Re: TIps on increasing the accuracy and robustness of icp odometry

Posted by matlabbe on
URL: http://official-rtab-map-forum.206.s1.nabble.com/TIps-on-increasing-the-accuracy-and-robustness-of-icp-odometry-tp10685p10700.html

dinossht wrote

When I enable deskew, the odometry does not work. Have you experienced this?

What kind of warnings/errors do you have? Note however that I am note sure if the timestamps set in livox point cloud are as expected (I tested mostly on velodyne and ouster lidars). If you can share a small rosbag of livox point cloud and imu, that would be useful to check if the format is as expected.

dinossht wrote

And what does this do exactly: PointToPlaneGroundNormalsUp? If I walk with the lidar sensor, does this apply as well?

If the lidar is always held upwards, that could be used. When computing normals of point on the ground, if the distance between two rays is too large, the estimated normal can be upside down. That option will flip all normals upward if they are close to 0.9*-z axis. This is all done in base frame, so it is why the lidar is rotated, it won't work properly. I mostly enable this on ground robots or cars with a lidar fixed upside up.