glob2-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [glob2-devel] Creature AI


From: Martin Nyffenegger
Subject: Re: [glob2-devel] Creature AI
Date: Fri, 20 May 2005 13:19:55 +0200

A Jean-David post ! This is quite rare ;-)

A little precision to you Jean-David: Andrew was talking about the AIs of the globules (now just a decision Tree) and not about AI- Castor and friends.

He wanted to improve the behaviour of the Globules and not the one of the Opponent player !

Barock

Le 20 mai 05 à 11:39, Jean-David Maillefer a écrit :

Hello,

Improving the AIs is an old TODO, but one of the most challenging.
I tried it, but stopped rapidly.
Why ?
The AI concepts are relatively simple, but the implementation is far more difficult
To successfully implement the AI, you need:
1) a good measure of quality of the AI. This may sound quite easy, but it takes often too many computational power or memory space. (It's a common mistake to underestimate this when creating genetic algorithms) 2) a deep understanding of how glob2's core is programmed, since you need access to many part of it. 3) a lot of time (I think programming will take less than 50% of the total AI creation time) 4) a very good AI design (should be fully agreed by the core developers. I think here to NCT and Nuage, who wrote the existing AIs)
5) the AI needs to be better than the existing AIs (logical).

If you think you can (with help) satisfy these five points, it will work. I know NCT will be pleased to respond to all questions related to points 2 and 5 (if he has time to ...)

I hope this list not discouraged you and help precise the AI requirements.

Jean-David



_______________________________________________
glob2-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/glob2-devel







reply via email to

[Prev in Thread] Current Thread [Next in Thread]