ustc-ais-lab / si-lio

SI_LIO

SI_LIO is based on the invariant-KF theory and our code is implemented on S-FAST_LIO.

The accuracy is close to FAST-LIO2, but the speed is improved by 30%.

1. Dependices

Sophus Eigen PCL livox_ros_driver

2. Build

Clone the repository and catkin_make:

cd ~/catkin_ws/src
git clone https://github.com/USTC-AIS-Lab/SI-LIO.git
cd ../
catkin_make -j
source ~/catkin_ws/devel/setup.bash

3. Run

We recommend using the M2DGR dataset.

roslaunch inva_fast_lio mapping_velodyne_m2dgr.launch
rosbag play street_04.bag

4. Acknowledgements

Thanks for the authors of S-FAST-LIO and FAST-LIO.

si-lio's Issues

What is the SE_{4}(3) group?

Interesting work! In your displaed supplementary video on Youtube, you introduce that you utilize the Lie exponential between the matrix Lie group SE$_4(3)$ and its Lie algebra for the linearization of the estimation error system and measurement residual, avoiding the introduction of the linearization errors in both estimate propagation and update process. What is the SE$_4(3)$ matrix Lie group?

What's new feature？

What's new feature？ The difference with (s-)fast-lio?

耗时统计时迭代次数恒定为1？

你好，请问这个时间对比是在maximum_iter均为0的前提下统计出来的吗，我看你代码中将maximum_iter直接设置0，请问有做这方面的对比实验吗

have you benchmarked CPU usage ?

how much CPU usage SI-LIO is?

Which type of constraint are you using? Is it point-to-point or point-to-plane?

Which type of constraint are you using? Is it point-to-point or point-to-plane?