Reproduce “CELLmap: Enhancing LiDAR SLAM through Elastic and Lightweight Spherical Map Representation” 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

Yifan Duan, Xinran Zhang, Yao Li, Guoliang You, Xiaomeng Chu, Jianmin Ji, Yanyong Zhang

CELLmap stores local geometric maps as plane summaries indexed by fixed spherical directions, pairing each summary with a pose that can change during backend registration and loop closure. The system uses point-to-plane constraints and pose-graph optimization without changing plane parameters. On KITTI, F-LOAM’s mean ATE fell from 0.86% to 0.63%; storage was 0.7% of raw points, with results limited to KITTI.

Key insights

  • Separating local geometry from CELL pose makes the map elastic: loop-closure corrections can reposition local map units without reconstructing their stored surfaces.
  • Directional spherical indexing enables compact map storage and avoids maintaining a separate three-dimensional search structure for every local region.
  • Bidirectional constraints provide a backend refinement layer that can be attached to different LiDAR odometry methods, rather than requiring a specific front end.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.