This image is from my cloth simulation I've created as the latest addition
to my game engine. It is not that complex of a simulation--it is basically
a large grid of masses connected to the eight adjacent masses by Hooke's-law
springs. The images shown are of a 30 meter square piece of cloth weighing
20kg suspended at four points in various stages of animation. The somewhat
lumpy lower-left image is after the cloth has been shaken quite a bit.
Currently there is no collision detection with itself, so the cloth can
overlap itself if it is moved too quickly. However, in the context of my
game, this is not a problem. It is used mainly for large cinematic-like
in-engine sequences, specifically when a large tarp is cut loose at one or
more corner. It works extremely well under these circumstances.
The explicit integration method of spring resolution is used here because of
the specific circumstances I am using it in. This means that if the mass of
the cloth is too small, it will explode. However, it works great when it
does work and provides accurate enough results. Even when it is stable,
however, some roughness occurs in the cloth, and I overcome this by
filtering the normals to provide a smoother look.
I've attached the source code to my cloth class and a small sample
application (with one DLL). The cloth instance used in the sample program
is created like this, in case you're interested:
g_pCloth = new Cb3dCloth(40, 40, 30, 30, 10, 20, 0.5f, c);
The immediate future of this simulation is collision with rigid, having the
cloth form around such objects and fold appropriately. Up until this point,
the folding has occurred almost without my intervention, but I'm not so sure
about that when it is colliding with other objects.
Let me know what you think!
~BenDilts( void );