Wrong inertia properties extracted from data class using python interface.

[TJBetter]

Hello all,

Thanks for this making this great repo public, we really like it.

I wonder if you could clarify what does the class data contains? We are using the latest version, and we are trinh to extract the inertia properties from the data class from the cheatsheet (here):

def main():

    urdf_filename = ('2d_planar.urdf')
    # Load the urdf model
    model = pinocchio.buildModelFromUrdf(urdf_filename)

    data = model.createData()
    print('data mass is {}'.format(list(data.mass)))

if __name__ == '__main__':
    main()

But the above code gives me data mass is [-1.0, -1.0, -1.0], which doesn't match the urdf file:

 <?xml version="1.0" ?>

<robot name="myexample" xmlns:xacro="http://www.ros.org/wiki/xacro">
    <link name="base_link">
        <visual>
            <origin rpy="0 0 0" xyz="0 0 10"/>
            <geometry>
                <box size="1 1 1"/>
            </geometry>
        </visual>
        <inertial>
          <mass value="1"/>
          <origin rpy="0 0 0" xyz="0.0 0.0 0.121825"/>
          <inertia ixx="0.0217284832211" ixy="0.1" ixz="0.0" iyy="0.0217284832211" iyz="0.0" izz="0.00961875"/>
        </inertial>
    </link>

    <joint name="joint_0" type="revolute">
        <axis xyz="1 0 0" />
        <limit effort="1000.0" lower="-3.14" upper="3.14" velocity="0.5" />
        <origin rpy="0 0 0" xyz="0 0 0"/>
        <parent link="base_link"/>
        <child link="link1"/>
    </joint>

    <link name="link1">
        <visual>
                <origin rpy="0 0 0" xyz="0 0 0.0"/>
                <geometry>
                    <cylinder radius="0.35" length="0.4"/>
                </geometry>
        </visual>
        <inertial>
          <mass value="2"/>
          <origin rpy="0 0 0" xyz="0.0 0.0 0.121825"/>
          <inertia ixx="0.0217284832211" ixy="0.2" ixz="0.0" iyy="0.0217284832211" iyz="0.0" izz="0.00961875"/>
        </inertial>
    </link>

    <joint name="joint_1" type="revolute">
        <axis xyz="1 0 0" />
        <limit effort="1000.0" lower="-3.14" upper="3.14" velocity="0.5" />
        <origin rpy="0 0 0" xyz="0 0 1.0"/>
        <parent link="link1"/>
        <child link="link2"/>
    </joint>

    <link name="link2">
        <visual>
            <origin rpy="0 0 0" xyz="0 0 0.2"/>
                <geometry>
                    <cylinder radius="0.35" length="0.4"/>
                </geometry>
        </visual>
        <inertial>
          <mass value="3.42"/>
          <origin rpy="0 0 0" xyz="0.0 0.0 0.121825"/>
          <inertia ixx="0.0217284832211" ixy="0.3" ixz="0.0" iyy="0.0217284832211" iyz="0.0" izz="0.00961875"/>
        </inertial>
    </link>
</robot>

In addition, I wonder if you may know if there is a python api for this great repo? Thank you.

[fabinsch]

Hi @TJBetter, as you did not call any algorithm this behavior is expected. If you just want to get the mass, you could call

In [2]: pin.computeSubtreeMasses?
Docstring:
computeSubtreeMasses( (Model)model, (Data)data) -> None :
    Compute the mass of each kinematic subtree and store it in the vector data.mass.
Type:      function

and this will fill data.mass, see Section Center of Mass on the cheat sheet.

[TJBetter]

Hello @fabinsch,

Thank you for getting back to me. I understand I can get the mass or inertia by calling other functions.

I'm trying to understand when the member variables of a data class are updated accurately so that I can correctly apply the formulas provided in the Data section of the cheatsheet. The above code is just a simple example that shows how easy it is to get incorrect info from data class. As a beginner to this library, I wonder if you could provide more guidelines about using data?

Thank you.

[jcarpent]

A quick reply is that Data stores intermediate results for common algorithms in such a way that:

1. they are accessible to the end users
2. we avoid recomputing twice the same quantities
3. to avoid useless online memory allocations for efficiency.