FDN-VINS: Fault Detection Neural Network-Aided Visual–Inertial SLAM
IEEE Sensors Journal · 15.04.2026 · 26 (8) · 12384-12397
DOI: 10.1109/jsen.2026.3672981 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.04.2026. Verfasst von Yilin Zhao und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 40 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Beijing Natural Science Foundation.
Abstract (Originalsprache)
Multisensor fusion is widely adopted for resilient positioning, navigation, and timing (PNT) services. Visual–inertial navigation system (VINS) is a critical solution for achieving pose estimation in global navigation satellite system (GNSS)-denied environments as one of the low-cost fusion approaches. However, vision-based methods depend on accurate depth reconstruction, which is highly susceptible to environmental conditions. When the environment degrades, the system is prone to failure. It is worth noting that it is difficult to maintain highly accurate and robust localization because of the absence of an effective fault detection method in VINS. This article proposes a fault detection scheme for VINS based on a data-driven method. It enables the detection of fault conditions of VINS by using IMU data as a reference while simultaneously inputting the state of VINS. Furthermore, a delayed marginalization strategy is employed to incorporate the temporally sequenced fault status into the optimization thread, allowing the system to circumvent fault conditions. Comparative experiments based on extensive public datasets and self-collected data demonstrate that the proposed fault detection network can effectively identify fault conditions of VINS, significantly enhancing both the accuracy and robustness of VINS. It exhibits universal applicability across different platforms and scenarios. The code has been open-sourced for learning and usage at https://github.com/ColeZhao/FDN-VINS
Förderung
- National Natural Science Foundation of China
- Beijing Natural Science Foundation
Zitieren
APA
Zhao, Y., Fan, Z., & Zhao, L. (2026). FDN-VINS: Fault Detection Neural Network-Aided Visual–Inertial SLAM. IEEE Sensors Journal, 26(8), 12384-12397. https://doi.org/10.1109/jsen.2026.3672981
Vancouver
Zhao Y, Fan Z, Zhao L. FDN-VINS: Fault Detection Neural Network-Aided Visual–Inertial SLAM. IEEE Sensors Journal. 2026;26(8):12384-12397. doi:10.1109/jsen.2026.3672981
BibTeX
@article{Zhao2026,
author = {Zhao, Yilin and Fan, Ziyue and Zhao, Long},
title = {FDN-VINS: Fault Detection Neural Network-Aided Visual–Inertial SLAM},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {8},
pages = {12384--12397},
doi = {10.1109/jsen.2026.3672981}
}
EUVIMED