Preventing negative transfer in cross-building Wi-Fi fingerprint localization via validation-gated estimator selection

Cross-building transfer learning can reduce Wi-Fi fingerprint calibration effort, but it may degrade positioning when source and target environments have incompatible access-point deployments, propagation conditions, or spatial mappings. This paper proposes SafeTransLoc, a framework that treats transfer as an optional deployment decision: target-only and transfer candidates are compared on held-out target data, and transfer is accepted only when its validation advantage exceeds a safety margin. The principal UJIIndoorLoc and SODIndoorLoc evaluation comprises 120 cross-building configurations; under the deployment-level criterion, the best transfer candidate was worse than the strongest target-only alternative in all of them. A post-review controlled same-environment SODIndoorLoc experiment adds 270 RSS-offset/noise configurations and yields 6 beneficial-transfer and 264 harmful-transfer cases. With a zero margin, the gate accepted 1 of the 6 beneficial cases and rejected 89.0% of harmful transfers; increasing the margin improved harmful-transfer rejection but reduced beneficial-transfer sensitivity. Independent leave-one-environment-out calibration on the archived SOD results selected 0 m for CETC331 and 0.5 m for HCXY and SYL, producing one held-out false acceptance overall. These results support SafeTransLoc as a deployment safeguard while showing that conservative validation gating involves a measurable safety-sensitivity trade-off under sparse target supervision.

Authors

Institutions

Publication Details

Journal
Discover Electronics
Published
2026-11-15
DOI
https://doi.org/10.1007/s44291-026-00318-4
Primary Topic
Indoor and Outdoor Localization Technologies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Preventing negative transfer in cross-building Wi-Fi fingerprint localization via validation-gated estimator selection

Ahmad Hakimi Ahmad Sa’ahiry, N. T. T. Tran, Hongnian Yu, Chan Hwang See et al.
Discover Electronics
Indoor and Outdoor Localization Technologies
article

Preventing negative transfer in cross-building Wi-Fi fingerprint localization via validation-gated estimator selection

Ahmad Hakimi Ahmad Sa’ahiry, N. T. T. Tran, Hongnian Yu, Chan Hwang See, Qian Wang
article en

Abstract

Cross-building transfer learning can reduce Wi-Fi fingerprint calibration effort, but it may degrade positioning when source and target environments have incompatible access-point deployments, propagation conditions, or spatial mappings. This paper proposes SafeTransLoc, a framework that treats transfer as an optional deployment decision: target-only and transfer candidates are compared on held-out target data, and transfer is accepted only when its validation advantage exceeds a safety margin. The principal UJIIndoorLoc and SODIndoorLoc evaluation comprises 120 cross-building configurations; under the deployment-level criterion, the best transfer candidate was worse than the strongest target-only alternative in all of them. A post-review controlled same-environment SODIndoorLoc experiment adds 270 RSS-offset/noise configurations and yields 6 beneficial-transfer and 264 harmful-transfer cases. With a zero margin, the gate accepted 1 of the 6 beneficial cases and rejected 89.0% of harmful transfers; increasing the margin improved harmful-transfer rejection but reduced beneficial-transfer sensitivity. Independent leave-one-environment-out calibration on the archived SOD results selected 0 m for CETC331 and 0.5 m for HCXY and SYL, producing one held-out false acceptance overall. These results support SafeTransLoc as a deployment safeguard while showing that conservative validation gating involves a measurable safety-sensitivity trade-off under sparse target supervision.

Discover ElectronicsVol. 3(1)
Can Tho University (VN), Edinburgh Napier University (GB), Yanshan University (CN)
Indoor and Outdoor Localization Technologies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon