Abstract
With corrections transmitted through the E6 signal, the Galileo High Accuracy Service (HAS) provides the information necessary to execute a stand-alone precise point positioning algorithm in real time. Once fully operational, the service aims to deliver an accuracy of 20 cm and 40 cm (at the 95% confidence level) in the horizontal and vertical channels, respectively.
While most of the current literature focuses on analyzing the performance of HAS in static and open-sky signal reception scenarios, this study presents the results of tests conducted in both static and dynamic conditions, including open-sky and urban canyon scenarios. The tests clearly demonstrate that utilizing HAS corrections leads to a significant reduction in positioning error across all tested environments. Furthermore, a specific analysis of HAS message availability in a harsh environment indicates that the corrections obtained from the signal in space are available approximately 95% of the time during dynamic scenario tests.
- decimeter-level accuracy
- Galileo High Accuracy Service
- HAS corrections
- position accuracy
- static and dynamic tests
1 INTRODUCTION
Position, velocity, and timing (PVT) information is now indispensable in various fields, ranging from automotive to space applications. Global navigation satellite system (GNSS) technology is a key enabler for such applications because of its global and continuous coverage, as well as its user-friendly nature. The number of applications and users relying on GNSS-based PVT is increasing, as confirmed by a recent European Union Agency for the Space Programme (EUSPA) market report (EUSPA, 2024). GNSS-based applications are diverse, with varying requirements in terms of accuracy, continuity, and integrity. To meet these evolving needs, GNSSs are expanding their services and developing new ones to enhance their performance. In this context, Galileo is introducing two new services: the Galileo High Accuracy Service (HAS) (European Union, 2022) and the Open Service Navigation Message Authentication (OSNMA) (European Union, 2023a). The former focuses on increasing position accuracy and is the primary focus of this paper, whereas the latter provides a mean for verifying the authenticity of received data using cryptographic elements.
The position accuracy requirements of emerging applications, such as autonomous driving, are becoming increasingly stringent, necessitating the use of specific techniques and/or services. While traditional single point positioning (SPP), which provides accuracy on the order of meters (Kaplan, 2017), has been used in the past, more advanced techniques such as precise point positioning (PPP) and real-time kinematics (RTK), originally employed for geodetic and surveying applications, are now being adopted in mass-market applications. These techniques allow for precision on the order of a decimeter, but require more complex algorithms, external infrastructure, and information.
To address these limitations, the new Galileo HAS provides the information necessary to execute a stand-alone PPP algorithm in real time. This new open-access service was declared operational in January 2023 (EUSPA, 2023) and aims to enable decimeter-level accuracy globally, using corrections broadcast by Galileo on the E6 signal (European Union, 2022) and via the internet (European Union, 2023b).
The new service has generated significant interest in the GNSS community, leading to several studies reported in relevant conferences and journals. The design of the service has been described by Fernández et al. (2018), who detailed the architecture and elements of the service. The Reed–Solomon scheme adopted for correcting transmission errors has been analyzed by Fernández et al. (2020), whereas the limitations due to the propagation channel were experimentally assessed by Borio et al. (2019). The impact of the receiver’s location on data demodulation performance has been discussed by Susi et al. (2021), who analyzed data collected on-board a vessel traveling in the Arctic region. The coverage and achievable accuracy of the service have been described by Fernández et al. (2022). The availability of corrections in the signal in space (SiS) triggered analyses on the usage of HAS for space applications. Hauschild et al. (2022) explored a potential adoption of the HAS for real-time orbit determination of the Sentinel-6A satellite, highlighting that HAS corrections offer greater overall orbit accuracy than those obtained using broadcast ephemerides alone. Hadas et al. (2024) evaluated the quality of orbit and clock corrections provided by HAS, analyzing the performance achievable in selected real-time GNSS applications. Regarding the parsing of HAS data, a library (HASlib) capable of decoding only the processing of Septentrio binary files was presented by Horst et al. (2022). Additionally, an extended decoder, the Galileo HAS Parser, capable of converting the bits broadcast through the E6B signal into actual corrections, was presented by Borio et al. (2023).
In the positioning domain, the impact of Galileo HAS corrections has been analyzed via various positioning engines. Chamorro Moreno et al. (2023) demonstrated the ability to achieve sub-decimeter accuracy using Galileo HAS corrections and a proprietary algorithm. Parra et al. (2023) used a common PPP algorithm, demonstrating a 95% accuracy of 19 cm horizontally and 34 cm vertically. Chen et al. (2023) combined Galileo HAS corrections with the Hemisphere Atlas GNSS global correction service to provide dual redundancy. In addition to professional receivers, an implementation of Galileo HAS using software-defined radio was presented by Alfonso (2023). The benefits of Galileo HAS corrections were also demonstrated by Angrisano et al. (2023) using low-cost devices for SPP. Furthermore, the results of using the RTKLIB software for processing HAS corrections for a static user were presented by Prol et al. (2024). A first demonstration of RTKLIB’s capability to process HAS corrections for automotive users was discussed by (Cucchi, Gioia, et al., 2023).
