Long Live DEACCON
Archive for January, 2008
QCR No More
Friday, January 25th, 2008
So QCR has a real name now. No longer Qualtative convex regions it is now “Decomposition of Environments for Automated Creation of Convex-region Navigation-meshes”.
Combining works
Monday, January 21st, 2008
Just ironed out the last major bug in combining. I think it is producing fully usable decomps now. I wanna touch up my validation code and the superior spawning code that was mentioned using midpoints between obstructions to spawn but I think this can be applied to problems. That leaves 9 days to generate some use cases for it and add those into my paper. And of course editing the paper.
The initial seeding in this image was 1k by 1k so there were 25 initial seeds.
IT LIVES!!!!
Tuesday, January 8th, 2008
Got it working and tossed a massively complicted and dense case at it.
Hunter
Many more fixes.
Tuesday, January 8th, 2008
I think I cleared the overgrowth and hanging problems. Now I am doing some stress testing and then I will do seeding.
Hunter
PS apparently it does spheres now(least that what it looks like if the resolution is low enough
Almost got it
Thursday, January 3rd, 2008
I keep coming back to the same problem though this time it should be solvable. For some reason the application isn’t following the lines correctly. One of my coordinate finding methods has got to be picking up the wrong values. The problem has to be in my line following code. Since I took that down to 15 lines and made it readable I will probably try and draft someone into looking over it today. Aside from that the seed tossing needs turning back on. And I wanna put in some code to limit the number of edges a shape can form. Now it is theoretically possible to go all the way up to 8 sides, so I want some user control over how many sides it can split to. Also I really really need a better method of checking for nulls in my overloaded == and != methods, the current method of using try catch blocks and probing for a null first thing is a massive processor sink and it isn’t workable.
Hunter
Update – I found the problem where the lines were being wonky. I left out a multiplication when I was finding the Y-intercept of the line I wanted to grow along. That ment that my line equation was off but explained why the slope looked correct just translated. Easy fix once I located it. Now one of my special cases isn’t growing quite right but it is looking a lot better and smoother than it ever has before.
Implementation Done, Now to debug
Wednesday, January 2nd, 2008
Got all the code implemented for my new design now I have to debug it hopefully that will get done today. This should be easier since I have 150 lines spread accross 2 methods vs 1700 lines in one method in the old system. Also I no longer have to worry about so many special cases.
Note to self : When overriding the == operator don’t use the == inside itself to try and check for null inputs. Use try catch and eventually hope to find a better way since that one kinda stinks.