My specific area of interest is designing the AI of real-time multiplayer games, for example, RTS games. Moves by players happen simultaneously, so I was wondering whether or not it would be modeled with a sequential (extensive form) game tree or something else. I don't know if this is even the correct approach at all, and if some other algorithms are more applicable.
How much of Artificial Intelligence design is based off of game trees or extensive-form games, and are there any suggested textbooks which thoroughly cover the subject game tree application to AI or AI of RTS games?
Answer
I will give this Question a partial try.
I guess that Game Trees are the same as decision trees and the answer to the question is, sure, why not.
But this alone will not solve the riddle for a RTS AI, you will need a combination of ai techniques to get the task done.
Some techniques you may want to use/combine:
(warning, the most of this is pure clasical ai stuff)
Control
- decision trees
- behaviour trees
- statemachines
- fuzzy logic
- recursive depth or Breadth-first search for planing tasks
- neural networks for more fuzzy stuff
- maybe even genetic algorithms/genetic programming for more fancy ai / emergent behaviour
- maybe even this Designing Emergent AI
to name only a few.
Navigation
- Dijkstra
- A-star
Planing
I have absolutly no idea about this field because i never programmed a rts ai, but take a look in the book i linked below.
Books
A very good book for Game AI Programming is Artificial Intelligence for Games.
No comments:
Post a Comment