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

Posted by derektan1995 on
URL: http://official-rtab-map-forum.206.s1.nabble.com/Tips-needed-to-improve-on-RTAB-Mapping-for-legged-Robots-tp7417p7605.html

Hi Mathieu,

Appreciate your reply. After performing extensive testing, it turns out setting odom_sensor_sync to true did not affect the results much. The 3d map became much better after I performed localization using a 3d point cloud instead of 2d laser scan. Nevertheless, I will keep the `odom_sensor_sync` parameter to true.

I have a few follow-up question:

1) What is the difference between scan-to-gridmap vs scan-to-mapscan that you have mentioned? I understand that for AMCL, particle's laserscan measurements are compared to the robot's real laserscan measurements, such that particles' laserscan with low similarity to robot's laserscan will be more likely removed during the resampling process. How is this different from RTABMAP's scan-to-mapscan?

2) Can a rtabmap database obtained from a robot with a 2d lidar be shared with a robot with 3d lidar? I experimented with this, but got an error that says the following? I wonder if this issue is linked to (1)?

[FATAL] LaserScan.cpp:241::LaserScan() Condition (data.channels() != 3 || (data.channels() == 3 && (format == kXYZ || format kXYI))) not met! [format=XY]

3) When running rtabmap that subscribes to scan_cloud, it appears that I have the following warning at each timestep RTABMAP adds a node to the posegraph (i.e. Rtabmap/DetectionRate). The 0 0 0 0 message is rather undescriptive. I wonder what this means?

[WARN] [TIMESTAMP] 0 0 0 0

Thanks alot,
Derek