Since my current engine uses OBBs as the main source of collision for dynamic objects, I decided to write some swept volume code to make it "more better". What I decided to do was take into account (project) the velocity''s influence in each test. Meaning when I figured in the sum of the projections of the boxes I would project the velocity onto the current separating axis too. This is when things got odd. The way I set up tests is one box is axis-aligned and the other is rotated and translated by a matrix. Since the matrix is colum major, I was projecting the velocity onto the colums to do the tests. This turned out to be incorrect. To get the right projection, (like zero if the velocity being tested is orthogonal to the separating axis) I had to project the velocity onto the row of the matrix. I made up a bunch of orthogonal unit matricies on my notepad to see why this was, and came up with no answer. The reason I mention this is because I have to project the velocity onto the 9 axis of edge combinations, and I can''t come up with a proper solution. Does anyone have an Idea of what''s going on?

You know what's odd?

That's really close to the solution that I came up with on my own, six additional tests and everything.

Small victory for me, I guess I thought I'd share. I guess my Computer Science degree and my "almost" math degree might have helped after all.

Heh.

[edited by - BarnyardMessiah on January 11, 2004 12:41:05 AM]

Right,

I figured that from the onset, I just seemed to work backwards from the direction that the PDF took, that''s all.