• Lightweight modular framework augments outdoor multi‑cell fingerprinting using existing MDT records. • Spatial synthesis via KDE generates geographically coherent synthetic locations from sparse measurements. • Radio‑feature augmentation uses a KNN block to produce realistic per‑cell fingerprints. • Training‑free, interpretable design supports distributed, on‑premise operator deployments with privacy. • Validation on Italian operator data shows median error reduction up to 30% in sparse or complex regions. • Saturation observed in high‑density areas where extra synthetic samples yield diminishing returns.
Article Summaries:
- A new lightweight framework improves outdoor cellular positioning by augmenting sparse mobile data. The system separates spatial and radio‑feature synthesis: kernel density estimation (KDE) generates realistic synthetic locations, while a k‑nearest‑neighbor (KNN) block creates additional per‑cell radio fingerprints. Designed to be training‑free and interpretable, it can run on‑premise or in distributed operator deployments, preserving privacy. Tested on a real Italian mobile‑network MDT dataset across urban and peri‑urban areas, the KDE‑KNN augmentation lowered median positioning error by up to 30 % in the sparsest or most complex regions, though gains saturate quickly in high‑density zones. The approach offers a low‑complexity path for operators to enhance positioning services using existing data traces.
Sources: