Published October 20, 2026 | Version 4

LoRaWAN Path Loss Measurements in an Indoor Office Setting including Environmental Factors/Conditions

  • 1. ROR icon University of Siegen

Contributors

  • 1. University of Siegen

Description

This dataset was collected during a LoRaWAN measurement campaign in a multi-room indoor office environment at the University of Siegen, Germany. It contains over 2.66 million time-stamped records from 6 LoRaWAN nodes transmitting once per minute to our Kerlink gateway. Each record includes environmental parameters: temperature, relative humidity, barometric pressure, particulate matter (PM2.5), and carbon dioxide (CO₂); as well as device metadata such as RSSI, SNR, spreading factor (SF), etc. The dataset also includes the effective signal power (ESP) and the noise power (NP) for LoRaWAN propagation analysis. The dataset is designed to support research on indoor wireless propagation, distance estimation (localization), and environment-aware modeling, among other IoT use cases and applications in line with the sixth generation (6G) flagship demands. 

A detailed methodology and initial analysis are presented in our accompanying  IEEE Access paper published as:  Nahshon Mokua Obiri and Kristof Van Laerhoven, "A Comprehensive Data Description for LoRaWAN Path Loss Measurements in an Indoor Office Setting: Effects of Environmental Factors," in IEEE Access, vol. 13, pp. 83148-83170, 2025, doi: 10.1109/ACCESS.2025.3569164.

🔗 Additional Notes

  • File 1 (1.combined_measurements_data.csv) holds the uplinks that our gateway kerlink001 received from 1 October 2024 to 30 September 2025, sorted by node and time, without duplicates, with kerlink001's own RSSI and SNR and the transmit power. The original export from InfluxDB, every uplink from 26 September 2024 to 9 October 2025 with the metadata of all receiving gateways, is file 1 of v2 and v3 (1.unsorted_combined_measurements_data.csv).
  • The preprocessed version: filtered, sorted, and with variable renaming, aligned to the structure described in our journal paper, is included as 2.aggregated_measurements_data.csv.
  • The fully cleaned dataset, where anomalies have been removed, is available as 3.cleaned_dataset_per_device.csv. 

Version history

v1, May 2025. First release, 26 September 2024 to 22 May 2025. File 2 has 1,715,869 records. File 3 has 1,217,313 records, and SF7 to SF10. The IEEE Access article analyses the first six months of the campaign, 26 September 2024 to 1 April 2025.

v2, March 2026. Files 2 and 3 extended to one year, 1 October 2024 to 30 September 2025, with 2,663,641 and 2,079,534 records. File 1 runs to 9 October 2025. Files 2 and 3 start on 1 October 2024, so the first five days of the article's window are only in file 1 and in v1.

v3, October 2026. Correction. In v1 and v2, RSSI and SNR, and with them ESP and NP, were read from the first gateway listed for each uplink. For some uplinks, this was another gateway of The Things Network, not our gateway kerlink001, and its RSSI was about 55 dB too weak, and its SNR was about 19 dB too low. In v2, this affects 10% of the records in file 2 and 2.9% in file 3; in v1, 2.8% of file 3. In v3, both values come from kerlink001's own entry. File 3 contains all spreading factors (SF7-SF12), totaling 2,660,274 records. Anomalies are removed by an Isolation Forest on the five environmental sensors, at 0.1% per node. Pressure in files 2 and 3 is now in hPa. Files 1, 2, and 3 for v1 and v2 hold hPa values minus 655.36 because the nodes send 100 times the pressure as a 16-bit number, which wraps at 65,536. Add 655.36 to convert. File 1 is unchanged from v2. The IEEE Access article used the v1 and v2 pipelines, and the pressure values in Table 7 and Figure 7(d) are hPa minus 655.36.

v4, October 2026. Correction. The nodes leave ADR on and never set a transmit power, so The Things Network set it for every uplink: 14 dBm for 66.9 % of the records, 12 dBm for 32.5 %, and lower for 0.65 %. The exported uplinks do not record it; it is reconstructed by replaying TTN's ADR on the raw export, file 1 of v3 (1.unsorted_combined_measurements_data.csv), with notebook 02 in the GitHub repository, and added to files 2 and 3 as tx_power (dBm, nominal). exp_pl now uses it; in v1–v3 it assumed 14 dBm and is too high by 14 − tx_power (0.68 dB on average). Record counts and all other values are unchanged. File 1 is now the sorted file of the uplinks received by kerlink001; the original export stays file 1 of v2 and v3.

Intermediate processing steps and transformations are documented in the accompanying GitHub repository, where each stage of the pipeline is implemented and version-controlled. Users can generate intermediate versions as needed by following the processing scripts or reviewing the commit history.

Files

1.combined_measurements_data.csv

Files (3.2 GB)

Name Size
md5:39869f0eaeaa3e815111659f751cdb07
2.3 GB Preview Download
md5:153c9c139b648adf8d084bd98aa55af9
459.4 MB Preview Download
md5:638d918ffc3a15bb10564c852df6e6b0
449.5 MB Preview Download

Additional details

Related works

Is described by
Journal article: 10.1109/ACCESS.2025.3569164 (DOI)
Is supplement to
Preprint: arXiv:2505.01185v1 (arXiv)
Conference paper: 10.1109/EuCNC/6GSummit63408.2025.11037092 (DOI)
Preprint: arXiv:2510.04346v1 (arXiv)

Dates

Collected
2024-09-26
Start of collection campaign
Available
2025-05-22
End of collection campaign

Software

References

  • Nahshon Mokua Obiri and Kristof van Laerhoven, "A Comprehensive Data Description for LoRaWAN Path Loss Measurements in an Indoor Office Setting: Effects of Environmental Factors," in IEEE Access, vol. 13, pp. 83148-83170, 2025, doi: 10.1109/ACCESS.2025.3569164.