rsasaki0109 / lidar_localization_ros2

3D LIDAR Localization using NDT/GICP and pointcloud map in ROS 2 (Not SLAM)
BSD 2-Clause "Simplified" License
231 stars 43 forks source link

the direct positioning got stuck #35

Open ming1j2 opened 2 months ago

ming1j2 commented 2 months ago

Hello, I would like to ask why I entered a source point cloud and opened the real-time point cloud of the radar. However, during operation, there was a lag phenomenon during real-time radar positioning, and in the end, the direct positioning got stuck

rsasaki0109 commented 2 months ago

Sorry, I just noticed your comment. I would like to know more details. For example, whether the environment is indoors or outdoors, or if you have done any parameter tuning for that. Also, I would like the YAML file. It would be better if there are images or videos of the position estimation. Setting ndt_num_threads to 0 (meaning maximum allowable threads are used) can sometimes cause problems, so it might be good to adjust it.

ming1j2 commented 2 months ago

In the indoor environment, my yaml file only changed the input path of the source point cloud for map_path and the number of ndtnum_threads to 4. Other changes may have been made to the point cloud topic and imu topic. The effect is that after opening it, it can be repositioned and registered, but after a period of time, it will occasionally get stuck, and in the end, the positioning will just get stuck. I don't know if there was a problem with the change

ming1j2 commented 2 months ago

And there is also a strange issue, which is that after changing the number of ndtnum_threads to 4, but the info display result of ROS2 is still 0, which makes me not sure if the subsequent repositioning has undergone multi-threaded acceleration

rsasaki0109 commented 2 months ago

In ROS2, you need to rebuild to apply parameter changes. If you add the --symlink-install option during the build, you no longer need to rebuild.

ming1j2 commented 2 months ago

But what is the problem with this repositioning getting stuck in the end? Did you encounter this issue while trying out the indoor scene radar

rsasaki0109 commented 2 months ago

I'm not quite sure what the problem is, so I'd like to see an image or video.

ming1j2 commented 2 months ago

![Uploading 微信图片_20240417130219.jpg…]()

ming1j2 commented 2 months ago

![Uploading 1.jpg…]()

rsasaki0109 commented 2 months ago

I can't view the image. The link is not functioning correctly.

ming1j2 commented 2 months ago

![Uploading 1.jpg…]()

ming1j2 commented 2 months ago

1

ming1j2 commented 2 months ago

截图 2024-04-17 13-27-25

ming1j2 commented 2 months ago

Just like this

rsasaki0109 commented 2 months ago

Can you upload the rviz screen? I'd like to see the map, trajectory, scans, etc.

rsasaki0109 commented 2 months ago

The delta angle is large. By the way, have you been swinging the LiDAR around by hand?

ming1j2 commented 2 months ago

Yes, I did swing the LiDAR with my hand. Can't I swing it with my hand? If rviz is turned on, then the trajectory is the one I swing the radar with my hand

rsasaki0109 commented 2 months ago

When you manually wave LiDAR around, without the combination with IMU, scan matching might fail. Although I have implemented proper integration options for IMU in SLAM, they aren't yet available in this LiDAR Localization.