Informed, Constrained, Aligned: A Field Analysis on Degeneracy-Aware Point Cloud Registration in the Wild
Turcan Tuna, Julian Nubert, Patrick Pfreundschuh, Cesar Cadena, Shehryar Khattak, Marco Hutter
AI summary
Problem
ICP-based LiDAR registration degrades in geometrically ill-conditioned environments due to insufficient geometric constraints, yet current solutions largely bypass the optimization step by relying on external odometry rather than mitigating degeneracy within the ICP formulation itself.
Approach
The authors conduct a large-scale field and simulation study comparing active and passive degeneracy mitigation strategies, introducing and evaluating truncated singular value decomposition, inequality constraints, and Tikhonov regularization directly within the ICP least-squares minimization step.
Key results
- Field and simulation comparison of active vs. passive degeneracy mitigation
- First evaluation of TSVD, inequality constraints, and Tikhonov regularization for degenerate registration
- Sensitivity analysis of ICP least-squares minimization under different constraints
- Open-source framework consolidating degeneracy-aware LiDAR-SLAM methods
Why it matters
Provides field-validated guidelines for robust LiDAR localization in challenging environments, directly benefiting autonomous robots operating in tunnels, open fields, and unstructured construction sites.
Abstract
The iterative closest point (ICP) registration algorithm has been a preferred method for light detection and ranging (LiDAR)-based robot localization for nearly a decade. However, even in modern simultaneous localization and mapping (SLAM) solutions, ICP can degrade and become unreliable in geometrically ill-conditioned environments. Current solutions primarily focus on utilizing additional sources of information, such as external odometry, to either replace the degenerate directions of the optimization solution or add additional constraints in a sensor-fusion setup afterward. In response, this work investigates and compares new and existing degeneracy mitigation methods for robust LiDAR-based localization and analyzes the efficacy of these approaches in degenerate environments for the first time in the literature at this scale. Specifically, this work investigates i) the effect of using active or passive degeneracy mitigation methods for the problem of ill-conditioned ICP in LiDAR degenerate environments and ii) the evaluation of truncated singular value decomposition (TSVD), inequality constraints (Ineq. Con.), and linear/nonlinear Tikhonov regularization for the application of degenerate point cloud registration for the first time. Furthermore, a sensitivity analysis for the least-squares minimization step of the ICP problem is carried out to better understand how each method affects the optimization and what to expect from each method. The results of the analysis are validated through multiple real-world robotic field and simulated experiments. The analysis demonstrates that active optimization degeneracy mitigation is necessary and advantageous in the absence of reliable external estimate assistance for LiDAR-SLAM, and soft-constrained methods can provide better results in complex ill-conditioned scenarios with heuristic fine-tuned parameters. The code and data used in this work are made publicly available to the community.