The majority of existing literature has primarily focused on analyzing the performance of Galileo HAS in static and open-sky signal reception scenarios. However, automative applications are among the intended applications for HAS; in such cases, the receiver is dynamic, and the signal conditions may vary significantly. In urban scenarios, single or multiple measurement outliers can be present, potentially degrading the navigation solution’s performance despite the corrections provided by HAS. Thus far, few studies have investigated the performance of HAS in such conditions. An example has been reported by (Cucchi, Gioia, et al., 2023), who presented the results of an automotive test and assessed the positioning performance achievable using HAS corrections in suburban conditions. Additionally, the use of HAS corrections by low-cost receivers is an unexplored field, despite the fact that automotive applications typically exploit such mass-market devices. To address these gaps, this paper analyzes the results of a series of tests conducted in both static and dynamic conditions, encompassing open-sky and urban canyon scenarios. This manuscript is an extension of the paper titled “Testing the Galileo High Accuracy Service in Different Operational Scenarios,” presented by the same authors at the Institute of Navigation GNSS+ 2023 conference (Cucchi, Damy, et al., 2023). In comparison to the conference paper, the following elements have been added in the current report:
The background and motivations of the work have been expanded, including a potential error budget for evaluating HAS benefits in different environments.
The literature review has been extended, incorporating additional and updated sources.
A detailed description of the setups used for the different tests has been added, including the challenge of establishing a reference solution.
Further details on the methodology are provided, including parameter settings and strategies used.
Additional preliminary data analyses are included to further characterize the environments of the test scenarios.
A more in-depth analysis is provided, including additional results.
The test setups were implemented by exploiting the Joint Research Centre (JRC) testing and demonstration hub for EU GNSS programs, described by Cucchi et al. (2021). Various grades of GNSS receivers were utilized for the tests, ranging from low-cost to high-end devices, including two HAS-enabled receivers, all of which offer multi-frequency and multi-constellation capabilities. For devices providing raw measurements, a post-processing analysis was conducted, considering different navigation solution strategies with and without HAS corrections. Specifically, the solution obtained after the HAS corrections had been applied to a common PPP algorithm was compared with the solution obtained using only broadcast parameters. The performance of the navigation solutions was assessed in terms of availability (defined as the percentage of time in which the solution can be computed) and position accuracy, evaluated for the 68th percentile of the horizontal and vertical channels, accounting for the fact that the service is currently in its initial phase.
The remainder of this paper is structured as follows. Section 2 describes the main characteristics and target performance of the HAS service. Section 3 outlines the methodology and presents the metrics used to assess the performance. The experimental setups used for the three different tests are described in Section 4. The results are discussed in Section 5, and finally, Section 6 presents conclusions and provides potential ideas for future work.
2 GALILEO HAS
Galileo HAS provides PPP corrections free of charge, with the goal of achieving a real-time user positioning error of less than two decimeters horizontally in nominal conditions, i.e., in open-sky scenarios with the HAS message status set to operational mode. Currently, HAS corrections cover orbits, clocks, and code biases for each Galileo and Global Positioning System (GPS) satellite, with the expectation of carrier-phase biases and atmospheric corrections being included in the HAS full service phase (Fernández et al., 2023). Although HAS corrections are accessible through the E6-B SiS, users without an E6-B-capable receiver can also retrieve the corrections through a terrestrial link via the internet. The corrections are provided in a format similar to compact-state space representation (CSSR) for different frequencies and frequency combinations of Galileo and GPS signals. The corrections are broadcast for Galileo E1/E5b/E5a/E6 and E5AltBOC, as well as for GPS L1/L5/L2 signals.
The HAS comprises two service levels. Service level 1 has global coverage and provides high-accuracy corrections for satellite orbits, clocks, and measurement biases (currently provided for the code, with carrier-phase bias to be included in the future). Service level 2 will have regional coverage and provide level-1 corrections plus atmospheric corrections (at least ionospheric) and corrections for potential additional biases. The main performance metrics for the service are defined in the HAS Service Definition Document (SDD) (European Union, 2023d), in terms of positioning accuracy and availability.
