The model used in gazebo seems to be different from the Genova04 model.

[georgwi]

When running a gravity compensation test with the gazebo simulation and the yarp bridge to the IHMC software I found that the torques I sent to the robot do not work well. I looked briefly at the sdf files and it seems like the gazebo simulation uses a pretty different robot model. Is that true? If so, what would be the best way to fix this inconsistency?

The chest in the Genova04 model that we use:

    <link name='chest'>
      <pose frame=''>0.005498 -0 0.082 1.5708 -0 1.5708</pose>
      <inertial>
        <pose frame=''>0.000189 0.073091 -0.047238 0 -0 0</pose>
        <mass>7.638</mass>

And the chest in the icub sdf from the package icub-gazebo

    <link name='chest'>
      <pose>-0.00549988 1.45945e-08 0.032 1.5708 -7.34642e-06 -3.14159</pose>
      <inertial>
        <pose>0.00222898 0.075 -1.18362e-17 0 -0 0</pose>
        <mass>4.81</mass>

I might be looking at the wrong files though.

[fiorisi]

Hello @georgwi, we are working to finalize the new model of iCub 2.5+ (IMU on the root link and high heel shoes).
The model in the icub-gazebo repository is not updated.
Since the new model is still WIP, we created a temporary repository that you can use in the same way you use icub-gazebo. Can you try to clone ad use the model in the repository https://github.com/fiorisi/icub-gazebo?
Let us know if you experience any issue, consider that also this repository is WIP.

[georgwi]

@fiorisi
This seems to work fine - I updated the model we use on our side as well as the gazebo model with the one from the repository you linked to. I now can run the gravity compensation fairly well. We should consider using the new model for the IHMC iCub package instead of the genova04 one once it is ready.
Thanks!