[ros-users] Design decisions behind libTF...?

[Tully Foote]

Hi alll,
Thanks for the comments.  I think that you will find that tf can support
most of your use cases as implemented, though it may take a little
creativity(like the camera inversion flag).  I've opened a few tickets on
actionable items from your comments.
https://code.ros.org/trac/ros-pkg/ticket/4423
https://code.ros.org/trac/ros-pkg/ticket/4422 and definitely will be reading
over these comments as we go forward.

Ivan, with respect to your comment about a way to automatically merge
multiple uncertain sources was thought about for integration into tf, it was
chosen to explicitly seperate that capability out of tf.  For although tf
could relatively easily apply a naive averaging filter, the parts of
robot_pose_ekf such as zeroing the height of the robot.  This is application
specific knowledge which can greatly improve performance of the system and
finding a way to make tf be able to handle every intricacy of every
application would make it way too complicated.  Instead you can just do the
application specific work separately and publish the result into the tf
structure.  If you want a more generic filter than the robot_pose_ekf you
could file a feature request ticket with what features you would
specifically want.

Thanks for the feedback and comments,
Tully
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100910/26a71472/attachment-0003.html>