Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.sprintlink.net!newsfeed.internetmci.com!news.mathworks.com!news.kei.com!wang!news
From: bruck@actcom.co.il (Uri Bruck)
Subject: Re: Cheating by AI in games
Organization: ACTCOM - Internet Services in Israel
Date: Sat, 7 Oct 1995 23:25:08 GMT
Message-ID: <DG3qDw.9q1@actcom.co.il>
References: <446muf$i0c@nyx.cs.du.edu> <448rul$asd@ixnews4.ix.netcom.com> <tgooding-2609951122550001@helser196.res.iastate.edu> <44bhsu$r2m@ixnews7.ix.netcom.com> <tgooding-2809950021420001@helser196.res.iastate.edu> 
Sender: news@wang.com
Lines: 51

Eric Dybsand (edybs@ix.netcom.com) wrote:
: In <DFxzpx.LIp@actcom.co.il> bruck@actcom.co.il (Uri Bruck) writes: 
: >
: > on the words that Eric Dybsand (edybs@ix.netcom.com) wrote:
: >
: >: >Heh. sounds like a search tree to me.. ;)
: >
[snip]
and then I added:
: >I think this can done locally, on  a small scale, and, if SIMM's
: serve,
: > has actually been suggested on the startegic dispositions thread a
: while
: >back.
: >Some games, uch as computer chess, do exactly that.
: >Uri
: >

: As another poster pointed out, search trees in chess have been used for
: quite a while, and IMO, are feasible due to the limited options that do
: exist in making a chess move decision (terrain is all the same, each
: type of piece is identical as a rook for one side always has the same
: characteristics as a rook for the other side, a limited sized map) and
: also due to the turn sequence and virtually unlimited time available 
: for the AI to make the decision.  If you consider the typical modern
: wargame (Panzer General, Harpoon, Empire, DuneII, C&C, etc.) you may
: note that the decision is based on much more information and variables.

That would depend upond the level of description you insist upon. There 
are many situations which at first glance may seem different but actually
could described the same way.
Such as :
enemy is d distance in one direction
friendly forces are d2 distance in another direction, it should take them
t1 time to get within effective range of enemy 
etc.
This kind of descritopn is much more effective than:
unit at (x1,y1)
unit at (x2,y2)
and so on.

In short, a little pre-processing of the situation, which is not very 
computationally intensive - if you take into account that information on 
further units need not be as precise - can be used to formulate such 
descriptions - to which rules can be applied that would suggest a course of
action. The feasibility of such actions could be estimated using 
a short 'simulation' of the move.

I could give a more detailed explanation of this, should anyone request it.
Uri

