Abstract
Gyroscope scale factor errors play a critical role in determining the dynamic performance of fiber-optic inertial navigation systems and directly affect measurement accuracy. However, the observability of scale factor errors is highly maneuver-dependent, and unobservable subspaces may arise under insufficient excitation. Conventional approaches typically adopt high-dimensional Kalman filters to enhance estimation stability, which may retain weakly observable scale factor states and adversely affect overall filtering performance. In this study, a novel variable-dimension Kalman filtering algorithm is proposed. A gyroscope confidence factor derived from prior noise characteristics is introduced to assess maneuver sufficiency and autonomously identify unobservable scale factor states in real time. Unlike existing methods that rely on structural modification of system or measurement models, the proposed approach realizes online dimension adaptation solely through covariance initialization and propagation, preserving filter continuity and implementation simplicity. Experimental results demonstrate that the proposed method improves heading accuracy from 0.1° to 0.02° and achieves a horizontal positioning accuracy of 0.89 nmile/h, corresponding to a 55.5% performance enhancement compared with conventional fixed-dimensional filtering approaches.
Funding Information
  • Beijing Postdoctoral Science Foundation (3254042)
  • China Postdoctoral Science Foundation (2024M764081)
  • China Postdoctoral Science Foundation (2025T181121)

This publication has 27 references indexed in Scilit: