Hi, does anyone here have a good knowledge of how bots are done in those games ? How do they get the bots to remember where all the weapons etc are on NEW maps ? Do these games use genetic algorithms ? do any games ?

I read a book that a little Ai in it. One of the things the author did was to have his bots remember where weapons, etc. were when they encountered them while walking around the level. The bots would also remember things such as where the enemy was. Then, when it came time for the bots to attack, they simply looked at their struct/class to see where the play was. This way, the same code works for all maps, even those not shipped with the game. He also had things such as the ability for bots to trade information with eachother, using a system where the most neede information and the best information was swapped (I think).

The author didn''t actually implement all of this, but he did sugeest how games did such a thing, and made a demo with ants who walked around, ate, and swapped information.

Ben

I am not entirely certain, but I think I read somewhere that (possibly the UT) bots use a series of waypoints to navigate the levels.

I could be wrong of course.

IIRC, you can view most of the UnrealScript code that governs the bots in UnrealEd! You definitely could look at the primitive ones in Unreal (as opposed to UT).

Bots in Unreal move around following a serious of predefined waypoints. Basically, they decide what they want and try and go and get it - pretty much the same as the old ReaperBot (no surprise, Steve Polge being the author of both UT''s bots and the Reaper) and others. A degree of randomness is added in.

The Quake 3 bots are basically souped up Gladiator bots, by Mr. Elusive.