To illustrate the impact of HAS corrections on positioning accuracy, Table 1 summarizes the different GNSS ranging error sources, as reported by Teunissen et al. (2017), Sanz Subirana et al. (2013), European Union (2023c), and European Union (2023d). These values are intended to convey the relative magnitude of the different errors rather than their exact absolute values. To focus on the gain provided by HAS, a dual-frequency (DF) PPP solution using broadcast ephemeris is compared with a similar solution that also includes HAS corrections.
GNSS Error Sources for Comparison Between a DF PPP Solution Using Broadcast Ephemeris and a Similar Solution That Includes HAS Corrections
As a float PPP solution is considered in the analysis, the impact of the phase ambiguities is captured. According to Teunissen et al. (2017), an error equal to half the wavelength of the DF narrow-lane combination is considered, corresponding to approximately 5 cm. The first-order ionospheric error is corrected by the DF combination, leaving only higher-order components. The tropospheric error is estimated by the PPP algorithm, and the residual contribution is considered negligible. The effects of the receiver antenna, Earth deformation, and phase wind-up are also excluded from this analysis because of their relatively minor contributions to the error budget. However, the antenna phase center offset and variation could be corrected via the International GNSS Service (IGS) antenna exchange format (ANTEX) model, and the phase wind-up could be compensated for if the satellite/receiver antenna orientation were known. Finally, the receiver corrects the relativistic clock effects, while the differential code biases are not required as a dual-frequency solution is considered.
As shown in Table 1, when local errors remain small, clock and orbit errors are the primary contributors to positioning inaccuracies. However, in degraded reception conditions, local error contributions become dominant. Thus, it is expected that the gain from HAS corrections is more important in open-sky conditions, while this gain may be considerably limited in more degraded conditions. The potential benefits of applying HAS corrections in various scenarios for receivers of different grades are further evaluated in the remainder of the paper.
3 METHODOLOGY
The methodology adopted in this work is described in the two following subsections, which focus on data acquisition and processing and on the definition of the performance metrics used in the experiments.
a Measurement Acquisition and Processing
Four receivers were used in the tests: the HAS User Terminal (HAS-UT) receiver (Pintor at al., 2022), PAULA receiver (Chamorro Moreno et al., 2022), Septentrio PolaRx5 receiver (Septentrio, 2022), and u-blox F9P receiver (u-blox, 2023). The former two receivers are prototype devices capable of receiving, decoding, and applying HAS corrections. Thus, the performances of these proprietary solutions were assessed directly. In contrast, for the commercial off-the-shelf (COTS) products (Septentrio PolaRx5 and u-blox F9P), a PPP solution was computed according to the processing scheme depicted in the orange box of Figure 1.
(Top) Workflow of the processing scheme applied for COTS receivers; (bottom) settings used for PPP via RTKLIB software
For these receivers, the proprietary binary files are parsed to obtain the receiver independent exchange (RINEX) navigation and observation files. Then, PPP solutions are computed with the RTKLIB software (version 2.4.3 (Takasu et al., 2009)), with the following as main inputs:
GNSS observables from the receiver (including code, carrier, and Doppler-shift measurements) in RINEX format.
Broadcast ephemerides for GPS and Galileo retrieved from the IGS network in the RINEX navigation files.
HAS corrections extracted from the Galileo navigation messages collected with a static receiver and processed via the HASlib software (Horst et al., 2022). The static receiver was connected to an antenna in open-sky conditions on the Ispra site of the JRC to ensure visibility conditions similar to those experienced by the prototype receivers.
While broadcast ephemerides are not commonly considered as a viable option for PPP because of their limited accuracy, the increasing demand for real-time PPP applications, combined with the smaller SiS range error of Galileo and the more stable and modernized clock of GPS, has prompted studies to investigate and assess the performance achievable by such solutions (Carlin et al., 2021; Ge et al., 2022). Furthermore, reporting the performance obtained with broadcast ephemeris allows for a better understanding of the expected benefits offered by HAS corrections, as presented in Table 1. Finally, the SPP solution from one of the prototypes (HAS-UT) is also provided for reference. The benefits of HAS corrections in the SPP solution have been discussed by Angrisano et al. (2023) and are not considered in this work.
The PPP solution was computed using RTKLIB software, with the settings outlined in the lower part of Figure 1:
The option “PPP static” is selected for static user cases, whereas the “dynamic” option is used for automotive user cases.
The ambiguity resolution is set to “Float.”
The elevation masking angle is set to 10°.
An ionospheric-free combination is used.
The tropospheric delay is estimated.
The carrier-to-noise (C/N0) threshold is set to 20 dB-Hz, meaning that measurements with C/N0 values lower than 20 dB-Hz are not used in the navigation solution computation.
