Page 2 of 2

Re: GeoEngine and movement improvements

Posted: Wed Oct 23, 2013 10:09 am
by Szponiasty
Szponiasty wrote:
Aristo wrote:Theoretically, this would greatly reduce the load on the CPU with a large number of characters.
Also, it will find a route to any visible distances without severe performance degradation.
I've already started writing JPS pathfinding engine. Have some experimental version done, based on some JPS code found on goole (currently it's without z support thou, so 2d pathfinding only). Looks very promising so far (It's very quick).

Anyway I have problem with moving very fat characters, for example around an obstacle. My current vectorengine path processing solves this, but only to certain max. collision radius :/ I was looking for some transformation algorithms, that I could use, but not found nothing interesting. So my idea is to process path this way: Image
Mby someone has better idea to solve this? :(
Ok. Implemented this. Results with newest Vector Engine v3 are fantastic. Character big collision radius is no longer a problem during movement :) Additionally, resulting paths are much better quality :)

http://youtu.be/OWjrV24IvLI

http://youtu.be/L_D5A4LDmDM

http://youtu.be/tQGXnDG_M-c

http://youtu.be/yZJNRJM-1BA

As for JPS, I've wrote pathfinding using this alghoritm, for other project. But it supports only 2D pathfinding, and I can't find any good info on making this work with 3D. Anyway, I'm rewriting it already, to be useable with L2J (since no z support, it will be used to find path on same layers, and between different layers, I'll use current CellPathfinding). Here is code, if anybody wanna try to do something with it, by himself.

http://pastebin.com/3myQs2h8

Re: GeoEngine and movement improvements

Posted: Sun Oct 27, 2013 4:16 pm
by StevenP
@szponiasty: I left a couple of PM to you..