search

robotology / human-gazebo

URDF models of humans created to perform human robot interaction experiments.
GNU Lesser General Public License v2.1
136 stars 40 forks source link

Fix hand links mass and inertia inconsistencies #12

Closed claudia-lat closed 4 years ago

claudia-lat commented 4 years ago

Issue https://github.com/robotology/human-gazebo/issues/11#issuecomment-603420750 revealed a bug in all the models stored in this repo. Models have to be regenerated by following these instructions https://github.com/dic-iit/human-model-generator#how-to-generate-models.

yeshasvitirupachuri commented 4 years ago

I generated the new models using the https://github.com/dic-iit/human-model-generator#how-to-generate-models, however I notice that all the fake links mass and inertia values are not zero, as they are in the models currently present in human-gazebo repo. The main difference seems to be in the mass and inertia values of the RightHand link, which is different from the LeftHand.

Screenshot from 2020-04-02 15-14-56

Screenshot from 2020-04-02 15-15-19

So, I think it is best if the RightHand link mass and inertia values are set similar to the LeftHand link in all the models and we correct them manually rather than regenerating all the models again.

@claudia-lat @lrapetti @DanielePucci

claudia-lat commented 4 years ago

@Yeshasvitvs the fake links in the new models do not have mass values equal to zero because the code creates fake links with small masses in order to be visualized in Gazebo https://github.com/dic-iit/human-model-generator/blob/master/human_23links/src/createXsensLikeURDFmodel.m#L290

yeshasvitirupachuri commented 4 years ago

@claudia-lat yes I understand and that is the reason I want to avoid the regeneration of models. If we regenerate, these new models may cause some other issues down the line in some code which assumed zero mass and inertial values for the links. So, instead we correct just the RightHand link values similar to the LeftHand link values of mass and inertia

claudia-lat commented 4 years ago

@Yeshasvitvs if you want models with zero mass for the fake links, launch the code with option false here https://github.com/dic-iit/human-model-generator/blob/master/human_23links/computeHumanURDF.m#L145

yeshasvitirupachuri commented 4 years ago

A PR is open to address this issue https://github.com/robotology/human-gazebo/pull/13

Further I double checked all the models with the test script from #11

@claudia-lat @lrapetti @DanielePucci