Skip to content
Open access

Adaptive Robust Visual-Inertial Localization Method in GNSS-Denied Environments

Nov 2026 · Micromachines · 0 citations

Abstract

With the rapid development of smart wearable devices and autonomous systems, high-precision and highly robust positioning in GNSS-denied environments—such as indoor spaces, underground areas, and tunnels—has become a critical engineering requirement. Nevertheless, several key challenges remain in practical applications. This paper establishes an adaptive and robust visual-inertial pedestrian navigation algorithm framework. For the inertial sensor, an online error compensation model is constructed, and IMU preintegration is employed to achieve pedestrian motion propagation. For visual observations, a lightweight front-end is designed to reduce computational burden. By introducing the Huber robust kernel function, the quality of visual observation information is evaluated in real time, and anomalous residuals are down-weighted or even rejected, thereby preventing abnormal observation data from degrading the state estimation and significantly enhancing the stability of the system in complex environments. Validation in real-world scenarios demonstrates that the proposed algorithm achieves a positioning error of less than 0.4% in normal environments while maintaining a positioning error within 1.5% in visually degraded scenarios, exhibiting excellent anti-degradation capability and good engineering applicability.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.