Stochastic AI systems

I'm pretty sure that Timkin, one of the moderators here, often mentions Stochastic models and Bayesian techniques in his AI discussions. Unfortunately there was a heated discussion here recently that seems to have changed his mind about posting in these forums anymore. You might want to do a search for Stochastic models or Bayesian techniques, which will inevitably lead to some of his posts, or try sending him a private message. I think he was one of the best resources for AI in this forum, gaming and otherwise.

Unfortunately, I don't think many people here come from your background or have deep experience with the technologies it brings to the table.

Recently there has been a lot of interest in bayesian techniques, though I can't recall any games off the top of my head which have applied them. I am sure they are out there though.

For an overview of technologies used in games, I would suggest looking at the books AI Game Programming Wisdom and AI Game Programming Wisdom 2. These books cover common solutions (and problems!) seen in games. If nothing else, you may find a few people to contact about their experiences with applying probability based techniques.

Hi!

Funny you ask because its exactly what I've been studying. I'm a computer engineering but I'm working with estimation and I of couse bayesian estimation. Seeing the potential it has for games, I did some research and now I intend to use Bayesian Networks for agent reasoning. There are games using this kind of approach, yes. From the top of my mind, I remember FutSim (a soccer simulation), but there are others...

Ah! Markov Chains also have its utilities. But I guess both MCs and BNs are not very common. At least not as much as regular decision trees and state machines!!!

Hi,

Thanks for the replies.

xlotl : Thanks - missed the search button originally - was sure there must be one somewhere. Timkin looks like a good person to get in contact with.

BrianL : Am waiting for a load of books from Amazon (I know there's a lot of info on the web, but the academic in me still prefers paper copy). Have just gotten programming gems 4, and there are some AI articles in their that pretty much address some of my initial ideas in the area - particularly about using Utility as a simple means to provide differing personalities.

S41NT : Sounds good - like the Markov chains idea - certainly hidden Markov chains seem fairly prevalent in the reading I've done so far - I've used these a lot in repeated measures modelling.

Need to learn a lot more - but hopefully I can bring some new ideas to the area as well. I think the hardest part for myself is getting too involved - for a game entity that might not last too long, I guess it's going to be a fight against wanting to create an intelligence that's too complex but hogs processor time.

Now RTS games - there's a prevailing AI I can get my teeth into!!