12
\$\begingroup\$

I'm interested in how strategic AI engines work, e.g. how to evaluate where to place troops in a turnbased strategy game etc.

I do get how goal based AI works, and I guess that is a good approach to decide what to build or research in a game like Civilization.

But tactical/strategical positioning of troops/units. What algorithms/strategies are used to evaluate those things? any reference papers/articles?

\$\endgroup\$

4 Answers 4

3
\$\begingroup\$

You might find reading the book "Artificial Intelligence for Games" very helpful. It gives you insight in the AI implementation of various kinds of games. (e.g. troop placement/movement, path finding, decision trees, ..)

See http://www.amazon.com/Artificial-Intelligence-Games-Second-Millington/dp/0123747317

\$\endgroup\$
2
\$\begingroup\$

Possibly not the best solution, but decision trees, specifically a minimax tree may be a good approach to take. See http://en.wikipedia.org/wiki/Minimax

You would need to create a heuristic function which can determine how good/bad a world state is. But this can become a bit more complicated if you are working with a large number of possible world states.

\$\endgroup\$
2
\$\begingroup\$

Artificial Intelligence is, unfortunately, an entire subsection of Computer Science. It's truly huge - something you can take a couple of semesters of in college just to get you warmed up.

There are a lot of common options - Decision Trees, Rules Engines, Neural Networks - the latter of which most people tend to find extremely interesting when they first hear about them (and invariably prove to be a horrible idea in the kind of scenario you're describing - look out for that). I recommend you look up some topics or books on basic AI development in general - it may not seem directly relevant to games specifically at first, but you'll appreciate understanding the fundamentals.

\$\endgroup\$
0
\$\begingroup\$

Have a look at Monte Carlo Tree Search (like min max but prunes the kind of large trees you see in games these days with large branching factors).

http://en.wikipedia.org/wiki/Monte_Carlo_tree_search

http://aigamedev.com/open/coverage/mcts-rome-ii/

http://www.aot.tu-berlin.de/fileadmin/files/lehre/diplomarbeit/BA_Barbara_Konz.pdf

\$\endgroup\$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .