Dynamic Feature-Fused Localization With Smartphones Exploiting 5G NR SSB and Wi-Fi for Indoor Environments

With the rapid development and wide deployment of fifth-generation (5G) and wireless fidelity (Wi-Fi) technologies, indoor positioning has entered a new era. Particularly, the salient properties of 5G new radio (NR), such as the synchronization signal block (SSB) with multiple beams, demonstrate superiority in utilizing multivariate fingerprints for positioning. However, existing research has rarely explored the fusion of 5G NR SSB and Wi-Fi with dynamic features for positioning in complex indoor environments. To address this gap, this study proposes a dynamic features-fused localization (DFF-Loc) framework that leverages the complementary advantages of 5G NR SSB and Wi-Fi to achieve accurate, reliable, robust indoor positioning. DFF-Loc consists of four modules: a dynamic signal filter that uses the extended Kalman filter (KF), prepositioning using a lightweight backpropagation neural network (NN), outlier detection through the local outlier factor (LOF), and dynamic weighting fusion based on improved particle filter (IPF) using signal features. Field experiments are conducted with two of smartphones in three typical indoor scenarios to evaluate the performance of DFF-Loc. Compared with 5G-based and Wi-Fi-based methods, DFF-Loc exhibits an average accuracy improvement of 28.23% and 20.38%, respectively. DFF-Loc outperforms classical machine learning algorithms in all tests, with average accuracies of 1.87, 3.97, and 1.23 m. Dynamic experiments reveal the same. The cumulative error probability diagrams demonstrate DFF-Loc’s excellent convergence ability, with the majority of points falling within acceptable error ranges. This solution can serve as a valuable reference for navigation with consumer-grade smartphones exploiting 5G NR SSB and Wi-Fi.