Integrity Monitoring of Commercial RTK Receivers in Static Open Sky and Kinematic Urban Environments Scenarios

authored by
Ali Karimidoona, Steffen Schön

1 Introduction
Currently, navigation of autonomous vehicles are increasing rapidly; GNSS provides as only sensor absolute positioning information. To meet lane level accuracy, Real Time Kinematic (RTK) (Odijk and Wanninger 2017) has to be used. Here, the distance dependent GNSS errors mainly ionospheric, tropospheric and satellite orbit errors can be modelled by a network of reference stations. (Wübbena et al. 2005). (Rizos 2002; Wanninger 2004). In this contribution, it is intended to evaluate the performance of geodetic RTK receivers in a changing environment and mainly on a moving vehicle in an urban environment.
Because of the complicated changing environments for the autonomous platforms and the importance of safety issues especially in urban areas, integrity monitoring is of great importance in navigation systems. The performance of a navigation system depends on accuracy, continuity, availability and integrity (ICAO Doc 9849 2005; RTCA 2006). The concepts of integrity monitoring have been well developed for aviation applications. However, we notice that the integrity monitoring concept has not yet been well developed for Network RTK. First results are given by (Weinbach et al. 2018), a general literature review of concepts for urban navigation are collected by (Zhu et al. 2018).
In this experiment three high-end and one low cost receivers were used. The four receivers are firstly tested in a static experiment and then using a van, the receivers are tested in a kinematic mode.
2 Methodology
2.1 Stanford Diagram
Stanford diagram is a tool which was introduced in the aviation application for WAAS integrity monitoring (Stanford University 2009). This tool uses mainly three parameters to establish the diagram: position error, protection level and alert limit. Based on these parameters the operation mode of the navigation system w.r.t. a known reference or nominal solution is classified.
The Stanford diagram based on the operation mode has six regions which are defined by the interaction of the given threshold (Alert limit), the confidence level of the estimated position (protection level) and the (unknown) position error (PE).
The deviation from the reference is defined as the Position Error (PE). By reference it meant the true position or sometimes it is called the ground truth. In reality we do not know any parameter perfectly, but we have some estimates of the parameters. So, when we speak about reference, we mean of a more accurate and reliable estimation of the desired parameter. For instance, in a static measurement, the reference can be the very accurate known coordinates of a stationary pillar. In kinematic applications, this reference can be obtained from sensor fusion e.g. integrating the GNSS data with IMU (Inertial Measurement Unit) using more accurate tightly coupled solution (Groves 2013).
In this experiment for calculating the protection levels, Coordinate Quality (CQ) which are position accuracies in 1 sigma level computed by the GNSS receivers, are used. So, the protection levels can be calculated from these values for horizontal and vertical components respectively by applying the appropriate scale factors, for the Rayleigh or Normal distribution, respectively. .
The Alert Limit (AL) is defined as the maximum tolerable error of the system (max of PE or PL). The AL is a parameter to be selected based on the positioning requirements. In this experiment for the kinematic part, the AL for horizontal positioning has been selected as 41 cm and 1.4 m for vertical component, respectively. These values are based on a research done by (Reid et al. 2019). In this research, the requirements for passenger vehicles in local streets have been calculated as 29 cm for the lateral and longitudinal components and 1.4 m for vertical component.
3 Experiment
This experiment consists of two phases, the static and kinematic one. In the static part four receivers: namely a Leica GS18 T, Trimble R12i, Septentrio Altus NR3 and Ublox ZED-F9P were installed on top of four pillars MSD1, MSD2, MSD5 and MSD3, respectively, on the rooftop of the Geodetic Institute of Leibniz University Hannover. The first three receivers are high-end receivers.
The pillars have very precise coordinates which have been calculated by post processing of very long static observation periods. This rooftop has a very good sky view without any blockage around.
All these receivers have internal cellular modems and used the SAPOS Network RTK service provided in Lower Saxony state applying the GPS + GLONASS corrections. The static part started at 6:45 UTC 29th July 2021 to 7:15 UTC the same day for half an hour. The three first receivers were logging 10 Hz data, while UBLX was set to 5 Hz. In the next step in the same day, the kinematic part was conducted from 13:51 to 16:18, driving an 8 shaped trajectory for twelve rounds in two sets of six rounds. This loop is approximately 1 km long, so the total length of the trajectory is 12 km.
In kinematic part, the four receivers were mounted on top of the van of the institute. In addition to these receivers, a reference antenna (NAX3G+C) was connected to the iMAR iPRENA IMU. This IMU is a high-end inertial measurement unit which was installed inside the van.
4 Results
The scatter plot of the fixed RTK results of the static part show almost a circular distribution of the horizontal results for these four receivers tested. The more circular the error ellipse is, the more consistent is the receiver in lateral and longitudinal components. Although the GS18 results are slightly elongated in north-west south-east direction but it has the smallest distribution among other receivers. The GS18 has the deviation from the reference as 6.8 mm, 1.2 mm and -7.1 mm respectively for East, North and Up components. These values are for TR12 3.9, 5.5, 6.5, for ALTS 1.3, -2.1 -4.9 and for UBLX 0.1, 9.1 and 7.4 mm. The results for the standard deviation for each of devices under test are as follows: GS18 3.4, 3.8 and 8.2, TR12 4.8, 4.0, and 8.1, ALTS 6.4, 5.4 and 9.1 and UBLX 4.7, 5.2 and 9.5 mm respectively for East, North and Up components.
All of these receivers have fixed solutions for this static part. Looking at 2D-CQ (horizontal Coordinate Quality) the overall range of precision can be considered; while GS18 and TR12 have results almost better than 1 cm, the ALTS results are in range of 2 cm. UBLX shows very good smooth results better than 2 cm.
As stated earlier, the kinematic experiment was conducted in twelve rounds. In areas with more open sky, all receivers show good results with fixed solutions. However in very narrow streets with buildings on both sides and also building at both ends, the visibility is not only adverse across the street, but also along the street. In these challenging situations, GS18 shows some missing results in the challenging street, and also some navigated results at one end, and the overall large number of code solutions. TR12 has some fixed solutions in challenging street, which is promising. And some float solutions in areas where e.g. ALTS has fixed solutions. ALTS experiences navigated solutions in critical points of both ends of the challenging street. UBLX is performing good in the first round where most of the path has fixed solution and in the difficult part of the challenging street has float solutions.
Looking at the Position Error (PE) and the Protection Level (PL), for the first round of the four receivers, an overall repeated pattern can be seen. Although there are some differences between these four receivers, but some patterns can be seen in PE that repeat in every receiver at the same location. In order to better understand the reason behind this performance, the logged GNSS raw data were analyzed in more detail.

Analyzing signal strength, the ALTS shows very low C/N0 values, this is because of 0 degree cut-off angle for tracking. The 10 degree cut-off angle is applied for PVT solution. GS18 shows higher values for the same signals in comparison to other. Another parameter to investigate is the signal continuity of the carrier phase signals. Overall, ALTS has less interruptions in continuity than the others.

Another parameter is Linear Combination (LC) of Melbourne-Wübbena (MW). It can be seen that satellites 1, 3, 22 and also 21 have very near zero values or very small fluctuations, which means these satellites have no multipath and also the signal has been continuous. These satellites have high elevation angle and are in good visibility condition. On the other hand, satellites 4, 31 and 32 show multipath effect, and the jumps in the LC indicate phase loss in the signal for the corresponding satellite.
All the Stanford diagrams for all the rounds have been calculated. In each Stanford diagram, different operation modes have a percentage. The Nominal Operation results are compared to evaluate the performance of the receivers in kinematic part of the experiment.
5 Conclusion
An experiment testing four integrated RTK receivers was conducted in static and kinematic modes using Network RTK correction service. In static part, receiver Leica GS18 T, had the best results, having least standard deviation of the RTK results of 3.4 mm and 3.8 mm for E and N components respectively. In the kinematic part of the experiment, the receiver Septentrio Altus NR3 had 13.6 % of the nominal operation results in Stanford diagram analysis which was more than the other candidates.

Institute of Geodesy
Conference contribution
No. of pages
Publication date
Publication status
Peer reviewed
ASJC Scopus subject areas
Aerospace Engineering, Electrical and Electronic Engineering
Electronic version(s) (Access: Closed)

Details in the research portal "Research@Leibniz University"