search

robotology / walking-controllers

Bipedal locomotion software for the humanoid robot platform iCub.
BSD 3-Clause "New" or "Revised" License
121 stars 46 forks source link

Tune the low level gains of the iCubGazeboV3 #106

Closed mebbaid closed 2 years ago

mebbaid commented 2 years ago

after yesterdays issues during the meeting, now with a clean build, using master or devel, and with those pidparams, it seems the simulation is more reproducible and stable than before (see attached video)

https://user-images.githubusercontent.com/36491318/153573524-4c82d5c2-8ec1-4c3c-a52f-97926c7b89c6.MP4

.

S-Dafarra commented 2 years ago

Actually, I think that we can target directly master, right @GiulioRomualdi?

GiulioRomualdi commented 2 years ago

Nice! At this point isnt better to commits those gains directly in icub models?

mebbaid commented 2 years ago

Nice! At this point isn't better to commits those gains directly in icub models?

Ok, if you think it is better to commit there then sure. The guys from HSP were installing everything individually and not through the superbuild, so that makes sense.

mebbaid commented 2 years ago

Nice! At this point isnt better to commits those gains directly in icub models?

Let me know if this PR is corresponding to what you suggested. In which case we can merge/close this one perhaps.

GiulioRomualdi commented 2 years ago

Hi @mebbaid did you open a PR in icub-models-generator as suggested by @traversaro in https://github.com/robotology/icub-models/pull/126?

Can I close this PR?

mebbaid commented 2 years ago

Hi @GiulioRomualdi

The suggested PR is not opened yet. The halt is due to me not being able to verify the same good behaviour in a different machine used by the HSP team. So I am spending sometime tuning and was going to close once the other PR is opened.

mebbaid commented 2 years ago

Opened a corresponding PR in icub-model-generators , see this PR.

Closing

Hi @GiulioRomualdi

The suggested PR is not opened yet. The halt is due to me not being able to verify the same good behaviour in a different machine used by the HSP team. So I am spending sometime tuning and was going to close once the other PR is opened.