Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!swrinde!newsfeed.internetmci.com!btnet!news.compulink.co.uk!cix.compulink.co.uk!usenet
From: glennc@cix.compulink.co.uk ("Glenn Corpes")
Subject: Re: Why is action AI so poor?
Message-ID: <Dnu6xC.C5t@cix.compulink.co.uk>
Organization: Compulink Information eXchange
References: <4hfuhk$6f8@engr.orst.edu>
Date: Wed, 6 Mar 1996 08:15:12 GMT
X-News-Software: Ameol32
Lines: 68

> In article <DnpHLw.KH2@cix.compulink.co.uk>,
> Glenn Corpes <glennc@cix.compulink.co.uk> wrote:
> >> WD> >Activision Studios, maker of such titles as "MechWarrior 2", 
> > > "Zork:
> >> 
> >> IMHO, the AI in MechWarrior 2 could be vastly improved. You just 
> > don't >feel the other Mech's are worthy opponents, especially when 
> > they try to >shoot their way through obstacles and run out of 
> > ammunition, and when >the game powers them up rather than make them 
> > more "intelligent" in >combat. Still like the game, though . . . >> 
> >> It's a common problem in flight sims and space combat type games. 
> > I've >not seen any of these style games where "one-on-one" would be 
> > fair to >the computer (ie in an equal fight, the human always wins). 
> > Has anyone >else?
> >> 
> >> The only exception I can think of is the "beat-em-ups" like Virtua 
> >> Fighter and Mortal Kombat. Even a lot of strategy games rely on the 
> >> computer "cheating" to make the game even.
> >> 
> >> Any professional game programmers out there want to tell us why? Is 
> >> good AI too computationally intensive, so that we'd only get 5fps 
> > with >MW2?
> >> 
> >> Neil
> >> ---
> >>  * RM 1.3 U0448 * Neil.Slater@Almac.co.uk
> >
> >I think that the main reason for simple AI is the current trend for 
> >simple doom clones, if a monster in one of these games had the ability 
> to >traverse the map properly to find you or dodge your rockets it 
> would not >enhance the game, it would just piss you off. If you want to 
> see some >more realistic AI look out for Dungeon Keeper and Syndicate 
> wars (sorry >about the advert).
> 
> Two Questions.
> 1) Just when _are_ these games coming out? I've been looking forward to 
> them.

Dungeon Keeper, april/may, Syndicate wars may/june.

> 2) Why should we believe that the AI in these games will be so great?

You shouldn't as everyone says it, this is not the side of the 
programming that I work on so I only know part of the AI story. Dungeon 
Keeper is a reverse role playing game where you control the dungeon, you 
have a bunch of monsters that you recruit who do their own thing in the 
dungeon when not under control, the main problem with this is that the 
most fundimental thing that the creatures need to do is to get to 
anywhere in the maze, we have a perfect maze solver (not unique in 
itself) which works in a the dungeon as you build new corridors and 
rooms, Syndicate wars uses a similar maze solving algorithm but here the 
problem is traversing a general polygon mesh with (destroyable) bridges 
and it works. I realise there is a lot more to AI than maze solving it 
does need solving before you can start on the other stuff, for instance 
in Syndicate wars you send a gang of agents around a city attacking, 
defending, escorting etc.. other characters, if you want to write a 
computer opponent who controls his gang in the same way you have to make 
sure that the computer's gang are going to get where they are sent, or at 
least inform the computer opponent that it's impossible. The individual 
characters in the city also run away from danger, form crowds, riot and a 
load of other stuff that I haven't seen done at an individual character 
level before.

_\/   __
  \  /  \  gcorpes@ea.com
|\ \/   /     Bullfrog
| \/   /___/_Productions
||    /___ \
