This repository is a modified version of LVI_SAM.
- Custom extrinsic parameters are adopted in the code.
- The original code assumes there are no translation between the sensors. Now extrinsic parameters including both translation and rotation among IMU, Camera, LiDAR in .yaml in the
configfolder are correctly used.- params_lidar.yaml: lidar to imu extrinsic (i.e. imu pose w.r.t lidar frame)
- params_camera.yaml: imu to cam, lidar to cam, lidar to imu extrinsic (sorry for the cubersome process...)
- Currently, the code is modified to use the
Ouster. - If you want to use the
velodyne, you have to uncomment here and here and comment here and here. - Also, you need to change the .yaml file.
- The paper claims the system is tightly-coupled for its lidar-depth association mechanism and initial odometry from VINS for LIO-SAM optimization.
- Yet, there are two odometries output from the system, one from VINS and one from LIO-SAM. You have to choose between the two and may require a hard switching, resulting in a sudden jump in the odometry.
- More tightly-coupled system with one single odometry output should be developed to avoid the issue.
- The original version is from LVI-SAM.