[ros-users] Different maps for navigation and path planning

Eitan Marder-Eppstein eitan at willowgarage.com
Tue Sep 7 05:33:18 UTC 2010


Eric,

You're right. Right now there's not a great way to specify a "forbidden
zone" in a map. If you specify the space as occupied, then sensors can clear
it out when it is observed to be free. I've always intended on adding a
different cost value that designates space as forbidden no matter what local
sensor information says, but honestly just haven't gotten around to it.
Since this seems like a critical thing for you, and something that a lot of
people will want, I've created a ticket for it here:
https://code.ros.org/trac/ros-pkg/ticket/4409.

In the meantime, as you alluded to, you can create a marking sensor that
asserts that the "forbidden" region is seen as occupied and hook it up to
the costmap. This will keep the LIDAR from clearing it out when the robot is
close to it. It isn't quite as easy as drawing on a bitmap, but it should
solve your problem. You can also have that sensor operate at a different
height than your LIDAR in the voxel grid version of the costmap which means
you'll only have to publish information about forbidden regions on startup.
If you do create the sensor that reads in maps to create forbidden zones, it
might be nice to share it with the ROS community, as the cost value fix
likely won't be released until D-Turtle.

Come to think of it, the sensor approach is actually much better for dealing
with the local planner's costmap which has no knowledge of the global static
map. I'll have to think a little on the best way to deal with that with the
special cost value approach.

Hope this helps,

Eitan

On Mon, Sep 6, 2010 at 10:41 AM, Eric Perko <wisesage5001 at gmail.com> wrote:

> Christian,
>
> I did exactly what you are trying to do just the other day and was able to
> successfully get it to run with one map for AMCL and another for the
> navigation stack. Launch file attached.
>
> We then ran into a separate problem, where the LIDAR in the navigation
> stack will clear out our static obstacles if it cannot sense them (glass
> walls or your 'forbidden zones'), as the static map seems to only be used to
> seed the costmaps, not continually make sure that they contain at least the
> static map. It also does not get assigned a height when using the voxel
> grids, so we were unable to just separate the static map and LIDAR into two
> different planes to prevent cross-clearing. Have you encountered this
> problem with LIDARs clearing your static_map forbidden zones? Also, do you
> load the static map into the costmap for the local_planner as well and if
> not, how do you give the local_planner knowledge of the forbidden zones? We
> are probably going to address these by creating a PointCloud sensor that
> reads in maps and can then be fed into the costmaps just like any other
> sensor. I'm interested to see if you came up with a different plan.
>
> - Eric
>
>
> On Mon, Sep 6, 2010 at 12:26 PM, Christian Verbeek <
> verbeek at servicerobotics.eu> wrote:
>
>>
>> > So you can remap map to planner_map in the launch file by doing
>> > <remap from="map" to="planner_map"/>
>> >
>> Adam,
>>
>> I am using the move_base node. Remapping would then also apply to amcl.
>>
>> _______________________________________________
>> ros-users mailing list
>> ros-users at code.ros.org
>> https://code.ros.org/mailman/listinfo/ros-users
>>
>
>
> _______________________________________________
> 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/20100906/f70bdadd/attachment-0003.html>


More information about the ros-users mailing list