pose drift largely about lidar_localization_ros2 HOT 17 CLOSED

[email protected]

I sent you again few seconds ago

from lidar_localization_ros2.

In my test I did not use imu data
Here is my parms:

`
/**:
ros__parameters:
registration_method: "GICP"
ndt_resolution: 1.0
ndt_step_size: 0.1
transform_epsilon: 0.01
voxel_leaf_size: 1.0
scan_max_range: 100.0
scan_min_range: 1.0
scan_period: 0.1
use_pcd_map: true
map_path: "/home/terra/ros2_ws/src/pcl_localization_ros2/map/map.pcd"
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_odom: false
use_imu: false
enable_debug: false

`

from lidar_localization_ros2.

Indoors, the following parameters should be reduced.

ndt_resolution (0.5 - 2.0 [m] )
voxel_leaf_size(0.1~1.0[m])

It might also be a good idea to set registration_method to NDT.

Also, is the initial pose correct?

from lidar_localization_ros2.

By the way, which package and which lidar did you use to create the map point cloud?

from lidar_localization_ros2.

Indoors, the following parameters should be reduced.

ndt_resolution (0.5 - 2.0 [m] ) voxel_leaf_size(0.1~1.0[m])

It might also be a good idea to set registration_method to NDT.

Also, is the initial pose correct?

registration_method to NDT also will drift.

I record a bag use velodyne 16p. use your slam bag to create a pcd map,and then play the lidar bag to test the localization.

from lidar_localization_ros2.

By the way, which package and which lidar did you use to create the map point cloud?

I record a bag use velodyne 16p. use your slam bag to create a pcd map,and then play the lidar bag to test the localization.

from lidar_localization_ros2.

By the way, which package and which lidar did you use to create the map point cloud?

I record a bag use velodyne 16p. use your slam bag to create a pcd map,and then play the lidar bag to test the localization.

give me a email I can share you my lidar data and the pcd file

from lidar_localization_ros2.

You might want to set voxel_leaf_size to 0.1 or 0.5.

from lidar_localization_ros2.

What is the SLAM you used, is it OSS?hdl_graph_slam?lio-sam?lidarslam_ros2?

You can share pcd and rosbag for me to see.
My email address is below.
[email protected]

from lidar_localization_ros2.

What is the SLAM you used, is it OSS?hdl_graph_slam?lio-sam?lidarslam_ros2?

You can share pcd and rosbag for me to see. My email address is below. [email protected]

lidarslam_ros2

from lidar_localization_ros2.

If lidarslam_ros2 is also parameterized for indoor use, a more accurate map can be produced.

from lidar_localization_ros2.

Hi author have you received my pcd and rosbag?

from lidar_localization_ros2.

I have not received those yet.

from lidar_localization_ros2.

@rsasaki0109 were you able to reproduce the pose drift?

from lidar_localization_ros2.

The following parameters allow for some localization indoors.
However, localization breaks down where there is not much of a point cloud
If you add a time series filter such as Kalman filter, it may be less likely to fail.

/**:
    ros__parameters:
      registration_method: "NDT"
      ndt_resolution: 1.0
      ndt_step_size: 0.1
      transform_epsilon: 0.1
      voxel_leaf_size: 0.1
      scan_max_range: 10.0
      scan_min_range: 0.1
      scan_period: 0.1
      use_pcd_map: true
      map_path: "/home/autoware/Downloads/pcd_loc_ros2/map.pcd"
      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_odom: false
      use_imu: false
      enable_debug: true

I tried setting registration_method: "GICP" and reducing voxel_leaf_size in various ways, but it didn't seem to work.

from lidar_localization_ros2.

The following parameters allow for some localization indoors. However, localization breaks down where there is not much of a point cloud If you add a time series filter such as Kalman filter, it may be less likely to fail.

/**:
    ros__parameters:
      registration_method: "NDT"
      ndt_resolution: 1.0
      ndt_step_size: 0.1
      transform_epsilon: 0.1
      voxel_leaf_size: 0.1
      scan_max_range: 10.0
      scan_min_range: 0.1
      scan_period: 0.1
      use_pcd_map: true
      map_path: "/home/autoware/Downloads/pcd_loc_ros2/map.pcd"
      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_odom: false
      use_imu: false
      enable_debug: true

I tried setting registration_method: "GICP" and reducing voxel_leaf_size in various ways, but it didn't seem to work.

thanks for your test and reply

from lidar_localization_ros2.

Using ndt_omp has prevented loss of localization

Addition ndt_omp PR
#31

from lidar_localization_ros2.