From http://cg.alexandra.dk/, they got a tutorial on cloth simulation using Verlet integration, and iterative constraint satisfaction. Video as below:
In class we use spring mass + Euler integration, lets see what is the different
Few Observation
* Euler integration is nicer with smaller time step
* System is not very stable and very subjective to forces and constants
* Spring mass constant has to be set carefully below some threshold (got to test it) to maintain stability
* Spring mass fail to spread out the position offset from collision detection.
This lead to i) over extend of few vertex
ii) CD obj pass thru cloth as vertex are sequeezed out to the side of CD obj
So these may be the reasons why Euler + Spring mass is not used
On the hand, iterative constraint satisfaction, to me, is like a moving average that spread out the old peak while maintaining new constraint like collision detection.
Anyway, here is my Source code for Euler Integration + Spring mass
No comments:
Post a Comment