search

andreadelprete / consim

GNU General Public License v3.0
14 stars 1 forks source link

Potential influcence of the cone projection #24

Closed jcarpent closed 4 years ago

jcarpent commented 4 years ago

The following lines are performing a projection on the friction cone: https://github.com/andreadelprete/consim/blob/c1d6b07ca4a7ae8ef64e70531ac32addd6fd8f4c/src/contact.cpp#L109-L122 but this does not correspond to an orthogonal projection. This might lead to a bias in the solution. I don't know how much this may influence the final solution, but for sure, if you want to use this step inside a gradient descent framework, it might lead to a feasible but not optimal solution.

andreadelprete commented 4 years ago

The idea here was to project the tangential force on the cone, without modifying the normal force. Considering the tangential force only, I think the projection is orthogonal to the cone. However, I agree that considering the 3d force the projection is not orthogonal to the cone, but I think that physically speaking this is sound.

jcarpent commented 4 years ago

Yes, I understand the idea. But if would use it inside an SOCP solver, for instance, this projection will lead to bad behaviour, nothing more to say I would say. But I guess this is not the objective here.