[ros-users] robot_pose_ekf /vo input units

[Jack O'Quin]

On Wed, Apr 28, 2010 at 9:35 AM, William Woodall <wjwwood at gmail.com> wrote:
> Wow, thanks Jack, the ROS specific part of your gps code will probably work
> for us.  I think your right though, ROS should have a standardized gps msg
> that can be linked through a standard gps-to-odom node that can do these
> calculations for you given just your gps topic.  I really appreciate your
> help, this should really save us a lot of time implementing and
> troubleshooting these calculations.

You're welcome to use any of that code, it is all open source. Please
report any bugs you find.

There is currently one known defect. I believe it is in the device
driver, not the ROS interface:

  http://code.google.com/p/utexas-art-ros-pkg/issues/detail?id=2

I believe the main reason ROS currently has no standard GPS message is
because all the WG work is done with indoor robots. If several
projects agreed on a common GPS message format, I expect the ROS
developers might agree to put it in a common place like nav_msgs. That
would allow us to better share code. Or, we could create a GPS stack
with some more cleanly separated packages in utexas-art-ros-pkg. A
common GPS monitoring package would be useful, for example. The GPS
support we have right now is not very general.

I like your idea for a common ROS package to translate GPS input into
standard Odometry and tf messages. For some applications extra latency
might be undesirable, but message bandwidth is not a problem. In our
system, most navigation and obstacle detection is done using Odometry,
there are only a few components (dealing with maps and such) that need
GPS information.
-- 
 joq