Hi , im completely new to AI and havent read much about it. however i was thinking about it the other day , just wondering how hard it would be to make an AI for eg a Redalert/warcraft type of game. my reasoning is like this (please explain why it will not work/is bad incase you bash it :P) i could define the logic in a tree kind of way , where each node is a goal. where the top/root node is "win the game" in order to achive that goal , you need to complete some sub goals , such as: "own all cities" or "eliminate all hostile units" in order to "own all cities" we need to "take over a city" (repeated task) in ordet to "take over a city" , we need to "gather an army strong enough to kill the defence in the city" to do so we either need to "produce new units" or "gather units nearby" and so on hope you get my point.. would something like this work ? and if so , what is this model called? //Roger

I believe it's called Goal-Directed Reasoning (Behaviour or Planning).
<edit> You might also want to search with Goal-Directed Markov
Decision Processes (GDMDP) and Decision-Theoretic Planning
(DTP?) but I haven't checked these.
However, there is an article in AI Games Programming Wisdom 1
which has a good, simple example and possibly examples in the
Articles section here.

thanks for the reply.

is this a common way to do ai for such games or are there other ways that are better for this kind of scenarios?

//Roger

You have the concept right but it is generally not that simple. In your example, you may have nodes of the tree that are in direct conflict with each other... or even just too similar that they rob resources from each other's progression. As I mentioned, though, this is a common sort of thing. I know that there are some RTS games where you can define the goal as "build Wonder XXX" - which may require you to advance through "Ages", etc. and the AI will be able to determine that whole chain of sub-goals that it would take to get there. Designed right, it can be very powerful.

I also would recommend the AI Wisdom books in this regard.

I have a good solution for you. Pick up the book, Artificial Intelligence: A Modern Approach, by Stuart Russell and Peter Norvig. After you read that book and do the exercises, you won't be new to AI. In fact, you will know enough to program the AI in most games effectively.

Game AI is pretty simple stuff as far as AI is concerned. Though, that might change in 10 or 20 years. Depends if people are still interested in playing the same game with better graphics.