The “receiver autonomous integrity monitoring” option is enabled.
These settings were selected because they are representative of generic HAS-based PPP processing and do not include any specific optimization.
The different navigation solutions evaluated in this analysis and their respective nomenclatures are reported in Table 2. The type of data output by the device, the positioning method (SPP or PPP), and the type of ephemerides used are also indicated.
Navigation Solutions Considered
NMEA: National Marine Electronics Association
b Metrics
The performance metrics were selected to be aligned with the minimum performance levels defined in the Galileo HAS SDD (EUSPA, 2023). In particular, the horizontal and vertical errors are computed after the convergence time, i.e., after the initial transient phase during which errors may rapidly change. The values used to assess the convergence of the solution are based on the Galileo HAS full service commitment (EUSPA, 2023), but are defined at the 68th percentile instead of the 95th percentile to reflect the fact that the service is still in its initial phase.
In particular, the following specific metrics are evaluated for each solution:
Horizontal and vertical position accuracy, evaluated in terms of the 68th percentile of the error. The error is computed as the difference between the reference position and the estimated position from the receiver or from postprocessing. For PPP solutions, the accuracy is estimated after convergence is reached.
Convergence time, defined as the amount of time the solution needs to steadily converge to the target accuracy. In this paper, the convergence time is measured as the first epoch at which both the horizontal accuracy (20 cm) and vertical accuracy (40 cm) are reached.
Solution availability, defined as the percentage of time during which a PVT solution is computed by the receiver.
HAS message availability, defined as the percentage of time in which HAS corrections retrieved from commercial navigation message (C/NAV) pages for the satellite/signal are available to the user (not accounting for the validity of the corrections themselves).
Note that the HAS message availability is introduced to quantify the impact of reception conditions on the retrieval of corrections from the SiS. The HAS message availability should not be confused with the HAS (service) availability, defined as the percentage of time over a reference period in which both the horizontal and vertical positioning accuracies are achieved, which is expected to reach 99% for full service.
4 TEST SETUP
The Galileo HAS was experimentally evaluated in three different scenarios, with a dedicated setup designed and implemented for each scenario. The main characteristics of the developed setups, such as the devices used and the generation of reference positions, are detailed in the following sections.
a Static Open-Sky and Obstructed Scenarios
For the static tests, two different locations at the JRC Ispra campus were used: one for the open-sky analysis and another for the obstructed scenario. Data for the open-sky scenario were collected from a geodetic-grade antenna, specifically the Trimble Zephyr 2 (Trimble, 2017), installed on the roof of the European Microwave Signature Laboratory (EMSL). This antenna is used as a reference for all tests carried out at the JRC, and its location is regularly evaluated, with observations collected by a high-end receiver over extended periods of time. The data are then post-processed in both PPP and RTK modes, and the solutions are compared to ensure that millimeter-level accuracy is reached. For this test, approximately 9 h of data were collected at 1 Hz, with all test devices connected to the antenna through a splitter.
For the obstructed scenario, an antenna is placed a few meters away from a tall building (hosting the EMSL), resulting not only in the obstruction of a portion of the sky, but also in the presence of medium to strong multipath effects. The antenna used for this scenario is of geodetic grade and was developed in the frame of the FANTASTIC project (Egea Roca et al., 2018). Approximately 4 h of data were collected at 1 Hz, with all receivers connected through a splitter.
For this scenario, because the reception conditions made it unlikely that the required accuracy would be reached via a GNSS-only solution, the antenna position was surveyed using a total station. The total station measures the position of a target using optical-based angle and distance measurements, guaranteeing millimeter-level accuracy. The station was set up using two reference points in open-sky areas, where the coordinates were computed using GNSS data processed in post-processing kinematic (PPK) mode, using the already georeferenced roof antenna.
The left panel in Figure 2 indicates the location of the antenna for the open-sky (blue marker) and obstructed (red marker) scenarios. The right panel in Figure 2 shows the setup of the GNSS antenna for the static obstructed scenario.
(Left) GNSS antenna locations for open-sky (blue marker) and obstructed (red marker) static tests; (right) antenna used for the static obstructed test
b Dynamic Urban Scenario
For the dynamic tests, a JRC van was equipped with state-of-the-art devices. The same GNSS antenna used for the obstructed scenario was mounted on the roof of the van and connected via a splitter to different GNSS receivers, including the two enabled HAS prototypes. The measurements from u-blox F9P could not be used because of an unexpected error that occurred during data collection.
The dynamic tests were performed in the city center of Milan, where the presence of obstacles limits satellite visibility and vehicle speeds are reduced. The entire path traveled in the Milan metropolitan area is shown in the following section. The test lasted approximately 40 min, reflecting typical automotive conditions, with an average speed of 30 km/h and frequent start and stop events.
