search

Huguet57 / LIMO-Velo

A real-time, direct and tightly-coupled LiDAR-Inertial SLAM for high velocities with spinning LiDARs
GNU General Public License v2.0
276 stars 54 forks source link

Leaf size is too small for the input dataset #26

Closed wpwebrobotics closed 1 year ago

wpwebrobotics commented 2 years ago

Hi, thanks for sharing the code. I am trying the slam using a rosbag made from Velodyne VLP-16 output and IMU Vectornav 100. The rosbag seems to me to have nothing special but I get this error: [pcl::VoxelGrid::applyFilter] Leaf size is too small for the input dataset. Integer indices would overflow. I have tried different values for downsample_prec (from 0.1 to 3) to no avail. What could it be? Thank you.

Huguet57 commented 2 years ago

Check the number of points in compensated. This is a problem that happens when using the pcl::VoxelGrid and a too small pointcloud.

You can try bigger deltas (try keeping it at 0.1 in the last stage of the Initialization), lowering the downsample_rate (to 1?) and check that the min_dist is adequate for your robot/aplication (maybe indoor robots do need points in a 4 meter radius) ideally the min_dist would cover just only the robot.

If it's a puntual occurrence, you could try to increase the Config.MAX_POINTS2MATCH hyperparameter (not referenced in the YAML): https://github.com/Huguet57/LIMO-Velo/blob/57f32058eddab9ed1461f856bb3987742c22f362/src/main.cpp#L81

Huguet57 commented 2 years ago

I would guess it's likely to be the min_dist value if it's an indoor application. The default value of 4 in the xaloc.yaml file is for a 2m x 1.5m autonomous car in an outdoor scenario.

Huguet57 commented 1 year ago

We have observed that this problem disappears if we require more IMU readings when initializing. Thanks for pointing it out and sorry for the delayed response!