Ubuntu == 20.04

ROS == Noetic. ROS Installation

(Other versions haven't been tested)

PCL >= 1.8, Follow PCL Installation.

Eigen >= 3.3.4, Follow Eigen Installation.

GTSAM >= 4.0.0(tested on 4.0.0-alpha2)

Follow livox_ros_driver Installation.

Remarks:

• Since the FAST-LIO must support Livox serials LiDAR firstly, so the livox_ros_driver must be installed and sourced before run any FAST-LIO luanch file.
• How to source? The easiest way is add the line source $Livox_ros_driver_dir$/devel/setup.bash to the end of file ~/.bashrc, where $Livox_ros_driver_dir$ is the directory of the livox ros driver workspace (should be the ws_livox directory if you completely followed the livox official document).

Assume you have installed CUDA, check this link for the CUDA Toolkit version and install here.

For LibTorch, follow libtorch installation, here i choose to use Stable-Linux-LibTorch-C++/Java-CUDA11.8.

For the question:

BEV-LIO-LC/include/REM.hpp:122:107: error: no matching function for call to ‘torch::nn::functional::GridSampleFuncOptions::mode(const torch::enumtype::kBicubic&)’
122 | auto options = torch::nn::functional::GridSampleFuncOptions().align_corners(true).mode(torch::kBicubic);

Change the code at Line 19 in libtorch/include/torch/csrc/api/include/torch/nn/options/vision.h

typedef std::variant<enumtype::kBilinear, enumtype::kNearest, enumtype::kBicubic> mode_t;

Remember to change the path of LibTorch in the CMakeLists.txt to yours.

We directly use the model from BEVPlace++, check /src/models/tool, you can use turn.py to transform your trained model to .pt format to be used in our codes at src/BEV_LIO/models/tool.

And change the path of resnet_weights.pth at 86 line in REM.hpp.

Clone the repository and catkin_make:

    cd ~/$A_ROS_DIR$/src
    git clone https://github.com/HxCa1/BEV-LIO-LC.git
    cd ..
    catkin_make
    source devel/setup.bash

• Remember to source the livox_ros_driver before build (follow 1.3 livox_ros_driver)
• If you want to use a custom build of PCL, add the following line to ~/.bashrc export PCL_ROOT={CUSTOM_PCL_PATH}

The MCD dataset can be downloaded here.

To run a sequence of ntu:

roslaunch bev_lio_lc mcd_ntu.launch

To run a sequence of kth or tuhh:

roslaunch bev_lio_lc mcd_kth_tuhh.launch

The NCD dataset can be downloaded here.

To run a sequence:

roslaunch bev_lio_lc NCD.launch

The M2DGR dataset can be downloaded here.

To run a sequence:

roslaunch bev_lio_lc m2dgr.launch

Please consider citing our work if you find our code or paper useful:

@inproceedings{cai2025bev,
title={BEV-LIO(LC): BEV Image Assisted LiDAR-Inertial Odometry with Loop Closure},
author={Haoxin Cai and Shenghai Yuan and Xinyi Li and Junfeng Guo and Jianqi Liu},
booktitle={Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
year={2025},
address={Hangzhou, China}
}

Thank the authors of BEVPlace++, FAST-LIO2, COIN-LIO, MapClosures and FAST-LIO-SAM for open-sourcing their outstanding works.

For the problem of FAST-LIO-SAM that crashes during long-term runs, we refer to FAST-LIO-SAM-LOOP.

• L. Luo, S.-Y. Cao, X. Li, J. Xu, R. Ai, Z. Yu, and X. Chen, “Bevplace++: Fast, robust, and lightweight lidar global localizationfor unmanned ground vehicles,” IEEE Transactions on Robotics (T-RO), 2025.
• W. Xu, Y. Cai, D. He, J. Lin, and F. Zhang, “Fast-lio2: Fast direct lidar-inertial odometry,” IEEE Transactions on Robotics, vol. 38, no. 4, pp.2053–2073, 2022.
• P. Pfreundschuh, H. Oleynikova, C. Cadena, R. Siegwart, and O. An-dersson, “Coin-lio: Complementary intensity-augmented lidar inertial odometry,” in 2024 IEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 1730–1737.
• S. Gupta, T. Guadagnino, B. Mersch, I. Vizzo, and C. Stachniss, “Effectively detecting loop closures using point cloud density maps,” in Proc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA),2024.
• J. Wang, “Fast-lio-sam: Fast-lio with smoothing and mapping.” https://github.com/kahowang/FAST_LIO_SAM, 2022.