Add model tests to check symmetry in geometry and intertial parameters
yeshasvitirupachuri opened this issue · 4 comments
In some recent tests related to human-dynamics-estimation, we suspected there may be some inconsistencies in the model geometry and inertial parameters. So, we decided to add a model test script that will assist in assessing the human models saved in this repo. A matlab script model_test.m is available at feature/model-test.
@prashanthr05 @lrapetti @claudia-lat @traversaro @kouroshD @DanielePucci
I added inertial parameters check d3c8f13, and there seems to a mismatch between the LeftHand and RightHand for all the models
Subject 01
Warning: Mismatch in mass
Warning: LeftHand link mass is 0.373200 Kgs
Warning: RightHand link mass is 1.033200 Kgs
Warning: Mismatch in center of mass position
Warning: LeftHand link center of mass is (0.000000, 0.031195, 0.000000)
Warning: RightHand link center of mass is (0.000000, -0.086363, 0.000000)
Warning: Mismatch in inertia
Warning: LeftHand link inertia elements are (0.003525, 0.000000, 0.000000, 0.000434, 0.000000, 0.003863)
Warning: RightHand link inertia elements are (0.009758, 0.000000, 0.000000, 0.001202, 0.000000, 0.010695)
Subject 02
Warning: Mismatch in mass
Warning: LeftHand link mass is 0.476400 Kgs
Warning: RightHand link mass is 1.136400 Kgs
Warning: Mismatch in center of mass position
Warning: LeftHand link center of mass is (0.000000, 0.040975, 0.000000)
Warning: RightHand link center of mass is (0.000000, -0.097741, 0.000000)
Warning: Mismatch in inertia
Warning: LeftHand link inertia elements are (0.004762, 0.000000, 0.000000, 0.000586, 0.000000, 0.005221)
Warning: RightHand link inertia elements are (0.011360, 0.000000, 0.000000, 0.001397, 0.000000, 0.012454)
Subject 05
Warning: Mismatch in mass
Warning: LeftHand link mass is 0.330000 Kgs
Warning: RightHand link mass is 0.990000 Kgs
Warning: Mismatch in center of mass position
Warning: LeftHand link center of mass is (0.000000, 0.028196, 0.000000)
Warning: RightHand link center of mass is (0.000000, -0.084587, 0.000000)
Warning: Mismatch in inertia
Warning: LeftHand link inertia elements are (0.003256, 0.000000, 0.000000, 0.000401, 0.000000, 0.003569)
Warning: RightHand link inertia elements are (0.009769, 0.000000, 0.000000, 0.001204, 0.000000, 0.010707)
Subject 08
Warning: Mismatch in mass
Warning: LeftHand link mass is 0.331200 Kgs
Warning: RightHand link mass is 0.991200 Kgs
Warning: Mismatch in center of mass position
Warning: LeftHand link center of mass is (0.000000, 0.028724, 0.000000)
Warning: RightHand link center of mass is (0.000000, -0.085964, 0.000000)
Warning: Mismatch in inertia
Warning: LeftHand link inertia elements are (0.003368, 0.000000, 0.000000, 0.000415, 0.000000, 0.003691)
Warning: RightHand link inertia elements are (0.010079, 0.000000, 0.000000, 0.001243, 0.000000, 0.011045)
I believe there is an error in the urdf template @claudia-lat
This explains the offset of the model center of mass offset in the y direction during neutral T pose
@prashanthr05 @lrapetti @claudia-lat @traversaro @kouroshD @DanielePucci
As a quick test, I set the mass and inertial values of the RightHand to be the same as the LeftHand in humanSubject02 model and the center of mass y value is 0
Yes @Yeshasvitvs you're right! They are wrong. By construction, the code in https://github.com/dic-iit/human-model-generator/tree/master/human_23links builds symmetric hands:
- see https://github.com/dic-iit/human-model-generator/blob/master/human_23links/src/subjectParamsComputation.m#L230
- see https://github.com/dic-iit/human-model-generator/blob/master/human_23links/src/subjectParamsComputation.m#L304
You are using the models stored in May, they for sure have been generated with an error. I would suggest you to generate the models again (for all the subjects) and overwrite the wrong ones.
Run the script https://github.com/dic-iit/human-model-generator/blob/master/human_23links/computeHumanURDF.m by having a data folder structured as described in the header of the script itself.
@claudia-lat please open an issue to create the new models in this repo. We can close this issue since we found the problem.