Hello-
I'm interested in implementing this paper for a cloth simulation I'm working on. The problem is that the "solver" section is fairly vague. Well, it's not vague, but it implies knowledge about Gauss-Seidel which -- even after a bit of research -- I cannot find a clear example of its usage. I studied numerical methods a bit in college, but I'm not familiar with all the intimate details of course.
I'm confused as to whether or not I'm supposed to be creating a giant matrix and solving that, or if Gauss-Seidel is just a nice name for a relatively simple practice (with a strong mathematical concept "behind the scenes").
I'm mostly looking to implement the cloth solution from this paper. Any direction, code samples, etc. should be useful. If someone were to walk me through an iteration of a basic 2x2 piece of cloth, I'd be very thankful
So say I have edge constraints:
C1(p2, p3), C2(p3, p1), C3(p1, p4) C4(p4, p2)
Should I also have diagonal constraints? (e.g. C5(p1, p2) & C6(p4, p3)) Reading the ppaer, I believe I should have C5(p1, p2) but not the other diagonal since it's not a triangle edge. Am I reading that correctly? Why not have the other diagonal?
And I know I'll have one bend constraints Cbend(p1, p2, p3, p4) between two two triangles /\(p1, p3, p2), /\(p1, p2, p4).
So, I have those 5-7 constraints.. How do I solve this thing?
Am I reading too heavily into this? I have a feeling I am.
Am I supposed to just go from constraint C1 through CN and "apply" (project) the constraint to p1-p4.
So C1 updates p2 & p3, C2 updates p3 & p1 (using the results from C1), and so on? In the case of cloth, this would be extremely similar to the Jacobson method except adding bend and collision constraints?
Is the real novelty here for cloth is that we are using predicted positions to modify the velocity ala the Nonconvex Rigid Bodies w/ Stacking paper (albeit in a non-impulse way)?
I'm new to a lot of this, so any help or guidance would be appreciated. While I value the mathematical theory behind things, all I've been able to find is theory so some concreteness would help!
Thanks,
-john