Answers are grounded in this paper and cite their sources.
Algorithmic validation reveals a 60.9% reduction in attitude error for autonomous vehicles using hierarchical split covariance filtering, suggesting improved multi-sensor navigation robustness.
Vehicle attitude determination is essential for autonomous driving and intelligent transportation. Gyroscopes provide continuity, accelerometers constrain pitch and roll to gravity, and multi-antenna Global Navigation Satellite System (GNSS) measurements provide an absolute reference. Because the processing branches share platform dynamics, state propagation and cross-sensor aiding information, the errors in their attitude estimates are not independent. Treating them as independent can double-count shared information and produce overconfident covariance estimates. We present a correlation-aware hierarchical architecture for multi-antenna GNSS/inertial navigation system (INS) attitude determination. Two local filters and a master filter integrate gyroscope-, accelerometer- and GNSS-derived attitudes using Split Covariance Intersection Filter (Split CIF), which bounds correlated information while preserving independent measurement information. Analysis showed weak raw inter-source correlations in pitch and roll (|r|≤ 0.19) but strong correlation between gyroscope- and antenna-derived heading errors (r = 0.975). Correlations between the local-filter outputs reached 0.805, 0.361 and 0.989 for pitch, roll and heading, respectively. On a suburban–urban route, with all methods evaluated on identical inputs against the post-processed SPAN-FSAS reference, the proposed architecture achieved pitch, roll and heading root-mean-square errors (RMSEs) of 0.192°, 0.447° and 0.105°, reducing the mean attitude RMSE by 60.9% relative to a centralized extended Kalman filter.
No takes yet. Share an insight, caveat, or question.
Jing et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: