Raw GNSS data collected using smartphones and low-cost receiver under optimal and sub-optimal conditions

Abstract

The miniaturisation and decrease of price are amongst the main current trends in the area of Global Navigation Satellite Systems (GNSS) receivers. Besides standalone receivers also receivers incorporated into Android devices can provide raw GNSS measurements thus enabling much wider options, formerly restricted to devices of much higher price. The article describes two datasets. The first was collected using a Xiaomi Mi 8 smartphone with and without application of a simple ground plane. In the second we compared a smartphone receiver (Google Pixel 5) with a standalone low-cost receiver (u-Blox ZED F9P). In both cases the datasets consist of multiple measurement sessions, also considering the conditions where the reception of GNSS signals was obstructed by trees’ canopy. The datasets are focused on repeatability (multiple measurements), influence of external conditions (canopy and foliage state) and the devices used.

Specifications Table

Subject	Computer Science: Signal Processing
Specific subject area	Raw GNSS data acquired using two smartphones and a low-cost receiver, collected during multiple static sessions under optimal and suboptimal conditions.
Data format	Calculated reference coordinates of test points -txt (reference coordinates) Raw GNSS data (in most cases converted to RINEX format) ­txt (raw smartphone GNSS measurements from GNSSlogger app – only for Xiaomi Mi8) ­*o (RINEX observation data from tested receivers and used base stations) ­*n, .*l, .*g, .*c, .*c (satellite system-specific RINEX navigation data) ­*p (multi-GNSSRINEX navigation data) ­ubx (u-Blox proprietary logging format – for u-Blox ZED F9P) ­obs (RINEX observation data) ­nav (RINEX navigation data – combined) ­sbs (RINEX SBAS data) Auxiliary files used for postprocessing -conf (configuration files used for postprocessing in the RTKLIB software) -py (Python script for batch postprocessing and plotting of results)
Type of data	Crucial data (rover and base GNSS data) are in the Receiver Independent Exchange Format – RINEX, organized in folders according to receiver type, conditions, and time of the recording.
Data collection	The data were collected on three test points representing differing external conditions (open area, partial and full tree canopy). For the first dataset, we used a Xiaomi Mi 8 smartphone. 10 minutes of measurements were recorded on each point with 40 repetitions - 40 days in approximately the same daytime. In the first 20 days, only a simple smartphone holder was attached to a tripod, while for the next 20 days, we employed a simple ground plane. For the second dataset, we used a Google Pixel 5 smartphone and a u-Blox ZED F9P receiver on a joint ground plane. The data were collected for 20 minutes on each point, with 10 repetitions (days) under leaf-on conditions and 10 repetitions under leaf-off conditions.
Data source location	The data were collected in the near vicinity of the main building of the Technical University in Zvolen, Slovakia (48°34′19"N 19°07′09"E). For precise locations of the test points, please, see the datasets.
Data accessibility	Repository name: Mendeley Data Dataset 1 DOI: 10.17632/5prmtwgph3.3 Direct URL to data: https://data.mendeley.com/datasets/5prmtwgph3/3 Dataset 2 DOI: 10.17632/83bvxzx3bj.3 Direct URL to data: https://data.mendeley.com/datasets/83bvxzx3bj/3
Related research article	J. Tomaštík, T. Everett, Static Positioning under Tree Canopy Using Low-Cost GNSS Receivers and Adapted RTKLIB Software, Sensors. 23 (2023) 3136. 10.3390/s23063136

1. Value of The Data

• •The data can be used to assess the usefulness of raw GNSS data from Android smartphone receivers, for example by comparison to the data from low-cost and professional GNSS receivers. Focus on repeatability – multiple measurement sessions in similar daytime, providing full phase and pseudo-range measurements in optimal and suboptimal conditions.
• •Possibility to test all GNSS postprocessing methods, new approaches and algorithms, and examine viable adaptations to deal with lower-quality data, coming from both receiver characteristics and suboptimal conditions.
• •Repeated, longer-term measurements are often complicated to conduct, moreover considering the combination of ideal and sub-optimal conditions (open area and tree canopy in our case). Such measurements could describe the behavior of the tested method under such conditions and facilitate the development of algorithms to mitigate the adverse effects.
• •Researchers, GNSS software and hardware developers, and professionals working with localization, especially under adverse conditions of tree canopy (nature conservation, forestry, urban greening, etc.)

