Login  Register

Re: icp_odometry with LOAM crash

Posted by huibuh on Sep 30, 2021; 9:59am
URL: http://official-rtab-map-forum.206.s1.nabble.com/icp-odometry-with-LOAM-crash-tp8261p8618.html

Hi Mat
First of all, thank you for your detailed post, very much appreciated. I was able to reproduce your results from the kitti dataset using the standalone test_velodyne.launch file. Next, I played back by own dataset from a Ouster OS0-64, which also worked like a treat. However, when I then tried to run rtabmap with floam using my own tf tree, icp_odometry with Odom/Strategy=11 did not work, but Odom/Strategy=0 did. Upon closer investigation I came to the conclusion that floam cannot handle an offset in z-direction of the lasers's frame_id ('velodyne' in test_velodyne.launch).

My tf tree roughly looks like this:
map->odom->base_footprint->base_link->os_lidar
where 'os_lidar' is the equivalent to 'velodyne' in test_velodyne.launch

If the z-offset between base_footprint and os_lidar is non-zero, floam produces incorrect output (offset between 0.1-0.7m) or just outputs zero values (offset >0.7m).

Any idea what causes this behavior?

Cheers,
Heiko