In general, I find it useful for any node that deals with pose estimation
to publish a PoseStamped / PoseWithCovarianceStamped in addition to tf.
Also, it's nice to have a flag which enables/deisables the pose and tf
publishing, letting the user choose which one they want. The rationale
behind this is to allow different components, which may publish partial and
conflicting observations about the pose in a given frame, to be used
together. If they publish poses, then can be used as input to a
higher-level package which fuses them, and in turn publishes a tf.
This applies to any odometry, visual odometry, ekf filters, slam, etc
pipelines.
I have adopted this convention for the laser_scan_matcher, with publish_tf
defaulting to true and publish_pose_stamped defaulting to false.
So in short:
+1 for a flag to enable PoseWithCovarianceStamped
+1 for a flag to disable tfs
Cheers,
Ivan
On Tue, Nov 27, 2012 at 3:59 PM, Stéphane Magnenat <
stephane.magnenat@mavt.ethz.ch> wrote:
> Hello,
>
> The current draft of REP129 for SLAM node API [1] only defines tf
> output. However, in many SLAM application the output pose of the
> registration is fused with additional sensors such as an IMU, typically
> through a Kalman filter. Hence, it might be useful for the SLAM nodes to
> output a PoseWithCovarianceStamped message in addition to the tf
> transform. Should this be an optional addition, or a mendatory output?
>
> Your opinion is warmly welcome, especially if you maintain a SLAM package.
>
> Cheers,
>
> Stéphane
>
> [1] https://github.com/ros-infrastructure/rep/pull/8/files
>
> --
> Dr Stéphane Magnenat
> http://stephane.magnenat.net
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
(text/plain)