Skip to main content

UWB Industrial Localization (UWB-Industrial)

Industrial UWB CIR localization dataset with anchor measurements and reference positions. WavesFM task id: uwb-industrial (dataset class UWBIndustrial).

Tasks supported​

  • Localization (Mean Localization Error in meters)

Targets​

  • Continuous (x, y) coordinates (regression); no class labels.

Split​

  • 80/20 random split on industrial_training.pkl (we do not use the official test set as the ground truth is not available).

Preprocessing​

  1. Load the pandas pickle files for train/test (each row is a CIR capture from one anchor).
  2. Group rows by burst_id and stack anchor CIRs into (2, A, L) real/imag tensors for each sample.
  3. Drop samples with fewer than 4 anchors (default) and fill missing anchors with zeros; write anchor_mask.
  4. Store cir, location, optional rec_time/burst_id, plus dataset-wide mean/std and loc_min/loc_max.

Script: preprocess_ipin_loc.py

python preprocessing/preprocess_ipin_loc.py \
  --data-path <IPIN_DIR>/industrial_training.pkl \
  --output <IPIN_DIR>/industrial_training.h5

Metric​

  • Mean Localization Error (meters) on the test split.