Outdoor-RSSI Collections

Abstract—Collision hazard detection in dynamic outdoor environments such as construction and mining zones remains a significant challenge due to unstable signal conditions, mobilityinduced fluctuations, and frequent non-line-of-sight (NLOS) scenarios. Traditional approaches like camera-based systems suffer performance degradation in dusty or visually obstructedconditions, while high-precision localization methods such as Ultra-Wideband (UWB) and GPS often prove cost-prohibitive and lack scalability. Bluetooth Low Energy (BLE) presents a costeffective and widely available alternative; however, its reliance on the Received Signal Strength Indicator (RSSI) introduces variability due to multipath effects and environmental dynamics, limiting its real-time reliability. This paper introduces AlertBLE, a hybrid BLE-based localization and hazard detection system designed to improve safety alert mechanisms in work zones. AlertBLE integrates an Extended Kalman Filter (EKF) with an Adaptive Moving Average (AMA) algorithm to stabilize RSSI signals dynamically, addressing LOS/NLOS variations. Rather than relying solely on precise localization, AlertBLE reformulate the problem as a hazard zone classification task, employing machine learning classifiers such as K-Nearest Neighbors (KNN) and Support Vector Machines (SVM) to enhance detection robustness under challenging signal conditions. The system incorporates wearable proximity sensors for personnel and portable, vehicle-mounted detection units to enable continuous monitoring and real-time alerts during vehicle movement and reversing. Field evaluations in realistic outdoor settings demonstrate that AlertBLE reliably identifies hazard zones, improves situational awareness, and significantly reduces collision risks in complex work environments.