I'm planning on doing a bigger update a little before the beta review, so this will probably be one of my shorter posts. What with quite a few projects going on this week, I was not able to get done everything for the Beta Review yet. However, over the next few days I have a mysterious lull in work from other classes, so I should be posting more progress in a couple days. As planned, I have finished up basic user interaction - the main difference from the final implementation with the pseudocode from last week is that I had to use impulse rather than straight-up force to modify fluid behavior because of the way force is currently zeroed at the beginning of each time step. (I might change this later). I finally put in an fps tracker as well - I have it display on the GUI status bar at the bottom - right now the simulation is ranging around 2 fps for the 1000 particle set, and around 20-some for the 100 particle set on my laptop.
I'm still working on the implicit reconstruction, mainly going back and doing a bunch more background reading on feasible solutions, such as:
http://research.microsoft.com/en-us/um/people/hoppe/psrecon.pdf
http://web.mit.edu/manoli/crust/www/crust.html
http://physbam.stanford.edu/~fedkiw/papers/stanford2001-03.pdf
In particular, this paper discusses a method which might be useful which is particularly suited to being faster and more memory light. Essentially, it uses a level set method, interpolating the zero isosurface of the level set function. A simplified version of the motion of the surface is also embedded in addition to the surface, which might be a good way to get fast deformations without having to recompute the entire mesh from scratch.
UPDATE:
Here is a video of user interaction working. The video also shows how clicking shoots a ray from the camera. The framerate is now displayed in the lower left corner of the status bar.
For next week and the beta review, same plan as before: focus on the implicit formulation.
No comments:
Post a Comment