Christian,

The persistence parameter keeps the obstacle in the costmap for X seconds no
matter what clearing information is observed, and the marking sensor always
takes precedence. If you've found the costmap behaves otherwise... its a
bug, and you can file a ticket against me.

Hope all is well,

Eitan

On Mon, Jul 26, 2010 at 10:57 AM, Christian Verbeek <
verbeek@servicerobotics.eu> wrote:

>  Dear Eitan,
>
> Then the next question is: Who is stronger. The marking or the clearing
> sensor? When my IR sees an obstacle and I set observation_persistence=10 I
> would expect the obstacle to be in the costmap for at least 10s. Or does the
> clearing sensor remove the obstacle as soon as the dense point cloud is
> published? The same question applies to the laser clearing IR readings.
>
> Regards
> Christian
>
> Am 26.07.2010 19:29, schrieb Eitan Marder-Eppstein:
>
> Christian,
>
> The reason that your IR sensors do not clear out of the costmap is that
> they are at a different height than the laser and you've chosen (implicitly
> by not setting the "map_type" parameter) to use a 3D voxel grid as your
> underlying map structure. So, the laser will never clear out IR hits because
> it lies strictly above them. Your IR sensors can clear out old hits, but
> with only 9 points to raytrace to... its quite likely that many old hits
> will not be seen through and thus are not safe to clear.
>
> You have two options to clear these hits:
>
> 1) If you want your laser to be able to clear out hits from your IR sensors
> once they are no longer observed, you could switch to the "map_type"
> parameter to use "costmap": see
> http://www.ros.org/wiki/costmap_2d#Map_type_parameters. This will use a 2D
> representation of the world where everything is assumed to be on the same
> plane.
>
> 2) You can continue using the 3D voxel grid and write your own sensor that
> publishes a dense point cloud at the height of the IR sensors and is only
> used for clearing. This sensor could publish the cloud at whatever rate you
> want your IR hits to clear out.
>
> Both of these options aren't strictly safe, consider the case where you see
> an object with IR, move a little bit, clear that object out, don't see it
> again, and hit it, but should probably help you do what you want.
>
> Hope this helps,
>
> Eitan
>
> On Mon, Jul 26, 2010 at 6:34 AM, Christian Verbeek <
> verbeek@servicerobotics.eu> wrote:
>
>> Dear all,
>>
>> my robot is equipped with a laser rangefinder and 9 infrared distances
>> sensors. The rangefinder is scanning at z=0.2 while the infrared sensors
>> are scanning at z=0.05. So it absolutely makes sense to put the readings
>> of the infrared sensors into the local costmap to avoid obstacles not
>> seen by the rangefinder.
>>
>> I publish a point cloud with 9 points in it (the distances read from the
>> 9 infrared sensors mapped to the base frame). The maximum range of the
>> infrared sensors is 0.4m.
>>
>> My costmap_common_params.yaml look like this
>>
>> observation_sources: laser_scan_sensor ir_sensors
>>
>> laser_scan_sensor: {sensor_frame: laser_link, data_type: LaserScan,
>> topic: scan, marking: true, clearing: true}
>> ir_sensors: {sensor_frame: base_link, data_type: PointCloud, topic:
>> ir_sensors, marking: true, clearing: true, obstacle_range: 0.4,
>> raytrace_range: 0.4, observation_persistence: 0.0 }
>>
>> The obstacles detected by the infrared sensors are correctly put into
>> the costmap. The problem is that the marker cells aren't cleared. So
>> when I move around my robot the robot is surrounded by occupied cells
>> and path planning does not work any more.
>>
>> What can I do?
>>
>> Regards
>> Christian
>>
>>
>> _______________________________________________
>> ros-users mailing list
>> ros-users@code.ros.org
>> https://code.ros.org/mailman/listinfo/ros-users
>>
>
>
> _______________________________________________
> ros-users mailing listros-users@code.ros.orghttps://code.ros.org/mailman/listinfo/ros-users
>
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
>