Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!gatech!howland.reston.ans.net!news.sprintlink.net!EU.net!sun4nl!hermes.bouw.tno.nl!usenet
From: sst@bouw.tno.nl (Tako Schotanus)
Subject: Re: Some thoughts on game AI: Doom, Descent, Heretic
Message-ID: <1995Mar20.092739.7457@hermes.bouw.tno.nl>
Sender: usenet@hermes.bouw.tno.nl (USEnet Postmaster id)
Nntp-Posting-Host: ruudnix
Organization: TNO Bouw
X-Newsreader: WinVN 0.92.6+
References: <3k2166$712@netnews.upenn.edu> <Bryan.Walls-1403951051020001@colin.msfc.nasa.gov> <3k4rok$j99@mailer.fsu.edu> <3k59pc$kda@pinot.callamer.com> <3kaghl$pne@kernighan.cs.umass.edu>
Date: Mon, 20 Mar 1995 09:27:39 GMT
Lines: 47

In article <3kaghl$pne@kernighan.cs.umass.edu>, jgrass@cs.umass.edu (Joshua Grass) says:
>
>In article <3k59pc$kda@pinot.callamer.com>,
>King_Claudius <claudius@callamer.com> wrote:
>>
>>What was interesting from an AI point of view was that the bees and turrets had
>>different "modes" of operation, and that they could change modes depending on
>>the situation and that they'd react to the environment differently depending on
>>the mode.
>>
>>The bees would either be transporting a crystal to the main base, or be waiting
>>for a crystal by hanging around either the player (which usually meant the bee
>>ran into the player, causing damage) or one of the turrets.
>>
>>The turrets would defend the spaceship, attack the player, or mine the
>>asteroids.
>>
>>Suffice it to say, the programming on the AI was astounding, in addition the
>>game had fantastic graphics and sound and was fast considering it had on the
>>order of 50 ships flying around doing various things.
>>
>
>This is called behavior-based AI.  And for lots of agents, it is a
>really good and time-cheap way of making them act smart as a group.  I
>use behavior based AI in Starbound 1.0 for each individual ship.  A
>computers personallity is based mostly on how often they assign
>specific behaviors.  It works well for fairly simple stuff, but the
>problem with Starbound's AI was that the agents didn't communicate
>together well enough.  Often a behavior based system will have this
>problem in exchange for speed.  See having the agents work as a group
>would mean all of them communicating with each other which would be
>N*(N-1) communications, and perhaps even more if you wanted
>communications to change the whole strategy.

This doesn't seem logical to me. We don't do it this way in the
real worl either. You have a "chain of command", so you talk to
your "buddies" in your "squadron" only and its "commander" talks
to his fellow commanders and they again have somebody they have
to take orders from etc. etc.

This way you can cut down on the number of communication lines a lot.

_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
_/ Tako Schotanus                TNO Building and Construction Research _/
_/ Phone : +31 15 842393 Fax : +31 15 122182  E-mail : sst@bouw.tno.nl  _/
_/ My employer is required,by Dutch law,to disagree with whatever I say _/
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
