More Brooks Subsumption???
DEACCON:
C++ Conversion
-pick up and finish implementing advanced seeding
-pick up and finish 3-2 if needed
Next Week:
3D DEACCON
Archive for June, 2008
Goals 6/9/08
Monday, June 9th, 2008
Reading:
More Brooks Subsumption???
More Brooks Subsumption???
6/3/08
Monday, June 9th, 2008
I tossed in some checks and tweaks and managed to speed up DEACCON algorithm today. I also have pulled and started reading some of the Brooks Subsumption papers. Its a little weird to be reading papers what were written before i started pre-school, the style and layout of the paper is very different from the standard format I am used to. Fixing the growth hangs and the combining algorithms next, then probably working on seeding issues tomorrow.
6/9/08 -Deaccon Preconversion & Reading
Monday, June 9th, 2008
I figured out how to implement a simple fix for DEACCON that should take long to do and will solve many problems. It involves removing the need for the slope component of the vertical line which is always undefined and then replacing it with a simple plug into the equation of the other line and checking the resulting points. I also just finishing writing up a couple of AI papers and the Belmont report. Very interesting reading though a lot of it is common sense except for the portion about the justice of the research and the less obvious section of the protection of vunerable populations.
06/06/08
Friday, June 6th, 2008
So I found the 2 lines of code that were totally destroying DEACCON. Apparently half of my intersection test was totally not working because I was relying on the slope of lines for the calculation and half the lines i feed into that algorithm are vertical and therefor have an undefined slope. I will dig up another more robust algorithm and i think it will work perfectly.
Also I have a nice stack of reading to do but all the papers have had at least a quick familarization reading pass.
6/5/08
Thursday, June 5th, 2008
Upgrading DEACCON with the ability to finalize shapes and then stop wasting CPU cycles trying to grow them. Cleaned up the rendering a touch as well. I have found a very enjoyable paper by Brooks where he presented the subsumption architecture though the style is a little strange. I found and fixed 3 growth bugs as well as a combining bug and a convex shape detection bug. And I think i have the major bug where the polygon fails to grow when it hits a shape localized and hopefully fixed very soon.
I also have the combining working a lot better DEACCON is now regularly making trapizoids however there are some issues that will prevent it from ever fully perfectly combining due to the lack of precision in the shapes and in how shapes grow. When combining I am checking for overlapping points, but some of the growth patterns can only yield close’ish points. The edges are still coplainer which is the most important thing but they don’t have exactly the same end points. I am afraid to “snap” the points together as that could introduce concavity issues, as well as exposing uncovered space which would violate my assumptions of never releasing claimed space, and destroy coplainer gateways.
The new seeding algorithm has been worked out on paper though I want to finish the growth issues first, that one may be integrated into the openGL conversion next week since these fixes are taking a little longer than planed to run down the bugs. Also I will probably have to rethink how the seeding algorithm is integrated into the program to fit it into the CGUL toolkit so it makes sense not to rewrite something twice.