LoRaIQ: Experimental LoRa Dataset with Annotated IQ Samples

Summary

This dataset contains more than 30,000 LoRa frames recorded within and around the EPFL campus in Lausanne, Switzerland. Each frame was captured by an average of 3.75 receivers using four remote radio heads (RRHs) installed on campus rooftops.

Each dataset entry provides the information required for synchronization, demodulation, and decoding of a single LoRa frame. Every frame is accompanied by an IQ sample file following the standard SigMF specification [1], together with its corresponding metadata and annotations.

The dataset includes both UAV-based and pedestrian measurements. For the UAV-based measurements, the start-of-frame location in the IQ samples is provided with sub-sample accuracy. For the pedestrian measurements, the provided frame start locations should be interpreted as indicative estimates.

Dataset Details

Recordings were captured by four RRHs deployed at the locations shown in Fig. 2 and indicated in rrh_locations.md. Each RRH consists of a USRP-2920, a Raspberry Pi 5, and a Quectel L80 GPS receiver. This deployment captures a wide range of propagation scenarios across the EPFL campus, including variations in receiver elevation, line-of-sight conditions, and urban obstructions.

The dataset includes transmissions from both an unmanned aerial vehicle (UAV) and a pedestrian-carried transmitter, with transmitter velocities ranging from 0 to 5.6 m/s. Measurements are labelled according to the different area types illustrated in Fig. 2.

• Line-of-sight (LoS): UAV Areas 1 and 2
• Non-line-of-sight (NLoS): UAV Areas 3 and 4; Pedestrian Areas 2 and 3
• Partial LoS: Pedestrian Area 1, Roaming
• Indoor: Pedestrian Area 4

The transmitter parameters used during data collection are summarized below:

Note: Boldface values indicate the parameters used for the UAV-mounted transmitter.

In addition to the synchronization parameters required for demodulation, such as carrier frequency offset, reception timestamp, and LoRa PHY parameters, each frame also includes the transmit location, transmit payload, and an SNR estimate. 

Dataset Entry 

All the keys of the dataset.csv is indicated below and Fig. 2 illustrates the spectrogram of one received  frame opened with Inspectrum [2], showing the LoRa chirps with annotations.

File Structure

• dataset.csv                                        (Dataset with all frame descriptions)
• load_and_plot_samples.py               (Exemple to load IQ samples)         
• rrh_locations.md                               (GPS positions of the four remote radio heads)
• zenodo_description_fig.png             (An illustration shown on Zenodo description page)
• sigmfs                                                (Folder containing the raw IQ samples)
  
  • <Date and time of measurement 1>
    • RRH1
      • 1.sigmf-data
      • 1.sigmf-meta
      • 2.sigmf-data
      • 2.sigmf-meta
      • ...
    • ...
    • RRH4
      • 1.sigmf-data
      • 1.sigmf-meta
      • 2.sigmf-data
      • ...
  • <Date and time of measurement 2>
    • ....

Contact

For any questions related to this dataset, please contact joachim.tapparel@epfl.ch.

References

[1] "SigMF specification,'' Accessed: May 19, 2026. [Online]. Available: https://sigmf.org/

[2] "Inspectrum", Accessed: May 19, 2026. [Online]. Available: https://github.com/miek/inspectrum