[Gtkradiant] AAS System
lordhavoc at ghdigital.com
Sun Dec 28 16:34:43 CST 2008
MrElusive once tersely described the Q3 AAS system as having two-level navigation - at a high level there are areas identified by the compilre which have all possible routes precomputed, then within
each area there are entrance/exit portals to other areas, the routing from the current leaf in the AAS bsp tree to an exit to another area is computed on the fly, but the routes from an entrance to an
exit in an area is cached (it caches a certain number of recently queried routes of this sort).
So basically, the bot figures out how to get from area to area, and queries a route cache for how to get from this leaf to the exit portal of the area, it follows this path (re-querying it each frame)
until it ends up in the next area and the route cache figures out the path from that entrance to the appropriate exit, and this process repeats until it ends up in the same area as the destination,
then it's caching the route from one leaf to another, soon it arrives at its destination.
It's kind of a two-level A* algorithm, similar to the variations of A* used in RTS games (which often use multiple levels of routing).
Timothee Besset wrote:
> Matthew Tang wrote:
>> I would like to understand how the area awareness system works. I am
>> interested in using such a system for navigation of bots. If someone
>> could point in to some resources that would be nice.
>> Matthew Tang
> Good thing that all the code is available .. have fun :-)
> I'm not aware that any "high level" documentation / introduction exists.
>> Gtkradiant mailing list
>> Gtkradiant at zerowing.idsoftware.com
> Gtkradiant mailing list
> Gtkradiant at zerowing.idsoftware.com
Author of DarkPlaces Quake1 engine and mod
"War does not prove who is right, it proves who is left." - Unknown
"Any sufficiently advanced technology is indistinguishable from a rigged demo." - James Klass
More information about the Gtkradiant