@article{ART003370769},
author={Am-Suk Oh},
title={AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2026},
volume={31},
number={8},
pages={209-217}
TY - JOUR
AU - Am-Suk Oh
TI - AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System
JO - Journal of The Korea Society of Computer and Information
PY - 2026
VL - 31
IS - 8
PB - The Korean Society Of Computer And Information
SP - 209
EP - 217
SN - 1598-849X
AB - Children left unattended inside school commuter vehicles continue to cause fatal heat-related and cold-related accidents every year, yet most of these vehicles are individually owned by small academies or kindergartens and are rarely retrofitted with dedicated safety equipment. This study instead implements a boarding and alighting notification system that relies only on a Bluetooth Low Energy (BLE) beacon and a smartphone, augmented with a lightweight on-device classifier that adaptively re-estimates the boarding/alighting decision boundary from short-window RSSI features (mean, slope, and variance) rather than a single fixed threshold. Each child carries a small BLE beacon; the accompanying teacher's smartphone extracts these features in real time and feeds them to a logistic-regression classifier trained offline on calibration data collected from three representative Korean commuter vehicles — an 11-seat van, a 15-seat van, and a 25-seat minibus — at both the driver's seat and the front passenger's seat, and the resulting boarding/alighting event is pushed to the parent's smartphone in real time through Firebase Cloud Messaging (FCM). Compared with the fixed-threshold baseline, the adaptive classifier raises the area under the ROC curve from 0.87 to 0.96 and improves notification accuracy in the 4–5 m and 5–6 m distance bins from 88% and 76% to 94% and 87%, respectively, while matching baseline accuracy within 4 m.
KW - Bluetooth Beacon;RSSI;Adaptive Threshold;Machine Learning;Boarding and Alighting Notification
DO -
UR -
ER -
Am-Suk Oh. (2026). AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System. Journal of The Korea Society of Computer and Information, 31(8), 209-217.
Am-Suk Oh. 2026, "AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System", Journal of The Korea Society of Computer and Information, vol.31, no.8 pp.209-217.
Am-Suk Oh "AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System" Journal of The Korea Society of Computer and Information 31.8 pp.209-217 (2026) : 209.
Am-Suk Oh. AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System. 2026; 31(8), 209-217.
Am-Suk Oh. "AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System" Journal of The Korea Society of Computer and Information 31, no.8 (2026) : 209-217.
Am-Suk Oh. AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System. Journal of The Korea Society of Computer and Information, 31(8), 209-217.
Am-Suk Oh. AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System. Journal of The Korea Society of Computer and Information. 2026; 31(8) 209-217.
Am-Suk Oh. AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System. 2026; 31(8), 209-217.
Am-Suk Oh. "AI-Based Adaptive Threshold Estimation for a Beacon RSSI Boarding and Alighting Notification System" Journal of The Korea Society of Computer and Information 31, no.8 (2026) : 209-217.
