[MUD-Dev] [MLP] NPC Complexity

Kwon Ekstrom justice at softhome.net
Mon Apr 29 20:44:06 New Zealand Standard Time 2002

From: <Daniel.Harman at barclayscapital.com>
> From: Kwon Ekstrom [mailto:justice at softhome.net]

>> You could add an NPC layer to the AI (which now that I think
>> about it, wouldn't be a bad idea anyway), which races propagate
>> events to when they're done handling them.  Then write an AI to
>> allow NPC's to transfer data between themselves.  Then you can
>> add an AI to the standard NPC init code allowing them to learn.

>> I'm using the "higher" level AI's for 2 things, they're a central
>> distribution point, and it's a pooled AI, so all the creatures of
>> that grouping have access to it.

> I quite like this idea, but I do wonder if identical behaviour
> couldn't be achieved through the use of emergent systems. Rather
> than trying to model the higher level functions of npc group
> interactions, one could look at what motivates people on an
> individual level, and by modeling them create these seemingly
> complex results.

There'd be a variety of ways to create similar behavior, the layered
AI system is meant setup for the game designer more than the
resulting behavior.  It's designed to use a minimum of system
resources while allowing the behavior of creatures in the realm to
be extended easily.  Since it's event driven, you can create
pseudo-intelligent reactions without creating a truely active AI.

> When groups of people do things, it seems to me that it is more
> down to the leadership than the bulk of the group. Perhaps we just
> need to study the motivations of leaders and then model them.

Yes, the leader of the group is the one primarily making the
decisions, or rather, the leader is a central dispatch point of the
decisions.  If you wanted to truely emulate behavior of a group, you
would have to give each mobile specific it's own set of goals, then
your "leader" mobiles would need to try and fill those goals in the
people following it.  The success of said leader will be in how well
he fulfills his goals, and keeps his followers happy.  If the leader
is trying to do something fairly popular, he could build a large
group of followers pretty fast.

> Of course thats skipping a step, the first stage is to look at why
> beings group and then stay grouped. Then to model why one creature
> might emerge as leader of the group (or subgroup), followed by his
> motivations to war/plant trees/build a castle/etc and his
> abilities to motivate others to do the same.

> It would be a layered AI in the sense that there would be a
> hierarchy, but at the same time you wouldn't have different rules
> for different entities, and I suspect it might have interesting
> behaviours you couldn't/wouldn't have thought of if you were
> modeling them explicitly at a higher level.

I don't see a problem with having different rules for different
entities.  In most rpg's the behavior of a character is heavily
colored by his or her race, and/or class.  you don't expect
individual to behave the same in most aspects.  There are a variety
of things where the majority of your people would behave in a
similar manor, most people would retaliate or run when attacked,
depending on the situation.

As I stated above, the purpose of the layered AI wasn't to emulate
true behavior, it was to create complex behavior in a way that was
easy to expand.  My job is to create an interesting world for
players to explore, not to create a simulation.  It cuts alot of
corners, but creates an interesting environment.  And it's easy to
adjust the behavior model, you have alot less factors to take into
account to achieve the desired behavior.

Alot of people seem to be trying to simulate "real" behavior, when
in truth I don't think that's a practical result.  The discussions
on how the sims handled their AI shows that it was just a priority
list of needs compared to how well a specific object handles those
needs, and for the most part, I hear people ranting and raving about
how well it portrays people.  It's a simple model to simulate a
complex interaction.

This is similar to Psychohistory in Asimov's Foundation
series... where the mathimatician Harry Seldon predicts the future
using mathimatical probabilities.  The theory states that the larger
the group of people the more accurate the results.  This is because,
the larger the group the less factors affect each member, when you
get down to an individual, the number of factors you have to take
into account is immense.  I don't believe that psychohistory will
ever become a reality, large groups of people are generally governed
by a small group.

-- Kwon J. Ekstrom

MUD-Dev mailing list
MUD-Dev at kanga.nu

More information about the MUD-Dev mailing list