search

robotology / icub-models

Official URDF and SDF models of the iCub humanoid robot.
Creative Commons Attribution Share Alike 4.0 International
33 stars 34 forks source link

iCubGazeboV2_* has `r_wrist_yaw` inverted #156

Closed Nicogene closed 2 years ago

Nicogene commented 2 years ago

As you can see from these videos, the rotation of r_wrist_yaw is the opposite w.r.t l_wrist_yaw:

https://user-images.githubusercontent.com/19833605/169014610-fa15450f-636d-41b3-8234-daf6549a2e41.mp4

On the other hand, on iCubGazeboV3, iCubGazeboV2_5_visuomanip and on the real robot left and right are moving accordingly:

iCubGazeboV3:

https://user-images.githubusercontent.com/19833605/169014672-a2e042c7-4540-4d8e-9d36-ca406b0c1e07.mp4

iCubGazeboV2_5_visuomanip:

https://user-images.githubusercontent.com/19152494/169041245-c7c3164e-9f62-4b86-916b-ee90ec0e94f9.mp4

traversaro commented 2 years ago

Thanks @Nicogene . I think this is an issue similar to https://github.com/robotology/icub-models/issues/70 , but for iCubV2 . Whenever the Simmechanics model is regenerated from the Creo model, the actual direction of the exported axis is non-deterministic (from what we are able to understand in this years), and so the axis direction should be carefully checked and in case it is necessary we should update the reverseRotationAxis (https://github.com/robotology/icub-models-generator/blob/872c9bddaa33d1d3010ef1e63eb98760ebd9696d/simmechanics/CMakeLists.txt#L191). Apparently this went unnoticed, even because the tests for iCubV2 only involve the torso axis (see https://github.com/robotology/icub-models-generator/blob/0b244da996b1f0f308e6596d7e156c176e79c918/tests/icub-model-test.cpp#L227).

By looking a bit at the history of generation, the last generation of the iCubV2 model was back in https://github.com/robotology/icub-models-generator/pull/122, so probably this bug was there in the last ~2.5 years, i.e. since icub-models 1.13.0 https://github.com/robotology/icub-models/releases/tag/v1.13.0 .

traversaro commented 2 years ago

Mentioning @robotology/iit-artificial-mechanical-intelligence as it can be of interested to anyone working with iCub**V2** models: on those models the r_wrist_yaw joint has the wrong direction of motion since 2019.

traversaro commented 2 years ago

The fix should be just to update the reverseRotationAxis option, volunteers to work on this are welcome. It would be great to also update the tests to detect regression of this kind, to avoid problems like that in the future.

Nicogene commented 2 years ago

I can take this issue and work on it on spare time since it should not be complicated 👍🏻

(Maybe also @marcello-goccia)

traversaro commented 2 years ago

Great, thanks!