Reproduce “SOCC-ICP: Semantics-Assisted Odometry based on Occupancy Grids and ICP” 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
Johannes Scherer, Sebastian Hirt, Henri Meeß
SOCC-ICP integrates LiDAR odometry with 3D semantic occupancy grid mapping, leveraging an adaptive Iterative Closest Point (ICP) formulation and semantic information to achieve robust pose estimation. This framework simultaneously constructs a unified map representation directly usable by downstream robotic applications, showing competitive accuracy on the KITTI benchmark and improved robustness in geometrically challenging environments.
Key insights
- A dynamically updated 3D semantic occupancy grid can effectively serve as a unified map representation for high-performance online LiDAR odometry, streamlining system architecture.
- Utilizing the fixed first-inserted point within each voxel as the ICP correspondence anchor is crucial for odometry stability and accuracy, outperforming other centroid-based approaches.
- Adaptive weighting of ICP residuals based on local planarity and incorporating semantic confidence significantly enhances robustness, especially in environments with geometric degeneracies or sparse features.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.