Given that HAS usage is expected to deliver a positioning accuracy at the decimeter level, the estimation of an accurate true path is of paramount importance. For this test, a Novatel PwrPak7-E2 receiver integrated with an inertial measurement unit (IMU) was used (Novatel, 2023), which, per specification, offers accuracy performance up to 1 cm in live RTK processing and open-sky conditions. A reference trajectory with centimeter-level accuracy can also be obtained in degraded environments with the use of post-processing tools. Thus, the collected data were post-processed in PPK mode with Inertial Explorer software (Novatel, 2020), together with the data collected from the base station at the JRC campus. The tool includes ad hoc optimization of the inertial measurements (e.g., level arm estimation, smoothing, forward and reverse processing), leading to centimeter-level accuracy during GNSS outages. This tool outputs the position accuracy estimation, with an estimated average of 3.3 cm (1 sigma) for the dynamic test case (excluding tunnels).
5 RESULTS
The results of the test campaigns are presented hereafter for the static open-sky and degraded scenarios, and results related to the dynamic scenario are discussed.
It is important to note that the goal of the analysis is not to directly compare the results of different receivers’ outputs in terms of accuracy and availability, but rather to evaluate the benefits brought by the application of HAS corrections in different environments. Indeed, a direct comparison among receivers would be unfair for several reasons. Firstly, two receivers (HAS-UT and PAULA) have proprietary implementations that include HAS corrections as well as other internal algorithms or fusion with inertial sensors in the case of the PAULA receiver. Secondly, the results from the COTS receivers are based on a solution evaluated with the RTKLIB software, which also applies specific internal post-processing logic. Finally, the PPP strategy implemented in the two prototypes could be different from the strategy used in RTKLIB. These considerations are further discussed in the conclusions.
a Static Tests
i Preliminary Data Analysis
This section presents a preliminary data analysis that showcases the representativeness of the selected setups for the open-sky and obstructed scenarios.
In Figure 3 (top), the number of visible Galileo and GPS satellites for the open-sky and obstructed tests is shown as a function of the data collection time (expressed as coordinated universal time [UTC]). The figure demonstrates that the number of satellites is stable in open-sky conditions, whereas rapid fluctuations are evident in obstructed environments because of the surrounding buildings and trees that mask GNSS satellite signals. In open-sky conditions, the number of visible satellites varies between 10 and 12 for GPS and between 7 and 10 for Galileo. In the obstructed scenarios, a larger variation can be noted, with a few epochs showing only 3 available satellites for GPS.
(Top) Number of visible GPS and Galileo satellites for the open-sky and obstructed scenarios; (bottom) EDOP and NDOP time evolution for the two scenarios
The effect of the obstruction due to the building on the satellite geometry is also evident, as shown in Figure 3 (bottom). Larger values of the east dilution of precision (EDOP) and north dilution of precision (NDOP) are observed for the obstructed case, as well as rapid variability. The choice of analyzing the EDOP and NDOP separately is based on the fact that the position DOP is a combination of the DOP elements along different components, which results in mixing of information. In contrast, the EDOP and NDOP allow for the identification of any weak directions, which is particularly relevant for the current analysis. In open-sky conditions, the NDOP is usually larger than the EDOP because of the constellation design for middle-latitude users, as shown in Figure 3. However, this condition is not expected in the obstructed case, where the presence of the building masking almost half of the sky results in larger EDOP values, reducing the east– west accuracy.
In addition to the number of available satellites and geometric conditions, a fundamental element that impacts the accuracy of the PVT solution is the presence of multipath. In Figure 4, the multipath linear combination (MLC) for E1 and L1 signals is shown as a function of time, with red markers representing values observed in open-sky conditions and blue markers representing values computed for the obstructed scenario. As expected, multipath is significantly larger in the obstructed scenario, with MLC values reaching tens of meters. This result is further confirmed by the standard deviation of the MLC, which is 0.79 m in obstructed conditions, compared with 0.26 m for the open-sky case. This analysis demonstrates that the selected setups for the two scenarios are indeed representative of the targeted reception conditions.
