The measurement environment consists of an office-like room measuring 11 m × 8 m, containing typical office equipment such as whiteboards and tables. The reflective walls create significant multipath propagation. The environment features a defined walking zone in which targets move at a slow pace of approximately 1 m/s on random paths.
The system operates at a center frequency of 4.5 GHz with a bandwidth of 499.99 MHz, using 6 DW1000-based UWB nodes forming a mesh network. Four nodes act as transceivers and two nodes act as receivers, forming a multi-static radar setup with 14 total links. A TDMA protocol assigns non-overlapping time slots for transmissions. CIRs are recorded for each link at 40 Hz, limited to 90 samples at 1 ns resolution, covering approximately 30 m of multipath propagation. Ground truth positions are provided by an optical reference system at sub-centimeter accuracy.
This dataset contains CIR measurements from all 14 links alongside IMU measurements and ground truth position references. Data was recorded from 3 different people to facilitate generalization studies.
IMU data is stored in csv files, while CIR and reference data is provided in a HDF5 file, that contains a CIR and a POS dataset with the following fields
CIR |
|||
|---|---|---|---|
| FIELD | SHAPE | TYPE |
DESCRIPTION |
| CIR_R | (90) | int32 | Real component of the CIR |
| CIR_I | (90) | int32 | Imaginary component of the CIR |
| Time stamp | () | uint64 | Timestamp of the recording |
| fpi | () | float64 | First path index, estamated by the DW1000 |
| B_ID | () | int32 | ID of the burst |
| REC_ID | () | uint8 | ID of the receiving node |
| TRANS_ID | () | uint8 | ID of the transmitting node |
POS |
|||
|---|---|---|---|
| FIELD | SHAPE | TYPE |
DESCRIPTION |
| POS X | () | float64 | X-Position |
| POS Y | () | float64 | Y-Position |
| POS Z | () | float64 | Z-Position |
| ROT X | () | float64 | X-Rotation |
| ROT Y | () | float64 | Y-Rotation |
| ROT Z | () | float64 | Z-Rotation |
| Time stamp | () | uint64 | Timestamp of the reference measurement |
Scripts for pretraining and training a passive channel chart can be found here <link>. The data can be downloaded here:
| Split | Person | duration |
Samples | used in paper |
|---|---|---|---|---|
| take1 | P1 | 40 min | ~70,000 | Train / Valid |
| take2 | P2 | 20 min | ~35,000 | Test P1 |
| take3 | P3 | 10 min | ~17,000 | Test P1 |
This work is licensed under a CC BY-SA 4.0 license, see
If you use our dataset for your research, please consider citing:
@inproceedings{poeggel2025passive,
title = {Passive, Channel Charting: Locating Passive Targets using a UWB Mesh},
author = {Poeggel, Raffael and Stahlke, Maximilian and Pirkl, Jonas and Ott, Jonathan and Yammine, George and Feigl, Tobias and Mutschler, Christopher},
booktitle = {2025 International Conference on indoor Positioning and Indoor Navigation (IPIN)},
pages= {1- -6},
year= {2025}
}
This work is licensed under a CC BY-SA 4.0 license, see
Fraunhofer Institute for Integrated Circuits IIS