feat: apply ground friction on predictive velocity correction too #7
Conversation
|
Noticeable improvement in the 2D examples. The material doesn't seem to slide as on ice, see the screenshots below. Before: After: |
|
Hmm, it does look like the fluid example is worse off as the number of escaping particles is increased. I'm not sure why that is. |
| @@ -125,25 +126,29 @@ fn update_single_cell( | |||
|
|
|||
| if apply_friction { | |||
There was a problem hiding this comment.
This could be a better name
| if apply_friction { | |
| if apply_friction_and_cancel_penetration { |
| let tangent_vel = cell_velocity - normal_vel * normal.into_inner(); | ||
| let tangent_vel_norm = tangent_vel.norm(); | ||
|
|
||
| let new_normal_vel = normal_vel.max(-dist_with_margin / dt); |
There was a problem hiding this comment.
Even if we zero the normal velocity out completely, particles can still approach the boundary as their non-zero-kernel space goes further. We may want to extend the margin s.t. we cover the whole range of an approaching particle.
There was a problem hiding this comment.
This amounts to changing the line above to let dist_with_margin = (dist - cell_width * 1.5).max(0.0);.
This also restores the previous behavior in the fluid example btw.
Before this change, ground friction was only applied on grid cells below the ground. Now, just like grid cells below the ground induce a friction impulse proportional to the normal velocity, the grid cells above the ground now apply friction proportional to the removed normal velocity due to the predictive contact.