HERO-LO: Heterogeneous-Odometry-Enhanced Robust LiDAR Odometry with Observable-Subspace Correction for Tunnel-like Environments
Applied Sciences · 23.10.2026 · 16 (21) · 10509
DOI: 10.3390/app162110509 ↗Kurz gefasst
Fachartikel in der Zeitschrift Applied Sciences, online erschienen am 23.10.2026. Verfasst von Dongjie Wu und 2 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Fujian Provincial Science and Technology Program Guiding Project, Jimei University Scientific Research Startup Fund Project.
Abstract (Originalsprache)
Tunnel-like environments remain challenging for Light Detection and Ranging (LiDAR) odometry because repetitive geometry and weak axial constraints can lead to degeneracy, accumulated drift, and unstable local pose estimates. This paper presents HERO-LO, a heterogeneous-odometry-enhanced robust LiDAR odometry framework with observable-subspace correction for tunnel-like environments. The proposed method detects LiDAR degeneracy online in the tangent space of SO(3)×R3 by interpreting the Hessian eigenstructure from a state-estimation uncertainty perspective. When the LiDAR geometry is well constrained, HERO-LO uses scan-to-scan LiDAR odometry as the motion prior and applies LiDAR matching corrections normally. When tunnel-like degeneracy is detected, a heterogeneous odometry source, such as kinematic odometry, attitude-and-heading-reference-system (AHRS)-aided kinematic odometry, or visual-inertial odometry, provides the motion prior, while LiDAR corrections are applied only in the observable subspace and the prior is preserved in degenerate directions. The correction law is further interpreted as a projection-gain observer, and a Linear Matrix Inequality (LMI)-based Lyapunov analysis establishes its per-direction bounded error propagation and stability under mode switching. Experiments on a real tunnel dataset show that, within the degenerate tunnel segment, the default HERO-LO configuration with an encoder-based kinematic odometry (KO) prior reduces the local anomaly rate to 0.6% and achieves an endpoint horizontal error of approximately 3.6 m, comparable to the best open-source LiDAR odometry (LO), LiDAR-inertial odometry (LIO), and LiDAR-visual-inertial odometry (LVIO) methods given the approximately 2.5 m circular error probable (CEP) uncertainty of the single-point Global Navigation Satellite System (GNSS) reference. Runtime analysis shows that HERO-LO processes each LiDAR frame in approximately 9.5 ms on average, satisfying the real-time requirement of the 20 Hz LiDAR. Ablation results further confirm that robust tunnel traversal requires both a stable heterogeneous prior and observable-subspace correction.
Förderung
- Fujian Provincial Science and Technology Program Guiding Project
- Jimei University Scientific Research Startup Fund Project
Zitieren
APA
Wu, D., Liang, M., & Chen, X. (2026). HERO-LO: Heterogeneous-Odometry-Enhanced Robust LiDAR Odometry with Observable-Subspace Correction for Tunnel-like Environments. Applied Sciences, 16(21), 10509. https://doi.org/10.3390/app162110509
Vancouver
Wu D, Liang M, Chen X. HERO-LO: Heterogeneous-Odometry-Enhanced Robust LiDAR Odometry with Observable-Subspace Correction for Tunnel-like Environments. Applied Sciences. 2026;16(21):10509. doi:10.3390/app162110509
BibTeX
@article{Wu2026,
author = {Wu, Dongjie and Liang, Mingyan and Chen, Xianyi},
title = {HERO-LO: Heterogeneous-Odometry-Enhanced Robust LiDAR Odometry with Observable-Subspace Correction for Tunnel-like Environments},
journal = {Applied Sciences},
year = {2026},
volume = {16},
number = {21},
pages = {10509},
doi = {10.3390/app162110509}
}
EUVIMED