MLC values as a function of time for the obstructed (blue) and open-sky (red) scenarios
ii Results for the Static Open-Sky Test
This section presents the results of the HAS analysis for static open-sky conditions. The horizontal position solutions obtained using the two prototype devices are depicted in the left panel of Figure 5, along with dashed circles indicating the 68th percentile. For the HAS-UT, the SPP solution (blue) is reported along with the PPP solution (red), whereas the PAULA receiver provides only the PPP solution (yellow). It is evident from the figure that the PPP solutions provided by the two devices are more accurate than the SPP solution. This difference is attributed to the different positioning strategies, with SPP exploiting only code measurements while PPP uses both code and carrier-phase measurements. In addition, for both PPP solutions, a convergence toward the reference position can be observed, whereas the SPP solution shows more spreading. The dashed circles for the PPP solutions are very similar, with radii of 7 cm and 8 cm for the HAS-UT and PAULA receivers, respectively. For these two prototypes, it is not possible to compute the PPP solutions without HAS corrections.
Horizontal position solutions provided by the (left) HAS-UT and PAULA receivers and (right) PolaRx and u-blox receivers
The circles indicate the 68th percentile errors.
Therefore, the benefit of HAS corrections is evaluated using the two COTS devices, as shown in the right panel of Figure 5. The green and purple markers represent the solutions obtained using the PolaRx measurements with HAS correction or without HAS corrections (BRDC - broadcast), respectively. Likewise, the blue and red markers represent the solutions obtained by exploiting the u-blox measurements. For both devices, a clear reduction in error can be observed when the HAS corrections are applied, with the green and purple clouds being more concentrated. The circles represented by the 68th percentile of the horizontal error show that the HAS corrections allow a reduction from 18 cm to 9 cm for PolaRx, whereas the error remains at 15–19 cm for u-blox. The finer grid in the central part of the plot on the right allows for better appreciation of the improvements brought by HAS.
The typical convergence behavior of all solutions is further analyzed in Figure 6, which presents the horizontal error time series over a data collection duration of 9 h. An initial transient is observed for all solutions, followed by a convergence of the solutions. It can be observed that all solutions achieved a final horizontal accuracy below the 20-cm benchmark. In addition, it can be noted that after convergence, the errors remain fairly stable, at the centimeter level for the solutions using HAS and at the decimeter level for the solutions using broadcast products only.
Horizontal error as a function of time
Table 3 reports the horizontal, vertical, and three-dimensional (3D) positioning errors (68%) for all configurations. It is easy to observe the significant gain achieved by the PPP solutions (decimeter level) compared with the SPP (on the order of 1 m). The benefits of applying the HAS corrections are evident, with a reduction in the 3D error of 6 cm for the Septentrio device and 22 cm for u-blox. It can be noted that the gain observed for the u-blox solution primarily arises from the vertical component.
68th Percentiles of the Horizontal, Vertical, and 3D Positioning Error for the Static Open-Sky Scenario
The convergence time (as defined in the previous section) for the static open-sky case is presented in Figure 7. The convergence time varies from 28 min for the HAS-UT to approximately 5 h for u-blox without HAS correction. The application of HAS corrections leads to a significant reduction in convergence time for the COTS devices, with the convergence time decreasing by a factor of 3 for the Septentrio case and a factor of approximately 6 for the u-blox case.
Convergence time for the static open-sky scenario
It can be concluded that while the PPP solutions computed using broadcast-only products lead to a significant reduction in positioning errors, the convergence time needed to reach the defined thresholds is significantly longer than that of the HAS-based PPP solutions. Although some studies suggest that broadcast products can be used for PPP (Carlin et al., 2022; Hadas et al., 2019), the current results show that they may not be suitable for all applications, as the convergence time would be too long. For a receiver running a real-time stand-alone PPP implementation, these results illustrate the significant gain provided by applying the HAS corrections, as the performance of the COTS receivers and the receivers with a proprietary HAS solution all meet the service targets of 20 cm and 40 cm for the horizontal and vertical errors, respectively.
iii Results for the Static Obstructed Test
In this section, the results from the static obstructed test are presented. As previously discussed, the local effects in these conditions can significantly degrade the accuracy and availability of the navigation solution and can also prevent the PPP solution from converging. In the analysis hereafter, the horizontal error is divided between its north and east components to illustrate the effect of the building obstruction. The building in front of the antenna installation location is oriented north–south, leading to larger errors in the east component than in the north component, as discussed in the preliminary analysis of the DOP components. Because the east component is not able to converge for any of the receivers, Figure 8 presents only the evolution of the north error as a function of time.
Time evolution of the north error for the static obstructed scenario
From the figure, it can be observed that the error generally decreases over time for the PPP solutions; however, not all solutions are able to converge. Moreover, even if a solution reaches convergence, the error can increase again, as observed for the u-blox-HAS and u-blox-BRDC solutions. This behavior can be expected as local multipath effects create measurement errors that are likely to impact the PPP algorithm results. In this scenario, the configuration with the smallest error is the one using PolaRx measurements with the application of HAS corrections (purple). In this case, the benefit of applying the HAS corrections is evident, as the configuration using PolaRx observables without HAS correction (yellow) has a larger error. The presence of outliers also strongly impacts the solution of the HAS-UT, with re-convergences of the filter occurring at approximately 11:00 UTC.
Table 4 displays the horizontal, vertical, and 3D positioning errors (68%). As most solutions are not able to converge to the set benchmark, the convergence time is not reported for this scenario, and the accuracy is estimated over the full duration of data collection. In this case, a clear reduction in error can be noted from SPP to PPP solutions, and an additional gain from the use of HAS corrections is visible. In terms of the north error, HAS-UT, PAULA, and u-blox (with HAS corrections) exhibit similar accuracies, ranging between 30 cm and 43 cm, whereas the Septentrio receiver with HAS corrections presents the smallest error, at 6 cm. Regarding the east component, u-blox with HAS corrections presents the smallest error (64 cm), similar to the HAS-UT in PPP mode (approximately 70 cm). For the vertical error, the HAS-UT, PAULA, and Septentrio with HAS corrections show similar performances, whereas the u-blox presents a larger error. In conclusion, the application of HAS corrections to COTS devices provides a clear reduction in errors with respect to the broadcast-only case for all components. In particular, an improvement of almost 3 m in the 3D positioning error was observed for u-blox.
68th Percentiles of the Horizontal, Vertical, and 3D Positioning Error for the Static Obstructed Scenario
b Dynamic Test
i Preliminary Data Analysis and HAS Message Availability
The subsequent sections present the outcomes from the dynamic test. To assess the severity of the environment in which the vehicle travels, the multipath error along the path (as MLC) is reported in Figure 9. It can be noted that the multipath is very high in certain areas, reaching values close to 6 m. These substantial errors are expected to impact the positioning results, as discussed in the next section.
HAS message availability in the dynamic test and MLC value along the travel path
In addition to the effect of the environment on the measurements, the impact on the HAS message availability is also analyzed. The HAS corrections are retrieved from the Galileo C/NAV message transmitted on E6-B and recorded by the PolaRx receiver during the drive. The corrections are retrieved every 50 s for the code and orbit and every 10 s for the clock. Throughout the entire trajectory, HAS correction availabilities of 94.7% and 95.6% are observed, respectively, for the clock and orbit. The analysis also confirms that the majority of occurrences are concentrated in the areas with the most severe environment conditions, as well as within and just outside the tunnel (on the north–east part of the trajectory), as depicted in Figure 9. This finding is relevant, as the COTS solutions are computed in post-processing by injecting HAS corrections retrieved with a static receiver in open-sky conditions. This implies that HAS corrections are always available for these receivers. In contrast, the two prototype receivers retrieved the HAS products broadcast via the SiS and may be impacted by degraded reception conditions.
It should also be noted that this analysis does not take into account the validity interval of the HAS corrections (300 s for the code/orbit and 60 s for the clock), which would further increase the observed availability. Thus, it can be concluded that the challenging signal reception conditions have little effect on the HAS message availability and that the analysis of position accuracy when using HAS corrections is fair among all receivers.
ii Results for the Dynamic Test
This section introduces the positioning results, with Figure 10 presenting the evolution of the horizontal error for the different receivers as a function of time during a portion of the dynamic test. It can be observed that in this environment, the PPP solution sometime begins to diverge or becomes unavailable, as expected because of the reduced number of available measurements and their lower quality.
Horizontal and vertical error as a function of time for the dynamic test
It can also be seen that the PPP solution needs to re-converge after a data gap. The PAULA receiver (in yellow) integrates an IMU that is used to propagate the solution when measurements are not available. A drift from the IMU can be observed, for example, immediately after 14:16 UTC. The implementation of logic also appears to impact the availability of the solution, for example, at 14:19 and 14:21 UTC, which potentially limits the drift. The benefits of applying the HAS correction are less evident than in the other two test cases. For the configuration using the PolaRx measurements with HAS corrections (red), a small advantage can be noted with respect to the configuration without HAS corrections (purple).
The 68% horizontal, vertical, and 3D positioning errors are reported in Table 5. It can be seen that, even in the dynamic scenario, PPP solutions can significantly enhance position accuracy with respect to the SPP solution. Comparing the performances of the two prototypes, it can be noted that the PAULA receiver provides a more accurate solution in this scenario. This result may be attributed to the integration of IMU measurements. When considering the COTS device, the benefits of applying HAS corrections is clear, with a reduction of approximately 50 cm in the 3D error (68%).
