[MUD-Dev] FEAR Animat style AI for Quake 2

Ted L. Chen tedlchen at yahoo.com
Tue Nov 26 17:19:27 CET 2002


Frank Crowell wrote

> FEAR (Flexible Embedded Agent aRchitecture) is an Open Source
> project that is an AI based on animat or the AI of virtual
> creatures.  They have implemented a library, development kit, and
> an interface to Quake 2.  According to the developers,

I haven't tried FEAR specifically, but I know the type.

The game I'm working on partitions the environment and the AI,
allowing it only to pass through the same inputs as the player.
Similar to FEAR, there's no way for the AI to cheat (other than
having a more precise sense of the environment).  The underlying
control systems that map decisions to physical inputs is the easy
part, as is low-level automata.  They work really well for simple
test-cases like obstacle avoidance and movement, so I can understand
why the guys over at FEAR are so excited.

But, in terms of believable AI, unless you have a really good
decision-system in place, it's not going to give you much
"simplicity" or "realism".  The hard work is in defining the
decision space.  At a simple level, the stimuli-response of 'animat'
can easily be tricked, and actually produce funny results because
they have no overriding executive.

For example, assume we have a threat detection organ that tells the
rest of the brain that someone is attacking when they have a rifle
pointed towards us.  What happens when the guy with the rifle turns
rapidly toward and away from us?

The annoying thing is though, even if you have a decision-system in
place, chances are it doesn't take into account the
continuous-discrete nature of the combined AI-physics system.
You'll start getting overshoots of position, velocity, you name it,
and definately the behavior will be emergent (likely not
believable).  Not to say that it's not solvable, but it makes AI
development harder, not easier.

The old game-AI design rules still apply.  Cheat if you can get away
with it and definately cheat if the player isn't looking.  :)

P.S. I don't haven't anything against animats.  I'm just wary of
their hyped application towards human-like decision making.  There's
a reason why it's animat and not humat ;)

TLC



_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev



More information about the mud-dev-archive mailing list