Dataset of Pseudonymized GPS Telemetry Records from 100 Vehicles in Pakistan (Teltonika FMB920 Devices)
Description
This dataset contains pseudonymized vehicle-tracking telemetry from 100 vehicles fitted with Teltonika FMB920 tracking units and monitored across Pakistan by a commercial vehicle-tracking provider. It spans twelve consecutive months (March 2024-February 2025) and comprises 61,331,270 message-level records. Vehicle positions fall within roughly 24.2-35.9° N and 66.6-75.7° E. The dataset was compiled to provide an openly shareable, privacy-preserving record of real fleet operation for research in intelligent transportation and geospatial analysis. Typical uses include reconstructing per-vehicle trajectories and trips, analysing speed, stop and ignition patterns, and assessing GPS-fix quality. The data are also suitable for developing or benchmarking anomaly-detection and trust-scoring methods for Internet-of-Things systems. The records are supplied as twelve monthly archives named with ISO year-month dates (telemetry_2024-03.zip through telemetry_2025-02.zip); each holds one CSV file with the matching name (telemetry_2024-03.csv through telemetry_2025-02.csv). Each row is a single message from one device and has 15 fields: stable vehicle and device pseudonyms, device model, GpsTime, message identifier, longitude (X), latitude (Y), speed, travel direction, altitude, visible-satellite count, ignition state, main battery voltage, packet-validity flag, and Canal. GpsTime is the DateTime when the packet was created by the device. Canal indicates if the message is regular (1) or in response to server ping (0). Vehicle registration numbers and device IMEIs were removed before release and replaced with stable pseudonyms. Column definitions are given in data_dictionary.csv. A 100,000-row sample is provided in sample.csv for quick inspection without downloading the full archives. It uses the same 15 fields as the monthly files and contains 10,000 records from each of ten vehicles (vehicle_020, vehicle_037, vehicle_044, vehicle_047, vehicle_051, vehicle_075, vehicle_076, vehicle_077, vehicle_079 and vehicle_081), all drawn from the first five days of March 2024 (1-5 March 2024). The identifiers Vehicle_Id and Device_Imei_Id are stable across all months, so a vehicle or device can be followed through the whole period. Reporting cadence is not fixed: intervals between consecutive messages vary by vehicle and vehicle state, and Ignition_Status and VectorSpeed help separate moving, idling and parked periods. The data carry no event labels (attacks, faults, anomalies), so any labels needed for supervised tasks must be supplied by the user. The release also includes records_per_vehicle.csv, which gives one total record count per vehicle for all 100 vehicles and is ordered by Vehicle_Id, and number_of_vehicles_and_records_per_month.csv, which lists each monthly archive and contained CSV with vehicle counts, record counts, and a total row.