Reproduce “OKVIS2-X: Open Keyframe-based Visual-Inertial SLAM Configurable with Dense Depth or LiDAR, and GNSS” 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
Simon Boche, Jaehyung Jung, Sebastián Barbas Laina, Stefan Leutenegger
OKVIS2-X is an open-source, configurable visual-inertial SLAM system that tightly integrates dense volumetric mapping from learned depth or LiDAR, and dropout-tolerant GNSS fusion. The system achieves state-of-the-art localization accuracy, often reaching centimeter-level precision over kilometers, and generates usable dense occupancy maps in real-time across diverse outdoor and indoor environments.
Key insights
- Tight coupling between the state estimator and dense volumetric submaps through novel map alignment factors significantly enhances localization accuracy and map consistency.
- Uncertainty-aware fusion of learned depth, alongside LiDAR integration, proves effective in generating accurate and complete dense occupancy maps, improving robustness in challenging visual conditions.
- A dropout-tolerant GNSS fusion strategy, combined with global frame alignment and re-initialization, effectively mitigates drift over long trajectories even with intermittent global positioning data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.