Continuous-Time Fixed-Lag Smoothing for LiDAR-Inertial-Camera SLAM

CLIC is a tightly-coupled multi-modal SLAM algorithms based on continuous-time fixed-lag smoothing, it supports a variety of sensor combinations, like the LiDAR-inertial and LiDAR-inertial-camera SLAM systems, and online timeoffset calibration is also naturally supported. More importantly, benefiting from the marginalization and the derived analytical Jacobians for optimization, the proposed continuous-time SLAM systems can achieve real-time performance regardless of the high complexity of continuous-time formulation.

Factor graphs of multi-sensor fusion. For more details, please refer to the paper. [arxiv]

• ROS (tested with Melodic)

• Eigen3

• Ceres 1.14

• OpenCV 3.3

• yaml-cpp

  sudo apt-get install libyaml-cpp-dev

• livox_ros_driver

• Sophus has been included in src/sophus_lib

mkdir -p ~/catkin_clic/src
cd ~/catkin_clic/src
git clone https://github.com/APRIL-ZJU/clic.git
git clone https://github.com/Livox-SDK/livox_ros_driver.git
cd ~/catkin_clic
catkin_make
source devel/setup.bash

• Download ntu_viral_dataset or newer_college_dataset or LVI dataset or livox dataset.

• Configure parameters in the config/ct_odometry_***.yaml file.

  • config_path: the path of config folder
  • bag_path: the file path of rosbag

• Run on ntu viral dataset for example.

  roslaunch clic odometry.launch config_path:=config/ct_odometry_ntu.yaml

  The estimated trajectory is saved in the folder./src/clic/data.

This code was developed by Jiajun Lv and Xiaolei Lang from APRIL Lab in Zhejiang University.

For researchers that have leveraged or compared to this work, please cite the following:

@article{lv2023continuous,
  title={Continuous-Time Fixed-Lag Smoothing for LiDAR-Inertial-Camera SLAM},
  author={Lv, Jiajun and Lang, Xiaolei and Xu, Jinhong and Wang, Mengmeng and Liu, Yong and Zuo, Xingxing},
  journal={IEEE/ASME Transactions on Mechatronics},
  year={2023},
  publisher={IEEE}
}

• The spline module is adapted from basalt.
• The lidar fearure module is adapted from LIO-SAM and loam_livox.
• The visual module is adapted from VINS-Mono.
• The IMU initializer module is adapted from open_vins.

Thanks for their excellent job!

The code is released under the GNU General Public License v3 (GPL-3).