[MUD-Dev] New AI Engine released

Steve Houchard stevehou at engin.umich.edu
Wed Dec 8 23:42:56 New Zealand Daylight Time 1999

On Wed, 8 Dec 1999, Fabian wrote:

<stuff snipped about braniac>
> the only thing I found lacking was better step-by-step instructions on how
> to get it to work with your own custom code.  Has anyone seen any similar
> packages out there?

The system I work with is called Soar.  It is a very general production
matching system that's been researched for something like 25 years.  The
source code (ANSI C) is available as well as all sorts of documentation
	<URL:  http://bigfoot.eecs.umich.edu/~soar>

It does have a couple drawbacks.  First, the footprint is fairly large.
While you could put together some fairly intelligent agents with it, you
wouldn't want more than 5 or so running on the same machine at any one
time (depending, of course, on the complexity of the agents).  Current
research is looking into relieving this.  Second, the production system
has some real gotchas in it.  Some things like simple math are
ridiculously difficult.  Third, unless you wait a while for me to get
my act together, you'll probably have a hard time decoupling it from

On the other hand, once you learn Soar, you can have intelligent agents
running in a reasonable amount of time.  As always, mileage may vary.    
Furthermore, Soar is a fairly expressive language.  It even has a chunking
mechanism built in.  Previous and current research include flying military
aircraft in simulated theaters of war, modelling cognitive human behavior,
natural language processing, and even playing Quake2.  

Don't know if this helps (or interests) anyone.  If someone attempts to
use Soar for a mudding application, please let me know.  


MUD-Dev maillist  -  MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list