- 构建关键帧(完成)
- 将关键帧添加到因子图中(完成)
- 检测回环优化(完成)
- 回环优化后对姿态和局部地图进行重新矫正(回环优化后,出现多线程数据冲突问题,参考LIO-SAM来解决问题)已解决
- 新发现一个bug,建图出现重影,有两个,但是暂时找不到问题所在。(2023年6月20号)
有任何问题,都可以提出,欢迎交流!
S-FAST_LIO is a simplified implementation of FAST_LIO (Xu, Wei, et al. "Fast-lio2: Fast direct lidar-inertial odometry."), which is modified from FAST_LIO. This code is clean and accessible. It is a reference material for SLAM beginners.The main modifications are as follows:
- The Sophus is used to define the state variables, instead of the complicated IKFOM
- The gravity component is directly defined by a Vector3d, thus the complicated calculation of two-dimensional manifold can be omitted
- The code structure has been optimized, and the unnecessary codes have been deleted
- Detailed Chinese notes are added to the code
- Support for Robosense LiDAR has been added
In addition, the following links are also my previous works. I strongly recommend reading them, since they are the interpretation and detailed equation derivation of the FAST-LIO paper:
Ubuntu >= 16.04.
PCL >= 1.8, Eigen >= 3.3.4.
Follow livox_ros_driver Installation.
Source:Add the line source $Licox_ros_driver_dir$/devel/setup.bash
to the end of file ~/.bashrc
, where $Licox_ros_driver_dir$
is the directory of the livox ros driver workspace (should be the ws_livox
directory if you completely followed the livox official document).
We use the old version of Sophus,这个问题已经解决,可以使用最新的版本,并且跳过这个部分
git clone https://github.com/strasdat/Sophus.git
cd Sophus
git checkout a621ff
mkdir build
cd build
cmake ../ -DUSE_BASIC_LOGGING=ON
make
sudo make install
你需要自行搜索,并且安装gtsam优化库。
Clone the repository and catkin_make:
cd ~/catkin_ws/src
git clone https://github.com/zlwang7/S-FAST_LIO.git
cd ../
catkin_make
source ~/catkin_ws/devel/setup.bash
Here we provide some additional Avia Rosbags. They are collected by Arafat-ninja.
Files: Can be downloaded from google drive. You can also directly use the Avia Rosbags provided by FAST_LIO (google drive).
Run:
roslaunch sfast_lio mapping_avia.launch
rosbag play YOUR_DOWNLOADED.bag
Datasets are collected by a RS-Helios LiDAR and an Xsens IMU.
Files: Can be downloaded from google drive.
Run:
roslaunch sfast_lio mapping_rs.launch
rosbag play YOUR_DOWNLOADED.bag
The same as FAST_LIO.
Here we use the dataset (The Main Building in the University of Hong Kong) in FAST-LIO paper. The sensor suite is handheld during the data collection and returned to the starting position after traveling around 140m. The average number of effective feature points is 2015. All algorithms are tested on an Intel i7 2.3 GHz processor-based computer.
Performance Comparison | Average processing time | Drift |
---|---|---|
FAST-LIO | 7.52ms | 0.035% |
S-FAST_LIO | 7.07ms | 0.037% |
Thanks for the authors of FAST-LIO.