LODESTAR: Degeneracy-Aware LiDAR-Inertial Odometry with Adaptive Schmidt-Kalman Filter and Data Exploitation

TL;DR

LODESTAR tackles degeneracy in LiDAR-inertial odometry by jointly addressing measurement sparsity, imbalance, and constrained optimization. It introduces DA-ASKF, a degeneracy-aware Schmidt-Kalman filter within a sliding window, and DA-DE, a degeneracy-aware data exploitation strategy that prunes and augments measurements based on localizability and Jacobian conditioning. Experiments across diverse degenerate environments show improved accuracy and robustness relative to state-of-the-art LO and LIO methods, with real-time performance. The method provides a practical solution for robust LIO in corridors, caverns, and high-altitude flights, with open-source release for reproducibility.

Abstract

LiDAR-inertial odometry (LIO) has been widely used in robotics due to its high accuracy. However, its performance degrades in degenerate environments, such as long corridors and high-altitude flights, where LiDAR measurements are imbalanced or sparse, leading to ill-posed state estimation. In this letter, we present LODESTAR, a novel LIO method that addresses these degeneracies through two key modules: degeneracy-aware adaptive Schmidt-Kalman filter (DA-ASKF) and degeneracy-aware data exploitation (DA-DE). DA-ASKF employs a sliding window to utilize past states and measurements as additional constraints. Specifically, it introduces degeneracy-aware sliding modes that adaptively classify states as active or fixed based on their degeneracy level. Using Schmidt-Kalman update, it partially optimizes active states while preserving fixed states. These fixed states influence the update of active states via their covariances, serving as reference anchors--akin to a lodestar. Additionally, DA-DE prunes less-informative measurements from active states and selectively exploits measurements from fixed states, based on their localizability contribution and the condition number of the Jacobian matrix. Consequently, DA-ASKF enables degeneracy-aware constrained optimization and mitigates measurement sparsity, while DA-DE addresses measurement imbalance. Experimental results show that LODESTAR outperforms existing LiDAR-based odometry methods and degeneracy-aware modules in terms of accuracy and robustness under various degenerate conditions.

Figures (5)

• Figure 1: Instances of LODESTAR under various degenerate conditions. In degenerate environments, LiDAR measurements are often sparse or imbalanced. LODESTAR mitigates these problems using DA-ASKF and DA-DE, which selectively utilize past states and measurements. States are classified into active and fixed based on their degeneracy level. Measurements from current and active states are pruned based on their localizability contribution. Then, measurements from fixed states in degenerate directions are exploited to resolve measurement imbalance.
• Figure 2: Overview and core components of the proposed LODESTAR. (a) Flowchart of LODESTAR. (b) Degeneracy-aware sliding modes in DA-ASKF (degeneracy-aware adaptive Schmidt-Kalman filter). (c) Selective data pruning and compensation in DA-DE (degeneracy-aware data exploitation).
• Figure 3: Condition number comparison of different degeneracy-aware modules. Baseline + GenZ-ICP diverged in [1]SubT-MRS--[1]Long Corridor, and is thus excluded. The **** and ** denote that p-values of the paired t-test are less than $10^{-4}$ and $10^{-2}$, respectively, indicating statistically significant differences.
• Figure 4: Mapping results of LODESTAR and the top three state-of-the-art methods on (a) [1]SubT-MRS--[1]Long Corridor and (b) [1]NTU-VIRAL--[1]SPMS 02 sequences. Despite imbalanced or sparse measurements, LODESTAR consistently mapped the environment with minimal drift, whereas others exhibited noticeable drift.
• Figure 5: Average computation time of LODESTAR and the baseline xu2022tro-fastlio2. Overall, the computation time remains comparable to the baseline.