This lab_slam is an exercise for me, both in SLAM theory and programing skills, thus it may be naive for some experts.

Currently, it completed a lidar odometry, loop clousre detecion based on pose, and pose graph optimization after a clousre is found successfully.

In the above picture, the blue line shows the pose path of purely scan-to-scan matching odometry, and comparatively the red one means the result after scan-to-map matching refinement, which is obviously better.

A loop clousre is indicated by a green line, which links the current pose with a historical one, and the the pose graph is optimized using gtsam, leading to a jump of current scan-to-map pose path(in red).

Please refer to /img directory for more demos.

• ROS(tested with ROS melodic 1.14.10)
• Eigen(tested with eigen-3.2.10)
• PCL(tested with pcl-1.8)
• Glog
• GTSAM(tested with gtsam-4.0.2)
• Ceres(tested with ceres-1.14.0)
• #OpenCV(just for visualization)

cd ~/catkin_ws/src
git clone [email protected]:TongxingJin/lab_slam.git
cd ..
catkin_make

source devel/setup.bash
roslaunch lab_slam run.launch

I use rosbag data provided in LIO-SAM to debug this lab_slam.

It's also avaliable from Baidu Netdisk at:
link:https://pan.baidu.com/s/1MqQD22d4sA3iUszlWg3C8Q password:2eyj