iCub3 masses updated

[ale-git]

This PR solves the real iCub3 Vs. iCub3 URDF model mass discrepancy. The real robot weights about 52 kg on the scale, while the mass calculated by the CAD - and thus exported in the model - is about 54.5 kg after the last debugging (hidden stray component removed, other missing component added).

Thus, the ICUB_3_all_options_gazebo.yaml.in has been modified in order to assign the part masses rescaled by a 52.010/54.992 factor.

[pattacini]

cc @S-Dafarra @salvi-mattia @mfussi66

[traversaro]

Hi @ale-git @salvi-mattia @mfussi66 , thanks a lot for the PR!

For better tracking this kind of models, could you report the cad-mechanics checkout from which the SimMechanics model was created? Thanks!

[pattacini]

@ale-git, better off referring to an issue in a public repo like robotology/icub-models#94 in the commit message, instead of pointing to our internal workflow.

[traversaro]

In the past, whenever a new model was regenerated we had tipically had regression due to the direction of the axis that changed, requiring then a change in the reverseRotationAxis parameters to be coherent with the axis direction expected in YARP (see robotology/icub-models#70, #184).

Did you happen to check that?

[traversaro]

Did you happen to check that?

Actually, this may be checked automatically thanks to #193, let's see the CI response.

[ale-git]

@ale-git, better off referring an issue in a public repo like robotology/icub-models#94, instead of pointing to our internal workflow.

I din't consider that the commit on my repo would have sneaked till here...

[traversaro]

If you want to inspect the differences in the generated model, you can check the log of CI.

[Nicogene]

Probably the changes done in ICUB_3_all_options_gazebo.yaml.in has to be done also in ICUB_3_all_options.yaml, because this fix is not gazebo-specific, what do you think about it @traversaro ?

[traversaro]

@pattacini pattacini requested review from Nicogene and traversaro 7 minutes ago

We can't review the PR until it is out of Draft state, there is anything missing?

[traversaro]

@pattacini pattacini requested review from Nicogene and traversaro 7 minutes ago

We can't review the PR until it is out of Draft state, there is anything missing?

Actually I was wrong, we can review even if in Draft. However, what is missing? From my point of view the PR is fine, the only request I have is the one in #207 (comment) .

[pattacini]

@traversaro, the PR is ok for being reviewed 👍🏻

[mfussi66]

For better tracking this kind of models, could you report the cad-mechanics checkout from which the SimMechanics model was created? Thanks!

@traversaro If I understood the question correctly, cad-mechanics-public master was used, specifically, up to the commit: icub-tech-iit/cad-mechanics-public@b176b2f

[traversaro]

For better tracking this kind of models, could you report the cad-mechanics checkout from which the SimMechanics model was created? Thanks!

@traversaro If I understood the question correctly, cad-mechanics-public master was used, specifically, up to the commit: icub-tech-iit/cad-mechanics-public@b176b2f

Great, thanks, this is what I needed.

[traversaro]

Probably the changes done in ICUB_3_all_options_gazebo.yaml.in has to be done also in ICUB_3_all_options.yaml, because this fix is not gazebo-specific, what do you think about it @traversaro ?

Sure, good catch!

[ale-git]

Probably the changes done in ICUB_3_all_options_gazebo.yaml.in has to be done also in ICUB_3_all_options.yaml, because this fix is not gazebo-specific, what do you think about it @traversaro ?

Sure, good catch!

Ok, I'm going to copy-paste the section.

[traversaro]

Probably the changes done in ICUB_3_all_options_gazebo.yaml.in has to be done also in ICUB_3_all_options.yaml, because this fix is not gazebo-specific, what do you think about it @traversaro ?

Sure, good catch!

By the way, this is a nice example on why relying on modified models to support quirks of specific physics engines (such as Gazebo/ODE, that has numerical errors with its default simulation parameters with small inertias that are otherwise perfectly valid) is not robusts, and the reason for which relying instead on simulation softare-specific plugins that modified the model as required during loading, like the one discussed in robotology/icub-models#33 would be useful.