UWB-BASED GEOFENCING: A ONE-CLASS CLASSIFICATION APPROACH WITH FINGERPRINTING AND TRILATERATION

Ultra-wideband (UWB) technology has emerged as a powerful solution for indoor localization, offering high accuracy, low power consumption, and robust penetration through obstacles. Among its applications, geofencing enables the creation of virtual boundaries for monitoring and security purposes. This paper presents a novel one-class classification (OCC) approach for UWB-based geofencing, named the k-Nearest Neighbours with Residual Norm Threshold (kNN-RNT) algorithm. The proposed method utilizes fingerprinting and trilateration techniques, operating in two distinct phases: an offline phase for constructing a reference fingerprint database and an online phase for real-time classification of a mobile tag’s location. The kNN-RNT algorithm determines geofence violations by analysing the distribution of nearest fingerprints and computing a residual norm to classify locations. A filtering mechanism enhances detection stability, mitigating noise and transient errors. Experimental validation in a controlled indoor environment demonstrates the effectiveness of the method, achieving over 99% accuracy within the geofenced area and significantly reducing classification errors in proximity zones. The proposed approach provides a reliable and efficient solution for real-time UWB-based geofencing applications.