AlertBLE: Alert Workzone Hazards Using Hybrid Filtering and Machine Learning Enabled BLE

Collision hazard detection in industrial work zones faces challenges from signal instability, mobility-induced fluctuations, and non-line-of-sight (NLOS) conditions. While Bluetooth Low Energy (BLE) offers cost-effective proximity sensing, its RSSI variability—fluctuating by ±10dBm even at fixed distances— limits reliability in safety-critical applications. This paper presents AlertBLE, a hybrid BLE-based hazard detection system that combines Extended Kalman Filter (EKF) and Adaptive Moving Average (AMA) algorithms to achieve up to 94% RSSI variance reduction in static NLOS conditions. The system introduces speed-aware safety thresholds based on reaction time and braking distance models, dynamically expanding hazard zones from 5m (static) to 8.19m (at 10 km/h), ensuring adequate safety margins across operational speeds. AlertBLE employs K-means clustering to identify LOS/NLOS propagation environments, integrating this context as features for supervised learning. Among four evaluated classifiers (KNN, SVM, Random Forest, XGBoost), KNN demonstrates optimal performance with an efficiency score of 12.8, balancing 82.74% recall with minimal computational requirements (156KB memory, 1.01 ms inference). Field evaluation using 44,774 samples across diverse outdoor conditions demonstrates 88.63% overall detection accuracy with 63 ms system latency—well below the 250 ms safety threshold. The multi-layered error mitigation framework, incorporating temporal smoothing, confidence thresholding, and state machine logic, achieves 75.6% reduction in combined false positives and negatives. Despite 170.8% average RSSI degradation under severe NLOS conditions, AlertBLE maintains 82% detection accuracy within the critical 5-meter zone. The paper also presents a comprehensive security framework addressing BLE vulnerabilities, providing a roadmap for production deployment enhancements.