Jondolf
commented
3 weeks ago Another issue noticed by @tim-blackbird: friction causes some bounce for a box with locked rotation and zero restitution. Increasing the friction makes the problem worse, even without locked axes.
https://github.com/user-attachments/assets/ca9fd53e-bebc-4d53-834f-c6d081803185
The current 3D friction implementation seems to be incorrect. As can be seen in the Gif in the README's "Usage example", the box stays essentially at the same spot despite dropping onto a corner.
Intuitively, you would expect the box to tumble in some direction based on which corner hit the ground and how. In fact, this is what happens with a similar setup using Rapier.
Notably, our current behavior can be roughly replicated in Rapier with a friction of zero. It almost looks like our friction only kicks in once the box is flat on the ground and the contact points are clearly sliding along the surface.
A similar setup in 2D does seem to work correctly.