy is improved by adding the third frequency to the traditional dual-frequency PPP solutions. For example, the horizontal accuracy of the G-TF PPP solution is improved by about 37% compared to the G-DF PPP solution. The 2D position of the GE-TF PPP solution is also increased by about 22% compared to the GE-DF PPP solution. Enhancement in the 2D position is also obtained through the GEC-DF model by about 20% with respect to the GEC-DF model. An expectation is the GC-TF processing model. This might be attributed to the influence of the BeiDou L1/L3 IFB and ISB.
Statistical analysis of 2D position for different PPP solutions
G-DF | 0.140 | 0.172 | G-TF | 0.107 | 0.109 |
GE-DF | 0.084 | 0.099 | GE-TF | 0.073 | 0.078 |
GC-DF | 0.097 | 0.115 | GC-TF | 0.123 | 0.129 |
GEC-DF | 0.064 | 0.085 | GEC-TF | 0.092 | 0.068 |
It is also shown that the 2D position accuracy of the triple-frequency PPP solutions is significantly enhanced compared to the widely used G-DF PPP processing model. For instance, the GE-TF PPP solution enhances the 2D position by about 55% in comparison to the G-DF PPP solution. Also, the horizontal positioning accuracy is improved by about 25% and 60% for GC-TF PPP and GEC-TIF PPP solutions, respectively, compared to the G-DF PPP solution.
To compare the accuracy of the proposed PPP solutions with the IMO requirements, RMSE of the horizontal positioning should be estimated at 95% confidence level as follows (El-Diasty and Elsobeiey, 2015):
Table 4 summarizes the estimated RMSE of the 2D position at 95% confidence interval and maximum 2D positioning errors obtained from the proposed PPP solutions. It is found that, compared to Table 1, the positioning accuracy of the dual- and triple-frequency PPP solutions achieved the IMO requirements for ocean, coastal, port approach, port, inland waterways, and track control navigation applications. However, none of the proposed PPP solutions achieved the automatic docking navigation application. For the IMO integrity requirements, the proposed PPP solutions fulfilled the IMO requirements for all navigation phases. An exception is the GPS and GPS/BeiDou PPP solutions, which did not fulfill the automatic docking application.
Maximum 2D position error and 2D RMSE at 95% confidence for the PPP solutions
|
|
||||
---|---|---|---|---|---|
G-DF | 0.346 | 0.420 | G-TF | 0.162 | 0.265 |
GE-DF | 0.215 | 0.242 | GE-TF | 0.149 | 0.190 |
GC-DF | 0.236 | 0.279 | GC-TF | 0.243 | 0.314 |
GEC-DF | 0.168 | 0.221 | GEC-TF | 0.221 | 0.175 |
The performance of the triple-frequency quad-constellation kinematic PPP solution using real-time CNES products in maritime applications has been evaluated. The obtained kinematic PPP accuracy has been validated with respect to the IMO positioning accuracy and integrity requirements. GPS/Galileo/BeiDou measurements have been collected from a moving vessel. Then, the observations have been processed using dual-frequency PPP and triple-frequency PPP solutions. Five satellite combinations have been considered namely, GPS-only, GPS/Galileo, GPS/BeiDou, and GPS/Galileo/BeiDou. The computed coordinates have been compared to those computed from the GPS-only differential solution. It has been found that the triple-frequency kinematic PPP accuracy has improved to be at a decimeter level compared to the dual-frequency PPP. In addition, the horizontal positioning accuracy has enhanced by about 37%, 55%, 25%, and 60% for the GPS-only, GPS/Galileo, GPS/BeiDou, and GPS/Galileo/BeiDou triple frequency PPP solutions, respectively, compared to the traditional GPS dual-frequency PPP solution. Furthermore, the IMO positioning accuracy requirement for ocean, coastal, port approach, port, inland waterways, and track control applications has been achieved by the proposed PPP solutions. The automatic docking application’s accuracy requirement, however, has not been fulfilled by any PPP solution. The IMO integrity requirement also has been fulfilled by the PPP solutions for all navigation applications, except the GPS and GPS/BeiDou PPP solutions, which have not achieved the integrity requirement for automatic docking application only.