Hi to all,
I am developing a custom collision algorithm for deformable meshes.
I developed Dbvt colliders similar to the btSoftColliders::CollideSDF_RS.
Sometimes, the collision algorithm crashes because the Evaluate function of the collider compute a null normal vector.
Any hints?
Thank you in advance,
Francesco
[SOLVED] btSparseSdf compute null normal
-
- Posts: 18
- Joined: Fri May 07, 2010 2:20 pm
[SOLVED] btSparseSdf compute null normal
Last edited by guernika on Tue Jan 28, 2014 12:06 pm, edited 1 time in total.
-
- Posts: 18
- Joined: Fri May 07, 2010 2:20 pm
Re: Custom collision algorithm: btSparseSdf compute null nor
Dear All,
I "solved" my problem using the function btGjkEpaSolver2::SignedDistance(...) instead of of the btSparseSDF::Evaluate(...)
Anyway, I'd like to understand why btSparseSDF::Evaluate(...) works correctly with convex shapes (e.g. btCylinder )
and doesn't work with convex triangle mesh shapes (e.g. a triangulation of a cylinder ).
The problem is a runtime error because btSparseSDF::Evaluate() computes a 0 length normal vector.
Any hint?
Do you think the use of btGjkEpaSolver2::SignedDistance(...) will increase computation times?
Thanks in advance,
Francesco
I "solved" my problem using the function btGjkEpaSolver2::SignedDistance(...) instead of of the btSparseSDF::Evaluate(...)
Anyway, I'd like to understand why btSparseSDF::Evaluate(...) works correctly with convex shapes (e.g. btCylinder )
and doesn't work with convex triangle mesh shapes (e.g. a triangulation of a cylinder ).
The problem is a runtime error because btSparseSDF::Evaluate() computes a 0 length normal vector.
Any hint?
Do you think the use of btGjkEpaSolver2::SignedDistance(...) will increase computation times?
Thanks in advance,
Francesco