SOLiD-PyICP-SLAM

This repository is the SOLiD-based Full Python SLAM for Narrowing your FOV with SOLiD: Spatially Organized and Lightweight Global Descriptor for FOV-constrained LiDAR Place Recognition. The results below are in order of KITTI 00, 02, 05, and 08 sequences.

Hogyun Kim, Jiwon Choi, Taehu Sim, Giseop Kim, Younggun Cho^†

Spatial AI and Robotics Lab (SPARO)

SOLiD based Full Python based SLAM

• SOLiD (Spatially Organized and Lightweight global Descriptor for LiDAR Place Recognition) is a lightweight and fast LiDAR global descriptor for FOV constraints situations that are limited through fusion with other sensors or blocked by robot/sensor operators including mechanical components or solid-state LiDAR (e.g. Livox).
• We estimate odometry using Point2Plane ICP in Open3D and optimize the pose graph using GTSAM.
• Purpose
  • This implementation is fully Python-based so slow and underperforming, but for educational purposes.
• Cpp version will be revealed soon.
  • (TBD) Integrated with A-LOAM: SOLiD-A-LOAM
  • (TBD) Integrated with LOAM-LIVOX: SOLiD-LOAM-LIVOX
• Scan Context fails in KITTI 08, which exists a lane-level reverse loop, but SOLiD detects this loop.

Prerequisite

• SOLiD
• GTSAM
• OPEN3D
• FFMPEG

Run

0. Download SOLiD-PyICP-SLAM.

$ git clone https://github.com/sparolab/SOLiD-PyICP-SLAM.git

1. Download KITTI in Datasets Folder.

$ mkdir Datasets

2. Just RUN!!

$ python3 main.py

Supplementary

• Arxiv
• Paper
• Video
• Project page

Main Contribution

• Hogyun Kim
• Jiwon Choi
• Taehu Sim
• Giseop Kim
• Younggun Cho

QnA

• If you have a question, you utilize a alphaXiv and comment here.

Special Thanks

Thank you Giseop Kim and MyeongHwan Jeon for providing the base code.

Citation

@article{kim2024narrowing,
  title={Narrowing your FOV with SOLiD: Spatially Organized and Lightweight Global Descriptor for FOV-constrained LiDAR Place Recognition},
  author={Kim, Hogyun and Choi, Jiwon and Sim, Taehu and Kim, Giseop and Cho, Younggun},
  journal={IEEE Robotics and Automation Letters},
  year={2024},
  publisher={IEEE}
}