2. Data Description

The low-cost GNSS receivers are currently the most prevalent, with two dominant directions of the research identifiable with this regard: smartphones and standalone receivers. Since 2016, the Android OS smartphones provide raw GNSS measurements [1], thus theoretically enabling all the post-processing methods available for higher tier GNSS devices. Vast number of studies focuses on this issue, ranging from theoretical aspects and evaluation of data quality to practical applications [2]. The standalone receivers with prices in few hundred Euros are also being continuously documented as able to provide centimeter-level accuracy, even with fast measurement methods (e.g. [3,4]). A frequent drawback, which we tried to address in our experimental design, is that the studies only take place under ideal conditions and the measurements are not repeated.

In the first dataset, acquired using the Xiaomi Mi8 smartphone, the data are organized in folders according to measurement day. The syntax includes month, day, and year. The “gp” suffix means that the data were collected with the application of a ground plane. Each “day” folder includes four subfolders and a combined navigation file with .*p suffix. Three subfolders are named according to measurement time, where the earliest contains data from the “open area” point, the middle from the “particular canopy” points, while the latest includes “full canopy” point data. These subfolders contain smartphone (GEO* prefix) and base (BBYS*) RINEX files. Data acquired using the smartphone and the GNSSlogger application can be found in the “GNSSlogger” folder, where the time order of the files is the same as the aforementioned. Besides the “day” folders, the root folder contains the “reference.txt” file with reference coordinates of the test points.

The second dataset is organized into three folders and a “reference.txt" file with reference coordinates. The “RTKLiB config files” contains configuration files for the smartphone (phone.conf) and the standalone receiver (f9p.conf), used for post-processing in the RTKLib package. The “python script” folder includes Python scripts for batch processing of the GNSS data (batch.py) and for the plotting of resulting coordinate errors (plotSolErrors.py). The measurement data are included in the “GNSS data” folder, which is further divided according to season into the “leafon” and “leafoff” subfolders, and subsequently based on device to “pixel” (for Google Pixel 5 smartphone), “F9P” (for the u-Blox ZED F9P receiver), and “base” subfolders. These are subdivided into “day” subfolders with the same syntax as in the first dataset (month, day, year). For the “pixel” and “F9P” folders, each “day” subfolder contains RINEX files for the three test points where, according to letter (pixel) or time (F9P) in the filename, it is the open area, partial canopy, and full canopy from the earliest to the latest. The same principle is used in the “base” folder, where we used another level of “time” subfolders due to different naming of the base data. The “batch.py” script is adapted to this data structure.

3. Experimental Design, Materials and Methods

The datasets were used as a basis for two studies [5,6] focused on applicability of GNSS data from low-cost and smartphone GNSS receivers.

As a basis for testing and comparisons, three test points were established under differing conditions. The first point was established on an open area without sky view obstructions. The second point was placed under a mixed tree canopy, however, with an overhead gap. The third point was placed under a canopy of coniferous trees, without any particular gaps. Their reference coordinates were calculated based on static GNSS measurement with a survey-grade receiver during the leaf-off season, in the case of the “full canopy” point in combination with total station measurements. For the second dataset, this point was reestablished due to accidental destruction. The reference coordinates were calculated using static GNSS method in the ETRS89 coordinate system (realization ETRF2000 epoch 2008.5). This coordinate system realization was used because the national SKPOS service broadcasts the coordinates of reference stations (CORS or VRS) in this realization as it is directly transformable into the national S-JTSK coordinate system. Therefore, when using the “base” data provided in the datasets for relative positioning methods, the resulting coordinates are in this realization.