68th Percentiles of the Horizontal, Vertical, and 3D Positioning Error for the Dynamic Test
To analyze the spatial distribution of the errors, the driving path is color-coded in Figure 11, based on the horizontal error reported by the PAULA receiver. To improve the readability of the figure, the color bar has been limited to 3 m. A visual inspection shows that the majority of the errors above 3 m appear to occur when the satellite visibility is low (e.g., in narrow streets) and when the IMU is used to propagate the positioning solution. It can be noted that the largest positioning errors correlate with the areas experiencing larger multipath errors (as shown in Figure 9), confirming that multipath can significantly affect PPP performance in urban scenarios.
Dynamic test path color-coded with the horizontal position error (PAULA receiver case)
Finally, the solution availability is computed over the trajectory, excluding the epochs during which the van was traveling through the tunnel. It is observed that the HAS-UT presents a high availability (over 97%) for both SPP and PPP modes, whereas the other solutions have a comparable availability of approximately 79%. Overall, in comparison to the static scenarios, the HAS-UT solution exhibits slightly larger errors than the PolaRx with HAS corrections, particularly in the vertical dimension, but significantly higher availability. In contrast, the PAULA receiver demonstrates lower availability, despite integrating an IMU, but higher accuracy. These results highlight the fact that, beyond the PPP estimation, different PVT provision strategies may be implemented in the receivers.
6 CONCLUSIONS
The main objective of this paper was to evaluate the benefits of applying Galileo HAS in various environments representative of potential user applications (e.g., urban environment for automotive applications). A dedicated test campaign was organized, and a methodology was proposed to compare the performance of HAS-based solutions against other real-time stand-alone PVT solutions. The test results in terms of position accuracy and availability for both static and dynamic conditions, including open-sky and urban canyon scenarios, have been presented.
The benefits of applying HAS corrections were assessed by using devices of different grades, including two HAS receiver prototypes. While a direct comparison of the tested receivers would be unfair because of the lack of understanding of the internal processing logic and algorithms implemented, the analysis clearly indicates that the usage of HAS corrections leads to a significant reduction in positioning errors for all receivers tested in all environments. In static open-sky conditions, it was found that all PPP solutions (with and without HAS corrections) achieve a final horizontal accuracy below the 20-cm benchmark. After convergence, the errors remain fairly stable, at the centimeter level for the solutions using HAS and at the decimeter level for the solutions using broadcast products only. The use of HAS corrections demonstrated a significant benefit in terms of convergence time, with the convergence time decreasing by a factor of 3 for the high-grade PolaRx receiver and an even a larger reduction for the u-blox.
As expected, the static receivers under obstructions were not able to converge to the set benchmark, likely because of the geometric conditions and the presence of multipath. For the COTS devices, the application of HAS corrections instead of broadcast products led to different results: the PolaRx exhibited similar performance, whereas a reduction of almost 3 m (3D error) was observed for the u-blox receiver.
Finally, the benefits of applying HAS corrections were evaluated in an urban environment with a dynamic vehicle. The analysis demonstrated a correlation between the magnitude of local errors and the performance of the PPP solutions. Despite this, a marginal positioning improvement of 50 cm was noted for the PolaRx receiver when using the HAS corrections versus broadcast products. The results for the HAS prototypes highlighted the different implementation choices made to balance the accuracy and availability of the positioning solution.
In addition to the positioning domain analysis, the availability of the HAS message in harsh dynamic conditions was assessed, and the corrections retrieved from the SiS were found to be available approximately 95% of the time. The possibility of increasing this availability by exploiting the applicability intervals of past valid corrections was also discussed. Based on these results, it can be concluded that the availability of HAS corrections from the SiS seems to be sufficient for automotive users in urban scenarios, without the need for remote assistance.
While this paper focuses on the use of products disseminated in the SiS, PPP can also be performed using different products and corrections from alternative service providers. A comparison of the PPP using HAS corrections with respect to other products, including IGS rapid and final products, will be considered in future work.
HOW TO CITE THIS ARTICLE
Cucchi, L., Damy, S., Gioia, C., Motella, B., & Paonni, M. (2024). Galileo high accuracy service: Tests in different operational conditions. NAVIGATION, 71(4). https://doi.org/10.33012/navi.665
ACKNOWLEDGMENTS
This study tested the Galileo HAS-UT, developed by EUSPA under the project “Galileo Reference High Accuracy Service (HAS) User Algorithm and User Terminal,” ref. GSA/OP/25/19, and the PAULA User Terminal, developed under the project “Precise and Authentic User Location Analysis,” funded by the European Commission Directorate-General for Defence Industry and Space (DG-DEFIS) under contract DEFIS/2020/OP/0002.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Jump to section
Related Articles
Cited By...
- No citing articles found.