A UWB Positioning Method Based on Clustering for NLOS Identification and Mitigation

Ultra-wideband (UWB) is a wireless technology that uses ultra-short pulse and multiple anchors for high-accuracy, interference-resistant positioning. However, most UWB positioning algorithms maintain high accuracy only in line-of-sight (LOS) environments, and their performance significantly degrades in non-line-of-sight (NLOS) conditions. To effectively mitigate NLOS interference, this article proposes a cluster-based identification adaptive Kalman filter (CBI-AKF). In each UWB ranging cycle, the distance measurements from multiple anchor nodes are grouped in sets of three, and a rough position estimate is calculated for each group using trilateration. The rough position estimates are then regrouped according to anchor node identifiers, each grouped dataset then undergoes Kernel density estimation (KDE) clustering to obtain cluster centroids representing ranging characteristics of the corresponding anchor nodes. All centroids are subsequently clustered using density-based spatial clustering of applications with noise (DBSCAN) to differentiate cluster, identifying node in NLOS conditions and discarding their relevant data. The distance between the two clustering results is used to identify NLOS ranging values, which are then discarded. A dynamic observation matrix Kalman filter is applied to link successive ranging cycles, the distance between the clustering results and their center is used to construct a dynamic errors matrix, with geometric dilution of precision (GDOP) introduced to adjust the errors, ultimately enhancing positional accuracy. The final NLOS experiment shows that CBI-AKF achieves a relatively high level of precision, demonstrating that the proposed algorithm can effectively mitigate the impact of NLOS effects on positioning accuracy and achieve superior positioning precision.