For the Dataset 1 [7], the data were collected using a Xiaomi Mi 8 smartphone. For the first 20 measurement repetitions, the phone was attached to a tripod using a simple phone holder (Fig. 1). For the next 20 measurements, we deployed a simple ground plane. The data were recorded using the Geo++ Rinex Logger application simultaneously with the GNSSlogger application. Ten minutes of raw data were recorded for every point for every repetition (approximately the same time during multiple days), regularly in order “open area” – “partial canopy” – “full canopy”. Correction (base) data were acquired using the continuously operating reference station (CORS) Banská Bystrica (BBYS) belonging to Slovak Real-Time Positioning Service (SKPOS).

Fig. 1.

Overall setup used for the first dataset (a) using simple ground plane (b) and smartphone holder (c).

For the Dataset 2 [8], we used a Google Pixel 5 smartphone and a u-Blox ZED F9P receiver. The antenna of the F9P receiver was placed on a simple ground plane together with the phone. The ground plane was oriented using a compass and the center of the F9P antenna was shifted 7,5 cm to the East while the center of the Pixel 5 phone was shifted 6,5 cm to the west from the reference position (Fig. 2). The Ge++ Rinex Logger application was used to record the data in the RINEX format for the Google Pixel 5 smartphone. The data for the u-Blox receiver were recorded using the u-center software in the .ubx format and subsequently converted to the RINEX format using the “rtkconv” module of the RTKLib package. The data were recorded for 20 min on every point, while the point order was the same as in the previous experiment. The measurements were repeated ten times during the leaf-on season and ten times during the leaf-off season. In this case, the correction data were downloaded using the Zvolen (ZVOL) CORS of the SKPOS service.

Fig. 2.

Overall setup used for the second dataset (a) and the detail on the devices placement (b).

In both cases, the ground planes consisted of a 2 mm thick steel plate. The white finish on the top side (paint for the Dataset 1 and glued paper for the Dataset 2) served as a prevention from overheating, especially considering the open area point during sunny days, and for devices’ centration marks.

The data were analyzed in the related papers, where we firstly evaluated characteristics of recorded signals (e.g. signal-to-noise density, multipath, number of satellites) and subsequently tested the positioning performance in the RTKLib software, using following basic settings:

• -Positioning mode: Static
• -Constellations: GPS, GLONASS, Galileo, Beidou
• -Frequencies: L1/L5 for smartphones, L1/L2 for the u-Blox ZED F9P

For more detailed settings we refer to related studies [1,2] and configurations files (.conf) in the Dataset 2.

Limitations

• -The correction data from CORS stations can be considered significantly different (since the ionospheric and tropospheric conditions at two location might be different) as the baseline to the BBYS station is ∼20 km, while ∼2.5 km for the ZVOL station. This was caused by unavailability of the ZVOL station during the first experiment.
• -The effects of the smartphone hardware causing noise and interference, and the impact of the different measurement conditions on the quality of the data. For example the presence of trees blocks or attenuates GNSS signals, which leads to errors in the positioning results. However, this was one of the main challenges to overcome in the mentioned studies.
• -The parameters of the antenna phase center (APC) of the smartphones were unknown. This can be a constraint for achieving the highest accuracy. We used the centre of the device as the reference point for comparison of coordinates.
• -It is needed to transform the reference coordinates from the ETRS89 (ETRF2000 2008.5) reference frame when using some other postprocessing methods, especially when using absolute (not relative) positioning methods. We suggest using the ETRF/ITRF Coordinate transformation tool [9] in such a case. The closest EUREF station with known velocities is the BBYS00SVK station.

Ethics Statement

The authors have read and follow the ethical requirements for publication in Data in Brief and confirm that the current work does not involve human subjects, animal experiments, or any data collected from social media platforms.

CRediT authorship contribution statement

Julián Tomaštík: Conceptualization, Methodology, Resources, Formal analysis, Writing – original draft. Matej Varga: Formal analysis, Investigation, Writing – review & editing. Tim Everett: Software, Validation, Visualization.

Data Availability

• Raw GNSS data, Xiaomi Mi 8, 10 minutes, 3 points, 40 days (Original data) (Mendeley Data)

• u-Blox F9P and Google Pixel 5 (repeated 20 minutes static measurements under variable conditions) (Original data) (Mendeley Data).