Per link padding and scaling not working about robot_body_filter

[peci1]

Please, provide also the urdf of your robot

…

On Fri, Jan 21, 2022, 21:39 Juan Solana ***@***.***> wrote: Hi, I tried applying a per link scaling and padding but does not seem to be working. Do I need to reinstall the library or is something wrong with my yaml file? # Filter config for segmenting out Robot from pointcloud. # The field of view of the sensor can only capture a few links, so the `only_links` parameter is set # to increase performance. Also, the `filter/keep_clouds_organized` parameter is important here to # retain the 2D structure of the RGBD camera output. # For more detail on the different parameters of the filter, refer to: # http://wiki.ros.org/robot_body_filter robot_filter_chain: - name: RobotBodyFilter type: robot_body_filter/RobotBodyFilterPointCloud2 params: frames/fixed: 'canonical_smpl_origin' frames/sensor: 'canonical_smpl_origin' filter/keep_clouds_organized: True filter/do_clipping: True filter/do_contains_test: True filter/do_shadow_test: True sensor/point_by_point: False sensor/min_distance: 0.01 sensor/max_distance: 2.5 only_links: [] ignored_links/everywhere: [] body_model/inflation/scale: 1.2 body_model/inflation/per_link/padding: {left_touchpoint: 0.072, right_touchpoint: 0.072} body_model/robot_description_param: '/robot_description' transforms/buffer_length: 15.0 transforms/timeout/reachable: 0.2 transforms/timeout/unreachable: 0.2 bounding_sphere/compute: True bounding_sphere/debug: False bounding_sphere/publish_cut_out_pointcloud: True bounding_box/compute: True bounding_box/debug: False bounding_box/publish_cut_out_pointcloud: True debug/pcl/inside: False debug/pcl/clip: False debug/pcl/shadow: False debug/marker/contains: True debug/marker/shadow: False — Reply to this email directly, view it on GitHub <#19>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AABMSBI2473L2TAYB7QQ5KTUXG77VANCNFSM5MQP6M5Q> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[juansolana]

This is the urdf of the robot. The collisions within ${arm_id}_touchpoint are the ones I'm trying to apply the filter config of padding and scaling.

`

<xacro:include filename="$(find franka_description)/robots/panda_arm.xacro"/>

<xacro:macro name="arm" params="arm_id:='panda'">

<xacro:panda_arm arm_id="${arm_id}_arm" connected_to="${arm_id}_arm_base_link" xyz="0 0 0" rpy="0 0 0"/>

<!-- Touchpoints -->
<link name="${arm_id}_touchpoint">
  <visual>
    <geometry>
      <mesh filename="package://description/meshes/touchpoints/${arm_id}_tp_100.stl"/>
    </geometry>
    <material name="blue">
      <color rgba="0 0  .8 1"/>
    </material>
    <origin rpy="0 0 0" xyz="0 0 0.021"/>  <!-- Millibar Offset -->
  </visual>
  <collision>
    <geometry>
      <cylinder length="0.2" radius="0.05" />
    </geometry>
    <origin rpy="55 90 0" xyz="0 0 0.05"/>
    <geometry>
      <mesh filename="package://description/meshes/touchpoints/${arm_id}_tp_100.stl" scale="1.05"/>
    </geometry>
    <origin rpy="0 0 0" xyz="0 0 0.021"/>
  </collision>
</link>

<joint name="${arm_id}_touchpoint_joint" type="fixed">
  <parent link="${arm_id}_arm_link8"/>
  <child link="${arm_id}_touchpoint"/>
</joint>

</xacro:macro>

<xacro:arm arm_id="right"/>
<xacro:arm arm_id="left"/>

`

[peci1]

The <collision> you posted is weird. It contains two <geometry> nodes, while I'm almost sure URDF only allows one geometry per collision. You can specify multiple collisions, each with a single geometry node.

[juansolana]

I guess that explains why only see the cylinder collision geometry in RViz. You think having both collision is what makes the per link scale and padding not work? The normal way that applies to all collisions works fine, and I can tell that the scaling and padding is not applied to the second collision that doesn't even show up.

[peci1]

That's my guess. As you're using invalid URDF, many things can break and this could be one of them. Try fixing the URDF and let me know if it fixed the issue.

[peci1]

Closing this issue because there is probably not much this library can do with invalid inputs.