Archived

This topic is now archived and is closed to further replies.

Your Biggest Problem when Programming Game AI?

This topic is 5333 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

As game programmers, what are the biggest problems you encounter when developing the AI for non-player characters? I''m not looking for specific problems, but common issues that arise more often than not (causing pain and grief ). No need for solutions either, just problems... I''m curious about the general trends at this point in time. Cheers! ajc

Share this post


Link to post
Share on other sites
quote:
Original post by Mr Nonsense
Developing an efficient representation of the environment/problem which allows npcs to extract all the relevant information needed for intelligent behaviour.



I was just pondering that myself...

-~-The Cow of Darkness-~-

Share this post


Link to post
Share on other sites
I would have to agree with the sensory and perception issue. Making decisions about a known world is what computers do natually and languages with their appropriate techniques do accordingly. However, it''s creating that "known" world in such a manner that the entity can understand it and interact with it that is the trick.

Dave Mark - President and Lead Designer
Intrinsic Algorithm -
"Reducing the world to mathematical equations!"

Share this post


Link to post
Share on other sites
To add further thought to this point... I find it difficult reducing a complex environment to a finite length string that can be of use in processing. There are conceivably an infinite number of ways to describe an environment... knowing which way is best doesn''t just come from experience...

How I process the string is irrelevant (almost... there are some techniques that just can''t handle long-winded semantics). My current project is bogged down in being able to express enough information about the domain in a compact form such that it doesn''t take an eternity to compute a singe decision for a single agent... oh, that and choosing the right information to give to each agent... I''m currently considering ways of automating the analysis of this choice.

Timkin

Share this post


Link to post
Share on other sites
Timkin: Indeed, I agree that finding relevant "features" would prove valuable research. Some people have done this in the context of neural networks... I remember fup telling me he wanted to try a genetic algorithms based approach.

That said, a statistical approach would probably do equally well at identifying features for other kinds of problems. Generate lots of random solutions with different inputs, and see which ones you get the most diversity for; those are likely to be important features... This would be great help for the AI developer!



Artificial Intelligence Depot - Maybe it''s not all about graphics...

Share this post


Link to post
Share on other sites
Structuring code so that it is easy to add to and debug.

I came up with a sloppy finite state machine based on Game Programming Gems I, but I have yet to apply it to a multi-level finite state machine. Plus, when I thought about how what I did would interface with everything else in the game, I shuddered. Way too many messy references. Ew.

I haven''t touched it since then. Developing a good methodology for designing classes around what I''d like to do would probably be a worthwhile venture. Problem is I add what I need from the standpoint of what the class itself needs, instead of designing a multi-class system knowing everything I need to accomplish.

Share this post


Link to post
Share on other sites