Xref:	news-dnh.mv.net comp.os.msdos.djgpp:4523 rec.games.programmer:37966
Newsgroups:	rec.games.programmer,comp.os.msdos.djgpp
From:	"gil" <hakim AT tx1 DOT elex DOT co DOT il>
Subject:	WARGAME ENGINE
Message-ID:	<DLDwyq.JHq@tx1.elex.co.il>
To:	Anybody.using.DJGPP??@Re
Encoding:	56 TEXT , 4 TEXT
Sender:	news AT tx1 DOT elex DOT co DOT il (NEWS SERVER TX1)
Organization:	Telrad Ltd.
Date:	Fri, 19 Jan 1996 02:12:01 GMT
Lines:	61
DJ-Gateway:	from newsgroup comp.os.msdos.djgpp

>I was wondering if anyone knew 1) how the AI worked 2) whether
>XCOM 3 (due in 1996 or 1997) will have a better AI? 3)  any suggestions
>that could have improved the performance of the XCOM AI?
>
>       Karl
>

Dear Karl, 
I am currently working (in my spare time, hahaha) on a tactical wargame.
The engine will be made of classes for each level of command, from the task
force HQ, let's say, to the Cies HQ, then for the platoon leaders, then for
the units themselves.

The units :will be responsible for their survival and immediate
environnement and responding (or not ) to the platoon leaders orders.

The platoon leaders : will be responsible for implementating their tactical
orders that is to say send orders to units according to the situation in
their (larger) environnement, and according to the Cies HQs orders. (I spare
you all the variations, that's a subject I have also pretty much to say)

The HQs: issue orders to platoon leaders according to their view of the even
larger situation without dealing with all the combat details and abtracting
their platoons to potentials

The Task Force HQ: Sees the whole picture and issues orders to all the part
of the front according to its understanding of the situation ( I'll explain
later) to Cies HQ.

Now we can have a set of determined actions that can be applied with pattern
analysis to a given situations. (remember that, globally, there's a limited
set of combat situations and maneouvers. I believe I can define most of them,
mostly from experience)

Talking about the situation awereness, I believe I can use "risk", "threat",
and "path" maps (list not exclusive): Those will be sort of two dimensional
maps with heat points varying on the value we want to stress out. For
instance, kill zones, movement difficulty zones, ennemy location... We can
use also probability maps for the position of the ennemy according to the
information provided by seen units also converted to heat maps. Then,
according to techniques used in image processing we can derive an action. (
techniques I can think of are pattern matching, edge detection, simulated
annealing, fluid dynamics, maybe more, I don't know...suggestions appreciated!)

Those maps would be for each level of groups (units, leaders, HQ...) and of
the size of their awereness environnement. unit sees less than Cie HQ, but
maybe with more precision. I don't know.

I strongly believe this approach to give a good close to life tactical engine.
Criticisms are welocomed !!! ;-)
Cheers.
Gil




Disclaimers: I'm only representing my own opinion, and the views expressed here don't represent my employer's. And conversely.
This disclaimer also absolves the author of any and all responsability for past, present and future events.

- Raw text -