wrong joint limits in generated files

[andreadelprete]

It seems that the joint limits for the two hip-roll joints are wrong. See for instance the left-hip-roll joint of generated/gazebo_models/iCubGenova03/iCubGenova03.sdf:

<joint name='l_hip_roll' type='revolute'>
      <child>l_hip_2</child>
      <parent>l_hip_1</parent>
      <axis>
        <xyz>-1 2.65359e-06 7.34641e-06</xyz>
        <limit>
          <lower>-2.07694</lower>
          <upper>0.296706</upper>
          <effort>84</effort>
          <velocity>100</velocity>
        </limit>
        <dynamics>
          <damping>0.1</damping>
        </dynamics>
        <use_parent_model_frame>1</use_parent_model_frame>
      </axis>
    </joint>

and compare the joint limits with the one of the model of the model /icub/icub.sdf in the icub_gazebo repository:

        <joint name='l_hip_roll' type='revolute'>
            <child>l_hip_2</child>
            <parent>l_hip_1</parent>
            <axis>
                <xyz>-1 5.30716e-06 1.26536e-05</xyz>
                <limit>
                    <lower>-0.34</lower>
                    <upper>1.6707963268</upper>
                    <effort>10000</effort>
                    <velocity>100</velocity>
                </limit>
                <dynamics>
                    <damping>0.1</damping>
                </dynamics>
            </axis>
        </joint>

The problem is the same for the right-hip roll and for the iCubGenova01 model.

[andreadelprete]

Also the ankle-pitch limits are significantly different. For the left-ankle pitch icub.sdf has:

<lower>-0.4490658504</lower>
<upper>0.8679448709</upper>

whereas iCubGenova03.sdf has:

          <lower>-0.733038</lower>
          <upper>0.366519</upper>

[traversaro]

cc @luca-fiorio @francesco-romano @jeljaik

[jeljaik]

Yes, we noticed just yesterday.

Jorhabib Eljaik Gómez,*Ph.D. Student - *Istituto Italiano di Tecnologia Dept. of Robotics, Brain and Cognitive Sciences Cognitive Humanoids Laboratory e-mail: [email protected] Cel: +39 388 352 33 85 Tel: +39 010 71781 420 Fax: +39 010 7170817

On Wed, Dec 10, 2014 at 10:20 AM, Silvio Traversaro < [email protected]> wrote:

cc @luca-fiorio https://github.com/luca-fiorio @francesco-romano https://github.com/francesco-romano @jeljaik https://github.com/jeljaik

— Reply to this email directly or view it on GitHub https://github.com/robotology-playground/icub-model-generator/issues/5#issuecomment-66423814 .

[iron76]

In my opinion it is not an easy task to define span [j_min, j_max] for all joints. It might be feasible only for some. There are situations in which the range of a joint depends on previous and therefore the configuration space is quite a complex manifold. If we really want something like [j_min, j_max], we can try to list some possibilities:

• "maximum span, considering all previous joint configurations": it might be too restrictive but it might also prevent the naive user from damaging the robot.
• "maximum span, regardless of previous joint configurations": it requires some other mechanism for checking collision avoidance (something like http://eris.liralab.it/wiki/MoBeE_installation).

[traversaro]

Ref: https://github.com/robotology-playground/icub-gazebo/pull/27 .

[traversaro]

This is a general instance of the problem of getting correct joint limits.

[traversaro]

cc @S-Dafarra