Hunter Hale

Fine tuning the exporter to work with sarge everything else is pretty well online and good to go.  Related works research should conclude tonight.  I haven’t found much of anything at all looking for Cell Decomp, Automatic waypoint generation, Freespace decomposition, also Freespace tessellation.  I will try and find a few others tonight cause I know we want some more supporting papers than the loosely related one I found and there must be some more out there.  There is one bit of wonkyness left in the generation but that should be fixed in five minutes.  Algorithim looks be be linear though horribly so.  O(number of regions * number of faces * growth cycles) + O(4n * growth cycles) in checks for overlaping.

The grand plan

Monday – Write the paper

Tuesday – hand it off for review

Wednesday – Revise and improve followed by submission

Once a few bugs were squished I have basically decomposed UCT level.  There are two growth artifacts I need to correct(either winding issues in the data or the growth having errors) killing those this afternoon.

I think I located the big nasty bug that was causing overlap.  I was handling the degenerate cases of a zero length edge in my collision detection algorithm by assuming the shape didn’t exist and always forcing false when attempting to collide with it.   Basically when i found a degenerate edge assume all edges are degenerate.  I did this intentionally when I first wrote the algorithm and it solves the initial case of nothing has grown yet very nicely on the first frame.  The problem occurs when the quad-splitting algorithm results in shapes composed of 5 points but that are actually triangles.  They usually have a degenerate 0 length edge or two.  Now I just need a quick work around and hopefully everything will start working.  I need to remember to handle cases like that better in the future.