InterestingITS/smartphone_in_vehicle_multisensor

Smartphone-in-Vehicle Multisensor Dataset

The smartphone_in_vehicle_multisensor dataset contains synchronized raw multisensor recordings collected to evaluate smartphones as low-cost in-vehicle sensing and vehicle-to-everything (V2X) terminals. Data were collected with two test vehicles, a Volkswagen Golf and a Volkswagen Passat, in three representative driving environments: highways, urban expressways, and urban roads. Each vehicle was driven for 2 h in each environment, yielding 12 h of onboard reference measurements and, through simultaneous recording by six phones, 72 h of smartphone observations.

The six phones cover recent and earlier generations from three vendors: Apple iPhone 15 Pro Max and iPhone 8 Plus, Samsung Galaxy Z Fold5 and Galaxy S10, and Xiaomi Mi 15 and Mi 11. All phones used the Sensor Logger app to record GNSS, motion and attitude measurements, and device-status information.

Dataset summary

• Vehicles: Volkswagen Golf and Volkswagen Passat
• Driving environments: highway, urban expressway, and urban road
• Recordings: 14 recording directories
• Smartphones: six devices from Apple, Samsung, and Xiaomi
• Reference systems: UM982 RTK-GNSS receiver and ELM327 OBD-II interface
• Data volume: approximately 92.6 GiB in 3,211 files
• License: Creative Commons Attribution 4.0 International

Experimental setup

The six smartphones were mounted on a common bracket near the upper central area of the windshield for synchronous data acquisition. A roof-mounted UM982 RTK-GNSS receiver and an ELM327 OBD-II interface provided positioning and vehicle-state reference streams. Both reference devices were connected to a host computer, and the RTK-GNSS system received CORS/RTCM correction data. Timestamps from the different sources were calibrated using a network time protocol (NTP) server.

Figure 1. Data acquisition hardware setup, including the six smartphones, onboard reference devices, host computer, and the two test vehicles.

Repository structure

smartphone_in_vehicle_multisensor/
|-- {VW_Golf,VW_Passat}/
|   `-- {Expressway,Highway,Urban_Road}/
|       `-- recording_<n>/
|           |-- Galaxy_S10/
|           |-- Galaxy_Z_Fold5/
|           |-- iPhone_15_Pro_Max/
|           |-- iPhone_8_Plus/
|           |-- Mi_11/
|           |-- Mi_15/
|           |-- GNSS/
|           |   |-- RTK_Results/
|           |   |-- Raw_Data/
|           |   |-- Trajectory_Data/
|           |   `-- RTCM_Data/
|           |-- Ground_Truth/
|           `-- OBD-II/
`-- assets/

Each smartphone directory contains separate CSV files for the sensors available on that device, together with Metadata.csv. The original ZIP exports are intentionally omitted because their extracted CSV files are included directly.

The GNSS directories contain:

• RTK_Results: receiver solutions, tabular speed results, and diagnostic plots.
• Raw_Data: raw binary/ASCII receiver logs and converted GNSS observation/navigation files.
• Trajectory_Data: trajectory logs, KML files, and KML visualization assets.
• RTCM_Data: RTCM correction logs and converted reference-station observation files.

Ground_Truth contains synchronized RTK/OBD reference tables. OBD-II contains timestamped vehicle-speed measurements.

Data sources and sampling rates

Data source	Device	Operating system	Recorded data and sampling rate
Apple	iPhone 15 Pro Max	iOS 26.3.1 / 26.5	Motion/Attitude (100 Hz); GNSS (1 Hz); Device Status (1 Hz)
Apple	iPhone 8 Plus	iOS 16.7.15	Motion/Attitude (100 Hz); GNSS (1 Hz); Device Status (1 Hz)
Samsung	Galaxy Z Fold5	Android 16	Motion/Attitude (100-500 Hz); GNSS (0.1-1 Hz); Device Status (1-25 Hz)
Samsung	Galaxy S10	Android 11	Motion/Attitude (100-400 Hz); GNSS (0.1-1 Hz); Device Status (1-25 Hz)
Xiaomi	Mi 15	Android 16	Motion/Attitude (100-500 Hz); GNSS (0.3-1 Hz); Device Status (1-25 Hz)
Xiaomi	Mi 11	Android 13	Motion/Attitude (100-400 Hz); GNSS (0.3-1 Hz); Device Status (1-25 Hz)
RTK-GNSS	UM982	—	RTK Results (20 Hz); Raw Data (1-20 Hz); Trajectory Data; RTCM Data
OBD-II	ELM327	—	Vehicle Speed (5-9 Hz)

Sampling rates may vary within the ranges shown above. Retain the original timestamps and consult each Metadata.csv when aligning data across devices and sources. Smartphone device id values have been replaced with stable, model-specific anonymous aliases; the alias remains consistent across all recordings from the same physical test phone.

Known data notes

• VW_Golf/Urban_Road/recording_2/Galaxy_S10/MagnetometerUncalibrated.csv is unavailable because the corresponding entry in the original source archive was irrecoverably corrupted.
• All 84 Annotation.csv files are empty placeholders because no manual annotations were recorded.
• All 14 Mi_15/Location.csv files are empty placeholders; use LocationGps.csv and LocationNetwork.csv for the available provider-specific location streams.
• VW_Golf/Urban_Road/recording_1/Ground_Truth includes two additional *_clean_1.csv and *_clean_2.csv interval products alongside the complete ground-truth file.
• VW_Passat/Urban_Road/recording_2/OBD-II includes two source segments and OBD_Data_20260520_combine.csv, which combines the segments and fills their intermediate gap.

Download

Install the Hugging Face Hub CLI and download the full repository:

hf download InterestingITS/smartphone_in_vehicle_multisensor \
  --repo-type dataset \
  --local-dir smartphone_in_vehicle_multisensor

To download only one vehicle or scenario, use an include pattern, for example:

hf download InterestingITS/smartphone_in_vehicle_multisensor \
  --repo-type dataset \
  --include "VW_Golf/Highway/**" \
  --local-dir smartphone_in_vehicle_multisensor

Dataset scope and intended use

This release is intended for research on smartphone sensing, GNSS/IMU characterization, vehicle-state estimation, multisource synchronization, cross-device comparison, V2X data quality, and traffic-flow or platoon analysis. Its multi-vendor, multi-generation, and multi-scenario design supports comparisons of sensing behavior across devices and driving environments.

The release contains the collected raw smartphone measurements and the corresponding RTK-GNSS and OBD-II reference streams. It does not include the processed smartphone speed-error dataset, fitted error-distribution parameters, or replication materials used in the accompanying stability analysis.

License

This dataset is released under the CC BY 4.0 license. You may share and adapt the material provided that appropriate credit is given.

Citation

Until a formal dataset DOI is available, cite the Hugging Face repository:

@dataset{interestingits_smartphone_in_vehicle_multisensor_2026,
  author    = {InterestingITS},
  title     = {Smartphone-in-Vehicle Multisensor Dataset},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/InterestingITS/smartphone_in_vehicle_multisensor}
}

Related publication

This dataset supports the study Are Smartphones Viable for V2X Transition? A Platoon Perspective Stability Analysis under Real World Errors. The formal dataset citation and article DOI will be added after publication.