Re: Tips needed to improve on RTAB Mapping for legged Robots

Posted by matlabbe on
URL: http://official-rtab-map-forum.206.s1.nabble.com/Tips-needed-to-improve-on-RTAB-Mapping-for-legged-Robots-tp7417p7609.html

Hi Derek,

If the timestamps between sensors and/or odometry are taken from different clocks not synchronized, odom_sensor_sync won't really help. However, this parameter is false by default as it has not be extensively tested in many scenarios, so it may not work as expected in all setups.

1) AMCL will localize against the occupancy grid map (a map from map_server for example), not the raw scans. RTAB-Map does it against the raw scans, by doing ICP between the current scan and the closest scans in the map.

2) Which rtabmap version are you using? I may test this case as it should not do a FATAL error. I had problems differencing between XYZ and XYI formats, but it should have been fixed now. However, I don't think ICP will work correctly when mixing 2D and 3D scans.

3) This has been fixed in January: https://github.com/introlab/rtabmap_ros/commit/21090899d423fd7795fd36c9d33d8efde28de7dc#diff-6a3ab41a835ba9b64ad85e420a4b9f270d8a4def12845afc786bc2c21f438eb6

cheers,
Mathieu