scanmatcher_node-1 process has died

[Tester2009]

Hi all.

Currently I'm using ROS2 humble, and using LS Lidar C32. I have recorded a pointcloud in a ros bag and try to play it back in our office.

First I run lidarslam. It run as normal

ros2 launch lidarslam lidarslam.launch.py

Next, I run RVIZ2 to visualise the path and modified_path. Also, have no issue.

After running lidarslam and rviz2, I play the ros2 bag. The way I play the ros bag is as following:

ros2 bag play file.bag

After few seconds, the vehicle is moving and now we can see the path.

Suddenly, there is an error occured in lidarslam. The error is as following:

[scanmatcher_node-1] [pcl::VoxelGrid::applyFilter] Leaf size is too small for the input dataset. Integer indices would overflow.
[ERROR] [scanmatcher_node-1]: process has died [pid 121510, exit code -8, cmd '/home/reka/ros2_ws/install/scanmatcher/lib/scanmatcher/scanmatcher_node --ros-args --params-file /home/reka/ros2_ws/install/lidarslam/share/lidarslam/param/lidarslam.yaml -r /input_cloud:=/c32/lslidar_point_cloud'].

Now, the lidarslam is crash. and the path is not moving anymore. Can I know what the reason behind this and how to fix it?

I tried to fix it by increasing the leaf size in params.yaml, but seems not working.

[rsasaki0109]

Can you set scan_matcher.debug_flag to true and check the terminal output just before and just after the node dies? It might contain anomalous values in the input scan for some reason.
You may also want to take a video of the scan with rviz.

scan_matcher.debug_flag
https://github.com/rsasaki0109/lidarslam_ros2/blob/humble/lidarslam/param/lidarslam.yaml#L28

[Tester2009]

Thank you @rsasaki0109 for your fast response. I have enabled debug_flag by change it to true value.

Following is the video that I capture when running lidarslam.

screencast.webm

[rsasaki0109]

@Tester2009 I would like you to provide me with the YAML file.

[Tester2009]

@rsasaki0109 Here is my YAML file.

scan_matcher:
  ros__parameters:
    global_frame_id: "map"
    robot_frame_id: "base_link"
    registration_method: "NDT"
    ndt_resolution: 2.0
    ndt_num_threads: 2
    gicp_corr_dist_threshold: 5.0
    trans_for_mapupdate: 1.5
    vg_size_for_input: 0.5
    vg_size_for_map: 0.1
    use_min_max_filter: true
    scan_min_range: 1.0
    scan_max_range: 200.0
    scan_period: 0.2
    map_publish_period: 15.0
    num_targeted_cloud: 20
    set_initial_pose: true
    initial_pose_x: 0.0
    initial_pose_y: 0.0
    initial_pose_z: 0.0
    initial_pose_qx: 0.0
    initial_pose_qy: 0.0
    initial_pose_qz: 0.0
    initial_pose_qw: 1.0
    use_imu: false
    use_odom: false
    debug_flag: true

graph_based_slam:
    ros__parameters:
      registration_method: "NDT"
      ndt_resolution: 1.0
      ndt_num_threads: 2
      voxel_leaf_size: 0.1
      loop_detection_period: 3000
      threshold_loop_closure_score: 0.7
      distance_loop_closure: 100.0
      range_of_searching_loop_closure: 20.0
      search_submap_num: 2
      num_adjacent_pose_cnstraints: 5
      use_save_map_in_loop: true
      debug_flag: true

[rsasaki0109]

Hmmm... I'm a little unsure.
I think I need to know the outline of the filtered scan point cloud and the number of point clouds when the error occurs.
I'm a bit busy so I can't get started right away, but if you provide a rosbag I can take a look at it in my spare time.

[Tester2009]

Hi @rsasaki0109 . Here is the link to the rosbag. The size of rosbag is 2GB compressed.

https://drive.google.com/file/d/1c8Aq0SliIqB5iGAl35CViehIKL8L74Zz/view?usp=share_link

Look forward to your reply soon!

[Tester2009]

Hi @rsasaki0109 good day!

May I know if there any update on this issue?

[rsasaki0109]

I apologize, it slipped my mind. At a glance, it seems that the voxel grid filter inside updateMap, which is running in parallel, might be the issue.
https://github.com/rsasaki0109/lidarslam_ros2/blob/humble/scanmatcher/src/scanmatcher_component.cpp#L422

Due to time constraints, I'll check again on a weekday evening or weekend.

[rsasaki0109]

I apologize for the inconvenience, but due to recent busyness and the complexity of the issue, I would appreciate some time to address it.

[colconbator]

I am facing the same problem, whenever I play rosbag2 play bag file command the bag is getting closed and rviz is not showing any map

[rsasaki0109]

memo

fix(ndt_scan_matcher): fixed a lock scope in update_ndt
autowarefoundation/autoware.universe#5951

[sbrodeur]

I had the same issue for my custom recordings with an OS1-64. I just needed to adjust the parameters vg_size_for_input and vg_size_for_map. In my case, increasing vg_size_for_map to 0.5 solved the crash.

[miku54]

I had the same issue for my custom recordings with an OS1-64. I just needed to adjust the parameters vg_size_for_input and vg_size_for_map. In my case, increasing vg_size_for_map to 0.5 solved the crash.

I also encountered the same problem during debugging. Which node is the parameter of vg_size_for_map? I can't find it.

[sbrodeur]

The parameter is documented here: https://github.com/rsasaki0109/lidarslam_ros2/blob/develop/README.md?plain=1#L85
You can set the value in the ROS parameter file that is configuring the scan matching node, as in here: https://github.com/rsasaki0109/lidarslam_ros2/blob/develop/lidarslam/param/lidarslam.yaml#L11

[rsasaki0109]

If it crashes, it's better to increase vg_size_for_map, but in terms of accuracy, it's best to keep it as small as possible.