RobotLocomotion / drake

Model-based design and verification for robotics.
https://drake.mit.edu
Other
3.35k stars 1.27k forks source link

Deprecation removal contact_solver #21225

Open jwnimmer-tri opened 7 months ago

jwnimmer-tri commented 7 months ago

The MbP has some deprecations related to the contact solver:

https://github.com/RobotLocomotion/drake/blob/f8f957f0817bcfc8d789452e1a80e6348d545f5f/multibody/plant/multibody_plant.h#L2235-L2238

https://github.com/RobotLocomotion/drake/blob/f8f957f0817bcfc8d789452e1a80e6348d545f5f/multibody/plant/multibody_plant_config.h#L64-L73

Normally I'd just remove the deprecated code myself, but I can't this time because I don't have the "information we collect from users".

Please remove the deprecated code / comments and adjust the nearby code as-necessary.

jwnimmer-tri commented 7 months ago

I think probably this ticket is holding up #21095, so I hope to have us resolve this sooner than later.

@amcastro-tri As of today, which of "similar" or "lagged" (or "sap?) is our best guess at the most useful default value? I can open the PR if you like, I just don't know which one to choose.

amcastro-tri commented 7 months ago

Per f2f: running kLagged would be ideal in that it does not introduce artifacts results of the convexification (i.e. "gliding" when sliding). However, kLagged introduced a "weak" coupling of friction and normal forces that might impact the perfomance (meaning predicted success/failure) of grasping tasks in simulation. In this regard, kSimilar is the safe bet, in that it solves most of the big issues we had with kSap and introduces Hunt & Crossley for the accurate and stable modeling of dissipation.

See #21294, thus far a WIP until we confirm all tests pass. I run a few tests locally and they succeded. Lets wait for Drake CI and Anzu CI for the final veredict.

jwnimmer-tri commented 2 weeks ago

@amcastro-tri Would you like to add this to our project board, for yourself?