EUVIMEDEUVIMED entdecken
Crossref

SR-TCN: an RSSI-based indoor localization algorithm via transformer–CNN fusion

Yong Hu, Xiaomin Yu, Sijia Wang, Mingxuan Li

Measurement Science and Technology · 31.07.2026 · 37 (30) · 306303

DOI: 10.1088/1361-6501/ae8bc4 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 28.07.2026. Verfasst von Yong Hu und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

With the rapid development of digital and information technologies and the widespread adoption of intelligent terminals, indoor localization technology has demonstrated significant value in scenarios such as underground parking navigation, intelligent manufacturing, emergency rescue, smart supermarkets, and museum guidance systems. However, due to temporal random fluctuations, multipath fading, spatial distribution heterogeneity, and the sparsity of received signal strength indicator (RSSI) features in complex indoor wireless environments, existing fingerprint-based localization methods still face substantial challenges in positioning accuracy and robustness. To address these issues, this paper proposes spatially and RSSI-aware Transformer–CNN , a high-precision indoor localization model that fuses convolutional neural network (CNN) and Transformer architectures, fully exploiting the global dependency modeling capability of Transformers and the local spatial feature extraction advantage of CNNs. First, an RSSI-aware cross-row masked attention mechanism is introduced to guide attention allocation based on RSSI masks, thereby enabling effective feature interaction among multiple access points (APs) and enhancing the global feature representation capability. Second, static and dynamic recurrent enhancement modules are designed to strengthen periodic feature representation through an adaptive recurrent replication strategy, thereby improving the model’s robustness to signal fluctuations and sparse samples. In addition, the topological relationships among APs are explicitly injected into token representations, providing the Transformer with geometric structural priors and significantly enhancing its spatial relationship modeling capability. Extensive experiments are conducted on a real-world indoor localization dataset, and the proposed method is compared with representative approaches, including support vector machine, CNN, GCN, and K-nearest neighbor. Experimental results demonstrate that the proposed model achieves significant advantages in both localization accuracy and cross-region generalization performance.

Zitieren

APA

Hu, Y., Yu, X., Wang, S., & Li, M. (2026). SR-TCN: an RSSI-based indoor localization algorithm via transformer–CNN fusion. Measurement Science and Technology, 37(30), 306303. https://doi.org/10.1088/1361-6501/ae8bc4

Vancouver

Hu Y, Yu X, Wang S, Li M. SR-TCN: an RSSI-based indoor localization algorithm via transformer–CNN fusion. Measurement Science and Technology. 2026;37(30):306303. doi:10.1088/1361-6501/ae8bc4

BibTeX
@article{Hu2026,
  author = {Hu, Yong and Yu, Xiaomin and Wang, Sijia and Li, Mingxuan},
  title = {SR-TCN: an RSSI-based indoor localization algorithm via transformer–CNN fusion},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {30},
  pages = {306303},
  doi = {10.1088/1361-6501/ae8bc4}
}
Kostenlos starten →