LaserScan Display is transforming messages in an unexpected way

[john-maidbot]

Hello,

Currently, the rviz2 LaserScan plugin converts laser scans to pointclouds for visualization using the laser_geometry package: https://github.com/ros-perception/laser_geometry/blob/71ba748322a38821eca58e79aeec475379026b68/src/laser_geometry.cpp#L416C23-L416C54

If I publish a laserscan, I would expect that it would be visualized as a pointcloud in the frame listed in the message header. (i.e. convert the range measurements to points in the sensor frame, and rigidly transform all those points into the frame I am currently visualizing).

However, since rviz2 is using transformLaserScanToPointCloud, something else is happening.
(https://github.com/ros2/rviz/blame/be68734b712dbc0f1f556d520bb993eb954fdaff/rviz_default_plugins/src/rviz_default_plugins/displays/laser_scan/laser_scan_display.cpp#L102)

transformLaserScanToPointCloud does the following (oversimplifying for brevity):

• convert the range measurements to points
• transform those points into the frame being visualized
• transform each point in the scan based on its time offset from the first scan (under the assumption that the points were not all captured at the same time) (https://github.com/ros-perception/laser_geometry/blob/71ba748322a38821eca58e79aeec475379026b68/src/laser_geometry.cpp#L322)

I think that this behavior distorts the users understanding of the information held in the laser scan. Additionally, the user might have a different algorithm for handling the per point transformation of points in the laser scan, which would conflict with the transformation rviz is doing under the hood (though in that case you could just set the scan time to 0).

Would you be opposed to providing users an option to enable/disable this behavior?

Thank you for your time.