SGT-LLC:

LiDAR Loop Closing Based on Semantic Graph with Triangular Spatial Topology

Shaocong Wang, Fengkui Cao, Ting Wang Xieyuanli Chen, Shiliang Shao

IEEE Robotics and Automation Letters

News

• 1 March 2025: Code updata
• 10 February2025: Accepted by IEEE RAL!

Getting Started

Instructions

The real-time demo of SGT-LLC was built on rangenet-ros . Please refer to rangenet-ros to configure the semantic segmentation environment.

Dependencies

• Ubuntu 20.04
• ROS Noetic (roscpp, std_msgs, sensor_msgs, geometry_msgs, pcl_ros)
• C++ 14
• OpenMP
• Point Cloud Library
• Eigen >=3.3.4
• DBow3

Compiling

Create a catkin workspace, clone the SGT-LLC repository into the src folder, and download the test semantic segment model into ./SGT-LLC/model/. Finally, Compile via the catkin_tools package:

mkdir ws && cd ws && mkdir src && catkin init && cd src
git clone https://github.com/ROBOT-WSC/SGT-LLC.git
catkin build

Execution

For your convenience, KITTI, MulRan and MCDcan be real-time test on SGT-LLC. For KITTI, the raw rosbag can be directly used for loop close detection. For MulRan and MCD, we provide some bags with semantic labels. These examples can be found here. To run, first launch demo via:

roslaunch rangenet_pp "kitti/mulrun/mcd".launch

In a separate terminal session, play back the downloaded bag:

rosbag play "bag's name" --clock

Citation

If you find SGT-LLC is useful in your research or applications, please consider giving us a star 🌟 and citing it by the following BibTeX entry.

@ARTICLE{10891171,
  author={Wang, Shaocong and Cao, Fengkui and Wang, Ting and Chen, Xieyuanli and Shao, Shiliang},
  journal={IEEE Robotics and Automation Letters}, 
  title={SGT-LLC: LiDAR Loop Closing Based on Semantic Graph With Triangular Spatial Topology}, 
  year={2025},
  volume={10},
  number={4},
  pages={3326-3333},
  keywords={Semantics;Topology;Point cloud compression;6-DOF;Data mining;Laser radar;Optimization;Encoding;Accuracy;Vectors;SLAM;localization;mapping},
  doi={10.1109/LRA.2025.3542695}}

Acknowledgements

We thank the authors of the semantic-histogram-based-global-localization and rangenet-ros open-source package.