Clipping with NewtonMesh

[Markus]

Hi,

for the domino simulation I'm programming I encountered the problem that the domino pieces will not stay upright on surfaces with a little slope. I'm loading the scene from a 3ds file into a TreeCollision and tried to place the pieces on a small ramp, but they start falling even at a small angle.

So I figured one could use NewtonMeshDifference to clip the domino pieces against the scene collision in order to adjust the bottom face to the underlying surface (See attachment for a picture).

I tried using the clip functions with a small test program:

Code: Select all

   tmp = identity;
   tmp1 = translate(0.0f, 2.5f, 0.0f);

   NewtonCollision* collision1 = NewtonCreateBox(g_world, 5.0, 5.0, 5.0, 0, tmp[0]);
   NewtonCollision* collision2 = NewtonCreateBox(g_world, 5.0, 5.0, 5.0, 0, tmp1[0]);
   NewtonMesh* mesh1 = NewtonMeshCreateFromCollision(collision1);
   NewtonMesh* mesh2 = NewtonMeshCreateFromCollision(collision2);

   NewtonMesh* mesh3 = NewtonMeshCreate(g_world);
   NewtonMesh* mesh4 = NewtonMeshCreate(g_world);

   NewtonMeshClip(mesh1, mesh2, tmp[0], &mesh3, &mesh4);
   // mesh3 and mesh4 are NULL


The result i get is that both mesh3 and mesh4 are NULL. Using NewtonMeshDifference, NewtonMeshIntersection and NewtonMeshUnion leads to the same result. I looked into the code and realized that all clip functions use void dgMeshEffect::TransformMesh (const dgMatrix& matrix)
which consists of "_ASSERTE (0);" Is this the problem or am I doing anything wrong? I could not find any documentation to these functions anyway...

Maybe there even is an idea much more simple for achieving the same thing?

[Julio Jerez]

you are doing it correct, There was a bug in the mesh difference that I fixed this morning, (I guess it is a case of serendipity)
basically the function will return bogus value if the input manes have coplanar faces, it is fixed now and check in. I discovered when I was making the Compound collision demo.

If you are sync to SVN just update and it should the test should work.

[Julio Jerez]

looking at the function I do not think will work, try this

Code: Select all

  matrix tmp = identity;

   NewtonCollision* collision1 = NewtonCreateBox(g_world, 5.0, 5.0, 5.0, 0, tmp[0]);
   NewtonCollision* collision2 = NewtonCreateBox(g_world, 5.0, 5.0, 5.0, 0, tmp[0]);

   NewtonMesh* mesh1 = NewtonMeshCreateFromCollision(collision1);
   NewtonMesh* mesh2 = NewtonMeshCreateFromCollision(collision2);

   matrix tmp1 = translate(0.0f, 2.5f, 0.0f);
   NewtonMesh* const mesh3 = NewtonMeshDifference (mesh1, mesh2, & tmp1[0]);

   // do not forget delete the meshes, of you will leave big memory leaks behind
   

[Julio Jerez]

Ha one more thing, if you are going to clip the dominos against the floor for better,
do not make a collision mesh of the terrain, and clip against the terrain, that will make the boolean very slow if the mesh have lots of faces.

I think it will be better to get the faces underneah the domino and make a temp NewtonMesh and then clip gainst that smaller mesh.
you can use the full mesh and get it to work, but if it is slow you can try that as an optimization phase.

[Markus]

Thanks Julio, it seems to be working great in my small test program!

I have one other issue, though: When clipping the white box (domino) against the red box (floor), I will end up with the blue collision, created with NewtonCreateConvexHullFromMesh(g_world, mesh3, 0.0f, 0);, where mesh3 is the result of the difference operation. This is obvious, because it is a convex collision. What would I need to do in order to get the yellow collision? Is there a way to do that in Newton?

[Julio Jerez]

to get the yellow collision ther is a another set of function.

firt clip boolean the mesh until you get teh shape you want
the you make a comve decompaosition using the function NewtonMeshTetrahedralization
then you call NewtonCreateCompoundCollisionFromMesh, and it will and make an optimal compound collision


whowever NewtonMeshTetrahedralization and NewtonCreateCompoundCollisionFromMesh are no workimng yet, I am make that demo and I am hopping to have ready by this week end.

[Markus]

That would be great, thanks. It is working nicely with a convex shape even on surfaces with a large angle (see picture).

I will be testing it with these function after you have implemented them. I'm looking forward to it.