Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!newsfeed.internetmci.com!news.exodus.net!news.wni.com!noc.netcom.net!news3.noc.netcom.net!netcom.com!nagle
From: nagle@netcom.com (John Nagle)
Subject: Re: Accurate Physical Modelling
Message-ID: <nagleDLxKD4.7D5@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
References: <19960128.001434.93@mageton.demon.co.uk> <4ef41j$fms@Venus.mcs.com>
Date: Mon, 29 Jan 1996 06:50:51 GMT
Lines: 28
Sender: nagle@netcom11.netcom.com

jorn@MCS.COM (Jorn Barger) writes:

>In article <19960128.001434.93@mageton.demon.co.uk>,
>Archmage <Archmage@mageton.demon.co.uk> wrote:
>>As far as I can tell (I only just thought of it, I could be
>>wrong) it would take an infinite amount of calculation
>>to properly (at a low level) model even a simple collision
>>of two objects [...]

>I think this is one of the Pons Asinorum's of AI (nothing personal!).

>To represent a situation you need only represent enough detail
>to distinguish it from *cases with different outcomes*.

>The outcome of a collision might be:

>- collide/ not-collide
>- collide-and-stop/ collide-and-rebound/ collide-and-glance-off/ not-collide
>- collide-and-glance-off-at-64-degrees-and-37-degrees/ etc etc etc etc etc

>You always pick the coarsest approximation you can get away with,
>and represent only waht's needed for that level.

>(I think there's probably a 'grey area' that you have to live with,
>where you can't be sure which will occur.)


>j
