Indoor Fingerprint Positioning Based on Spatiotemporal Feature Analysis of 5G RSSI in NLoS Scenes
Qiyuan Zhang, Haina Ye, Shan Yang · 2025
To address the accuracy issues of uplink time difference of arrival (UTDoA) methods in non-line-of-sight (NLoS) scenarios and the errors in traditional fingerprint localization based on static received signal strength indicator (RSSI) mean values in dynamic environments, this paper proposes a deep neural network-based fingerprint positioning method using RSSI spatiotemporal feature modeling. By analyzing the spatiotemporal variation trends of the RSSI data, a dynamic signal propagation model under multipath superposition is established, revealing the mechanisms behind trends such as RSSI increases, decreases, and V-shaped reversals. Additionally, a fingerprint-assisted UTDoA method based on RSSI spatiotemporal features is proposed to enhance positioning accuracy. Experimental results show that in NLoS scenarios, the proposed fingerprint localization method achieves 70% accuracy, while the fingerprint-assisted UTDoA method reduces cumulative positioning error to 1.71m, improving accuracy by 36.2% compared to traditional UTDoA.