Incorrect information in comment

[AlexanderDavid]

I think there's some confusion surrounding the different "modes" that the localization node can run in. This comment lists that if world_frame == odom_frame then the map_frame -> odom_frame will be published, otherwise the odom_frame -> base_link_frame will be published

https://github.com/cra-ros-pkg/robot_localization/blob/040dd17034d2fae55f99cc6d93598aeeb8b97b47/src/ros_filter.cpp#L825C1-L830C46

/*
  ...
  * The state estimation nodes in robot_localization therefore have two "modes."
  * If your world_frame parameter value matches the odom_frame parameter value,
  * then robot_localization will assume someone else is broadcasting a transform
  * from odom_frame->base_link_frame, and it will compute the
  * map_frame->odom_frame transform. Otherwise, it will simply compute the
  * odom_frame->base_link_frame transform.
  * /

However, this code seems to say the opposite since the filteredPosition's frame_id is set to the world_frame.

https://github.com/cra-ros-pkg/robot_localization/blob/noetic-devel/src/ros_filter.cpp#L1999C1-L2047C12

if (invertTransform_ || filteredPosition.header.frame_id == odomFrameId_)
{
  worldTransformBroadcaster_.sendTransform(
    invertTransform_ ? RosFilterUtilities::invertTF(worldBaseLinkTransMsg_) : worldBaseLinkTransMsg_);
}
else if (filteredPosition.header.frame_id == mapFrameId_)
{
  ...
    worldTransformBroadcaster_.sendTransform(mapOdomTransMsg);
}

I think the larger issue here is that the documentation lacks a nuanced explanation of what the various modes actually end up publishing and what the expected setup of the TF tree should be.

https://docs.ros.org/en/noetic/api/robot_localization/html/state_estimation_nodes.html#frame

[ayrton04]

You are correct. That comment is erroneous. Feel free to submit a PR.