Online Geometric Consistency Assessment of LiDAR Measurements via Recursive Bayesian Comparison
Online Geometric Consistency Assessment of LiDAR Measurements via Recursive Bayesian Comparison
This paper presents a recursive Bayesian framework for online assessment of LiDAR measurement consistency using accumulated geometric history. By coupling a Normal-Inverse-Wishart model with a weak isotropic reference, the method computes Normalized Drift Evidence (NDE) to distinguish corrupted from consistent sequences without an independent external reference. Evaluated across 14 real-world sequences totaling approximately 73 minutes from four platform types and ten UAV trials, the approach correctly identified all nine corrupted sequences while remaining silent on five consistent ones, with alarms occurring within 0.9 seconds of event references in representative scenarios. The system processes updates at 1 Hz with a mean latency of 4.2–6.1 ms per update and enables autonomous responses such as odometry isolation, landing, relocalization, and mission continuation upon detection.
The method uses recursive Bayesian comparison with a Normal-Inverse-Wishart model to detect LiDAR registration errors in real-world sequences, achieving detection within 0.9 s of reference events across four platform types while maintaining 4.2–6.1 ms processing time per update.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
elapsed_minutes
≈73 minutes
Reported trials
View original evidence
omparing the marginal predictive support of the same current points under both models yields Normalized Drift Evidence (NDE). Assessment precedes assimilation: positive NDE triggers an alarm and preserves the preceding historical state; otherwise, the current observations update the history. Across 14 real-world sequences totaling approximately 73 min from four platform types, alarms identified all nine corrupted seq
Open source S4processing_time_ms
4.2
View original evidence
ic support. Mean processing time is 4.2--6.1 ms per update at 1 Hz. In ten UAV trials, alarms initiated odometry isolation, landing, relocalization, and mission continuation.
Open source S6processing_time_ms
6.1
View original evidence
ic support. Mean processing time is 4.2--6.1 ms per update at 1 Hz. In ten UAV trials, alarms initiated odometry isolation, landing, relocalization, and mission continuation.
Open source S6elapsed_seconds
0.9
Reported trials
View original evidence
uences and were absent from all five consistent sequences. In four representative scenarios with event references, first alarms occurred within 0.9 s of the reference times, with no preceding alarms. Cross-backend evaluation demonstrates consistent responses to reconstructed geometry. Ablation and controlled perturbation results establish the roles of uncertainty modeling, recursive history, displacement, and geometr
Open source S5sequences
14
Reported trials
View original evidence
omparing the marginal predictive support of the same current points under both models yields Normalized Drift Evidence (NDE). Assessment precedes assimilation: positive NDE triggers an alarm and preserves the preceding historical state; otherwise, the current observations update the history. Across 14 real-world sequences totaling approximately 73 min from four platform types, alarms identified all nine corrupted seq
Open source S4corrupted_sequences
9
Reported trials
View original evidence
omparing the marginal predictive support of the same current points under both models yields Normalized Drift Evidence (NDE). Assessment precedes assimilation: positive NDE triggers an alarm and preserves the preceding historical state; otherwise, the current observations update the history. Across 14 real-world sequences totaling approximately 73 min from four platform types, alarms identified all nine corrupted seq
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
omparing the marginal predictive support of the same current points under both models yields Normalized Drift Evidence (NDE). Assessment precedes assimilation: positive NDE triggers an alarm and preserves the preceding historical state; otherwise, the current observations update the history. Across 14 real-world sequences totaling approximately 73 min from four platform types, alarms identified all nine corrupted seq
Open source S4
ic support. Mean processing time is 4.2--6.1 ms per update at 1 Hz. In ten UAV trials, alarms initiated odometry isolation, landing, relocalization, and mission continuation.
Open source S6
uences and were absent from all five consistent sequences. In four representative scenarios with event references, first alarms occurred within 0.9 s of the reference times, with no preceding alarms. Cross-backend evaluation demonstrates consistent responses to reconstructed geometry. Ablation and controlled perturbation results establish the roles of uncertainty modeling, recursive history, displacement, and geometr
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org