Re: [ros-users] Map Topic Idea

Excellent feedback, Thanks everyone!

~Dereck


“The greater danger for most of us lies not in setting our aim too high and
falling short; but in setting our aim too low, and achieving our mark.” -
Michelangelo


On Mon, Mar 25, 2013 at 8:13 PM, Michael Gratton <mikeg@cse.unsw.edu.au>wrote:

> On 26/03/13 03:10, Dave Hershberger wrote:
> > I wish I had time to actually work on this right now. Failing that I
> > will just suggest:
> >
> > Please don't break backwards compatibility. It makes unnecessary work
> > for lots of people down the road, and doesn't save that much work now.
> > It's hard enough migrating from one ROS release to the next already,
> > let's not (knowingly) make it worse.
>
> +1! While ROS clearly has no API/ABI stability contracts from release to
> release, the fewer breaking changes the better. Especially with such a
> rapid release cycle.
>
> > My suggestion is to make a new message, called CompressedOccupancyGrid.
> > It would look exactly like OccupancyGrid, but would have an "encoding"
> > byte which would support PNG and maybe some other lossless encoding
> > method that is more optimized for speed rather than compression size.
> > (A simple run-length encoding might do pretty well.)
> >
> > Next step might be to make a library for subscribing/publishing
> > OccupancyGrid and/or CompressedOccupancyGrid so that code doesn't need
> > to be repeated and so nodes can have a consistent message API going
> > forward. Maybe map publishers advertise "map" and "map/compressed" and
> > only publish to whichever topic gets subscribed to. Subscribers would
> > look for "map/compressed" and "map" and subscribe to compressed if
> > available and uncompressed if not. This way new code would work with
> > old code but be fast if possible.
>
> This sounds like the best approach to me.
>
> //Mike
>
> --
> Michael Gratton <http://www.cse.unsw.edu.au/~mikeg/>
> UNSW School of Computer Science and Engineering.
>
>
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>