Degeneracy Sensing Light Detection and Ranging‐Inertial Simultaneous Localization and Mapping with Dual‐Layer Resistant Odometry and Scan‐Context Loop‐Closure Detection Backend in Diverse Environments
Year: 2025
Type: article
Abstract: Light detection and ranging (LiDAR) has become a critical sensor in intelligent systems such as autonomous vehicles and unmanned robots. However, due to limited geometric constraints, the performance of LiDAR simultaneous localization and mapping (SLAM) degrades significantly in degenerate environments, such as tunnels, elongated corridors, and feature‐sparse open spaces. To address this limitation, a novel degeneracy‐aware LiDAR‐inertial framework is proposed that integrates a dual‐layer robust odometry frontend with a Scan‐Context‐based loop‐closure detection backend. Temperature compensation is first applied to enhance vehicle inertial measurements, thereby improving basic recursive accuracy. The frontend module assesses degeneracy status through Fisher information to adaptively adjust feature down‐sampling rates. Moreover, the odometry residual is redefined with feature scaling factors. The backend leverages scan‐context descriptors to detect global loop closures in environments lacking distinctive features robustly. Extensive experimental evaluation on public and private datasets, encompassing challenging scenarios such as linear indoor corridors, large outdoor playgrounds, and complex parking lots, demonstrates that the proposed framework effectively mitigates localization drift caused by degeneracy, resulting in significant improvements in accuracy and robustness.
Source: Advanced Intelligent Systems
Authors: Haoming Yang, Liye Zhao, Xukai Ding, Wenjun Li, Kehao Qi
Institution: Southeast University
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,970
Awards: G2490626666, G5664979477, G6183353344
OpenAlex ID: W7117296377