Jump to content

  • Log In with Google      Sign In   
  • Create Account

AI Architectures: A Culinary Guide (GDMag Article)

  • You cannot reply to this topic
11 replies to this topic

#1   Moderators   

3521
Like
17Likes
Like

Posted 01 November 2012 - 01:34 PM

*
POPULAR

This is reprint of the feature article I did for Game Developer Magazine for their August 2012 issue. This is slightly longer than what appeared in the magazine. (They also had a lot of cool pictures of Mexican food.)

AI Architectures: A Culinary Guide (GDMag Article)
Dave Mark - President and Lead Designer of Intrinsic Algorithm LLC

Professional consultant on game AI, mathematical modeling, simulation modeling
Co-advisor of the GDC AI Summit
Co-founder of the AI Game Programmers Guild
Author of the book, Behavioral Mathematics for Game AI

Blogs I write:
IA News - What's happening at IA | IA on AI - AI news and notes | Post-Play'em - Observations on AI of games I play

"Reducing the world to mathematical equations!"

#2   Members   

205
Like
2Likes
Like

Posted 04 November 2012 - 03:25 AM

Very nice article and helpful,

An open question for me are genetic algorithms. Where would they fit in your big picture...is the current state of the art to combine NN and genetic algorithms? I've read some papers about that but have not seen any non-theoretic example.

Or are they more or less currently only used for optimization problems, that is for example playing an optimizer role in a planing task?

Edited by hawkeye_de, 04 November 2012 - 03:27 AM.


#3   Members   

20235
Like
4Likes
Like

Posted 04 November 2012 - 06:19 AM

Genetic algorithms are optimization tools. Writing AI for a game is not primarily an optimization problem, so in my opinion they don't fit anywhere in the picture.

#4   Members   

650
Like
2Likes
Like

Posted 05 November 2012 - 01:13 AM

Great article! Thanks for sharing it. Posted Image

I thought that the AI system I'm working on is similar in its function to a behavior tree but it's actually an utility-based AI. My intention is to use genetic algorithms to train the AI to value each 'action'. It will be a while til I get that far though...

#5   Members   

624
Like
0Likes
Like

Posted 23 January 2013 - 10:30 PM

Thank you for this article, very helpful.



#6   Members   

630
Like
0Likes
Like

Posted 24 January 2013 - 02:01 AM

Very good article, didn't realize there was a term for the AI that I was writing (BT).

******************************************************************************************
Youtube Channel


#7   Members   

847
Like
1Likes
Like

Posted 31 January 2013 - 05:27 PM

In "Programming Game AI by Example", one of the architectures presented is called a Goal-based AI. Where does this particular architecture fit in?

 

The reason I ask is because I am working on a game using this particular architecture, but I have not encountered it anywhere else.

(I think it is a really good architecture...)



#8   Members   

20235
Like
2Likes
Like

Posted 31 January 2013 - 06:48 PM

In "Programming Game AI by Example", one of the architectures presented is called a Goal-based AI. Where does this particular architecture fit in?
 
The reason I ask is because I am working on a game using this particular architecture, but I have not encountered it anywhere else.
(I think it is a really good architecture...)

It's a planner (a.k.a. "fajita"). I haven't look at "Programming Game AI by Example" in a while, but my guess is their method is something like GOAP.

#9   Members   

847
Like
0Likes
Like

Posted 01 February 2013 - 09:17 AM

No, It's not a planner. It just has a misleading name...



#10   Members   

20235
Like
1Likes
Like

Posted 01 February 2013 - 10:15 AM

What is it, then?

#11   Moderators   

3521
Like
3Likes
Like

Posted 02 February 2013 - 12:11 AM

If you are talking about the section towards the end of the book, it looks a lot like an HTN planner (Hierarchical task network).


Dave Mark - President and Lead Designer of Intrinsic Algorithm LLC

Professional consultant on game AI, mathematical modeling, simulation modeling
Co-advisor of the GDC AI Summit
Co-founder of the AI Game Programmers Guild
Author of the book, Behavioral Mathematics for Game AI

Blogs I write:
IA News - What's happening at IA | IA on AI - AI news and notes | Post-Play'em - Observations on AI of games I play

"Reducing the world to mathematical equations!"

#12   Members   

847
Like
0Likes
Like

Posted 02 February 2013 - 08:22 AM

I think you are right.

 

The way I use it, I only let the tasks (called "Goals" in the book) do limited planning while they are executing. I prefer to plan most before setting them in motion.