[ros-users] Occupancy grid map with customized sensor model

Eitan Marder-Eppstein eitan at willowgarage.com
Fri May 14 17:03:53 UTC 2010


Sorry its taken a bit to get back to you. You are correct that costmap2D is
not representing the likelihoods of obstacles in the environment. We chose
to do this for efficiency reasons, but we do have the luxury of sensors that
have reasonable precision, so it hasn't been a big issue for us.
Unfortunately, gmapping's occupancy grid is, to my knowledge, the closest we
come to having code for the probabilistic case.

If you decide to write a probabilistic occupancy grid, I'd be happy to
answer any questions you have along the way. Also, we could work on getting
it into the navigation stack at some point as its code that would be nice to
have available to everyone.

Hope all is well,


On Wed, May 12, 2010 at 1:30 PM, Daniel Maier <
maierd at informatik.uni-freiburg.de> wrote:

> Hi everyone!
> I was wondering if I am correct in my assumption that the costmap2D is not
> representing likelihoods of obstacles in the environment but assumes the
> sensor readings are more or less perfect and the values in the grid cells
> represent likelihoods of collisions with the robot.
> (As opposed to the classical occupancy grid mapping proposed by Moravec
> and Elfes).
> If I am right, is there another ROS module for this job, preferably less
> sophisticated than gmapping?
> If I am wrong, is there a "default" way to include new sensor models for
> e.g. updates from camera, sonar, etc. to better model these sensor's
> characteristics and uncertainties?
> Thanks a lot!
> Dan
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100514/91926f0d/attachment-0003.html>

More information about the ros-users mailing list