Reproduce “LIO-GVM: an Accurate, Tightly-Coupled Lidar-Inertial Odometry with Gaussian Voxel Map” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xingyu Ji, Shenghai Yuan, Pengyu Yin, Lihua Xie
The LIO-GVM framework from Nanyang Technological University introduces a tightly-coupled Lidar-Inertial Odometry system utilizing a Gaussian Voxel Map with similarity-based outlier rejection and a variance-aware residual metric. This approach yields superior odometry accuracy (lower ATE and ARE), reduced map storage consumption, and enhanced mapping efficiency compared to existing LIO methods across various datasets.
Key insights
- Explicitly evaluating the similarity between matched features and rejecting low-similarity correspondences is critical for robustly mitigating false positives and enhancing LIO accuracy.
- Incorporating both spatial separation and probabilistic variance disparity into the residual metric, rather than just distance, leads to more reliable and precise pose estimation in filter-based LIO systems.
- An incremental Gaussian Voxel Map (GVM) implemented as a hash table provides O(1) time complexity for map operations and, with an adaptive weighted-average update scheme, ensures stable and efficient map maintenance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.