Revision history [back]

@lucasw I'm not sure that this is the correct answer either, but after some thought, it's the answer that makes the most sense to me. It's applicable to any sensor whose measurement device is "globally-anchored".

I started treating the marker (attached to the robot) and the estimate of the pose of that marker (determined by a camera mounted in the map frame) as separate entities. It makes sense because the physical link between marker and the robot is fixed, where as the link between the pose estimate of the marker and the robot is not guaranteed to be fixed (because of localization error). This method allows for comparing the global measurement to the fused odometry source the robot uses as its pose within the map.

map-->odom-->base_link-->ar_tag
 \----->canera---------------->arg_tag_estimate

A comparison between the physically-linked ar_tag and the ar_tag_estimate is just a transform to a common frame. Does that all sound reasonable?

@lucasw I'm not sure that this is the correct answer either, but after some thought, it's the answer that makes the most sense to me. It's applicable to any sensor whose measurement device is "globally-anchored".

I started treating the marker (attached to the robot) and the estimate of the pose of that marker (determined by a camera mounted in the map frame) as separate entities. It makes sense because the physical link between marker and the robot is fixed, where as the link between the pose estimate of the marker and the robot is not guaranteed to be fixed (because of localization error). This method allows for comparing the global measurement to the fused odometry source the robot uses as its pose within the map.

map-->odom-->base_link-->ar_tag
 \----->canera---------------->arg_tag_estimate
\----->canera---------->arg_tag_estimate

A comparison between the physically-linked ar_tag and the ar_tag_estimate is just a transform to a common frame. Does that all sound reasonable?