SuJin Baek, Journal of Internet Computing and Services, Vol. 26, No. 5, pp. 49-56, Oct. 2025
10.7472/jksii.2025.26.5.49, Full Text:  HTML
Keywords: UWB, NLOS(Non-line-of-sight), Indoor Positioning, XGBoost algorithm, Machine Learning, trilateration

Abstract

Ultra-Wideband (UWB) technology is a promising method for indoor positioning with centimeter-level accuracy; however, in non-line-of-sight (NLOS) environments with obstacles such as walls or humans, signal attenuation and multipath effects can lead to significant ranging errors. This study proposes a machine learning-based ranging correction framework to address these issues. The proposed framework employs an Extreme Gradient Boosting (XGBoost) model trained on distance measurements collected from each anchor. The model classifies NLOS and line-of-sight (LOS) conditions to characterize error patterns and performs precise regression to correct nonlinear NLOS ranging errors. Corrected distances are subsequently used in a trilateration algorithm to estimate the final positions. Experimental results show that the proposed approach reduces raw ranging errors by approximately 98%, with NLOS/LOS classification accuracy reaching around 99.5%. Compared to traditional Kalman filter-based methods, it achieves significantly higher positioning accuracy, demonstrating the superiority of the learning-based approach. The results indicate that the proposed method can enhance the performance of UWB-based indoor positioning systems and offers high reliability and applicability across various use cases.

Statistics
Cite this article