WHU-Smartphone Dataset

A large-scale multi-traffic scene GNSS dataset of Android smartphones and a low-cost GNSS module collected in China

To conduct a comprehensive analysis of the performance and accuracy of smartphone positioning, we released the open-source WHU-Smartphone dataset , which is a large-scale, multi-scene, and multi-mode GNSS dataset collected by four Android smartphones from different manufactures and a low-cost GNSS module with high-precision ground truth. Our dataset contains several hundred driving routes with more than 2500 km from three large cities, including open-sky or shaded urban roads and highways. Multifrequency observation space representation (OSR) corrections of real-time kinematic (RTK) positioning with multiple GNSS systems, particularly those that supported the BeiDou Navigation Satellite System (BDS)-2 and BDS-3 systems, were implemented. On the basis of WHU-Smartphone dataset, we proposed some tools for converting data format and evaluating position accuracy with self-developed RTK positioning algorithm. We hope that the open-source GNSS dataset can help researchers verify the effectiveness of proposed algorithms and compare accuracy between algorithms with a large amount of dataset in various environments. We also expect more extensive progress in robust positioning, integrity monitoring, and cooperative positioning, and more advanced studies in challenging environments.

If you have any question for our dataset, please contact me via email: [email protected]

Related Papers:

• Wang, Yixin, Chuang Qian, and Hui Liu. "WHU-Smartphone: A Large-Scale Multi-Traffic Scene GNSS Dataset of Android Smartphones and a Low-Cost GNSS Module." IEEE Transactions on Intelligent Transportation Systems (2024).

if you use our dataset for your academic research, please cite our related papers

1 Platform Equipment

The platform is equipped with the following sensors:

• Four Android smartphones [Xiaomi MI8; Huawei P20; Huawei P30; and VIVO X30]
• low-cost GNSS module (MXT906A,BDS/GPS,L1)
• GNSS/IMU reference benchmark (Trimble BD982 geodetic receiver and Honeywell Hg4930 IMU)

2 Device Information

The specific information of the smartphones and the GNSS module is provided in the table.

Device (Abb.)	Xiaomi 8 (XIM8)	Huawei P20 (HP20)	Huawei P30 (HP30)	VIVO X30 (VX30)	MXT module (MXT)
GNSS chips	Broadcom 47755	Kirin 970	Kirin 980	Exynos 980	MXT906A
L5	YES	NO	YES	NO	NO
ADR	YES	YES	YES	NO	YES
Duty Cycle	OFF	ON	OFF	\	\
GNSS Signals	GPS: L1+L5; GLO: L1; GAL: E1; BDS: B1	GPS: L1; GLO: L1; BDS: B1	GPS: L1+L5; GLO: L1; BDS: B1	GPS: L1; GLO: L1; GAL: E1; BDS: B1	GPS: L1; BDS: B1
BDS-3	YES	YES	YES	YES	NO
Other	\	Duty cycle occurs mainly in static mode	Abnormal carrier observation noise (0.2m-0.5m)	Unavailable mAccumulatedDelta-RangeState	\

• Base Station for RTK

  • base station with fixed coordinates: GPS L1/L2/L5, GLO G1/G2, GAL E1/E5, BDS-2&BDS-3 B1/B2/B3, 1Hz
  • virtual reference station (VRS) service：GPS L1/L2, GLO G1/G2, BDS B1/B2, QXFZ, provided by Qianxun Spatial Intelligence, Inc
  • RTCM3 format

  Notes: For data security reasons，the observation files of base station are available upon request by contacting the author([email protected]).

3 Driving Routes

The total length of the driving routes in the three cities was about 2500 km, with large coverage areas and rich test scenes, including open-sky/shaded urban roads and highways.

• Test Environments

• Beijing (39°26′N ~ 41°03′N)

  The driving test in Beijing was from October 14, 2020 to October 22, 2020, and the total length of the driving routes was 978.1 km.

• Wuhan (29°58′N–31°22′N)

  The driving test in Wuhan was from July 3, 2020 to September 2, 2020, the total length of driving routes was 713.3 km.

• Shenzhen (22°24′N–22°52′N)

  The driving test in Shenzhen was from September 17, 2020 to September 25, 2020, and the total length of the driving routes was 795.5 km.

4 Data Format

We provide the raw GNSS observation data in two formats:

• MXT GNSS module(YYYY_MM_DD_hh_mm_ss_bluetooth_data_log.txt):

  • RTCM3 format,you can use RTKLIB to convert RTCM file to RINEX file.

• Smartphones(YYYY_MM_DD_hh_mm_ss.sss_spare_log.txt):

  • JSON Format
  • Spare file

  Data Type	Description
  current_Start_Timestamp	Observation data timestamp of smartphones (local time)
  gga	Positions calculated by GNSS chipset (NMEA-0183 format1)
  imei	International mobile equipment identity of smartphone
  obsd_t	Raw GNSS observations at current epoch
  d	Doppler observation
  l	Carrier-phase observation
  lli	Loss-of-lock indicator
  p	Pseudorange observation
  prn	Satellite index
  snr	Signal-to-noise ratio
  type	Observation code (RINEX Version 3)
  vd	Variance of Doppler observation
  vl	Variance of carrier-phase observation
  vp	Variance of pseudorange observation
  satnum	Number of visible satellites at current epoch
  t	Observation data timestamp of smartphones (GPST)

1.5 Data Conversion Tools

• ConvSpare.bat: Extract the position results in NMEA-0183 format calculated by the GNSS chipset (chipset solution) from the spare file and then convert them into files with “pos”(RTKLIB format) and KML files(driving tracks).

  Notes:

  • 1.Put "ConvSpare.bat" in a folder with the spare files.
  • 2.You can only extract all the files of the same day at one time, the Year/Month/Day[2020/08/15] in "ConvSpare.bat" needs to be the same as that in the spare file.

• spare2rinex: The function converts available raw GNSS information extracted from the spare file into RINEX file (RINEX 3.x format), which can be used in RTKLIB.

  • Supporting Visual Studio 2017/2019/2021.
  • The functions are written in C language, and they can be extended and modified in accordance with one’s needs.

Footnotes

1. A voluntary industry standard developed for GNSS receivers, which contains the position, velocity, and time message.(https://en.wikipedia.org/wiki/NMEA_0183) ↩