osmag-wifi-localization

Official implementation of "WiFi-based Global Localization in Large-Scale Environments Leveraging Structural Priors from osmAG"

This project presents a novel WiFi-based localization framework for autonomous robotics in large-scale indoor environments where GPS is unavailable. By leveraging ubiquitous wireless infrastructure and OpenStreetMap Area Graph (osmAG) structural priors.

Environment

• Python: 3.10
• Ubuntu 22.04

Quick Start

1) AP Localization

python ap_localization.py --config ./config/ap_localization_config.yaml --log-level INFO

• Input (defaults in config):
  • file_paths.input_osm_file: ./map/wifi_data.osm
  • file_paths.template_osm_file: ./map/base_map.osm
• Output:
  • file_paths.output_osm_file: ./map/AP_MAP.osm (estimated AP positions are written)
  • Console statistics with final/initial errors and improvement

2) Robot fingerprint KNN Localization

python robot_fingerprint_localization.py \
  --fingerprint ./map/wifi_data.osm \
  --test ./map/Non-FingerprintedAreas.osm \
  --polygon ./map/wifi_data.osm \
  -k 5 --log-level INFO

• Add --no-boundary to disable polygon boundary constraint
• Prints localization metrics to console

3) Robot access point Localization

python robot_AP_localization.py \
  --ap-map ./map/AP_MAP.osm \
  --fingerprint ./map/Non-FingerprintedAreas.osm \
  --polygon ./map/wifi_data.osm \
  --iter 10 --log-level INFO

• Optional model parameters (with defaults): --rssi0, --n, --wall
• Outputs analyzed results to console

Project Structure

• ap_localization.py: AP localization entry
• robot_fingerprint_localization.py: KNN-based localization entry
• robot_AP_localization.py: Iterative robot localization entry
• core/: data loading, preprocessing, signal model optimization, position estimation, result saving, fingerprint algorithms
• algorithms/: lower-level point estimator and RSSI optimizer
• io_layer/: OSM parser and writer
• utils/: configuration, building constraints, geometry, filtering, wall learning
• map/, config/: sample paths referenced by defaults

Citation

If you use this work in your research, please cite:

@misc{ma2025wifibasedgloballocalizationlargescale,
      title={WiFi-based Global Localization in Large-Scale Environments Leveraging Structural Priors from osmAG}, 
      author={Xu Ma and Jiajie Zhang and Fujing Xie and Sören Schwertfeger},
      year={2025},
      eprint={2508.10144},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2508.10144}, 
}