Evolutionary approaches to evolve AI scripts for a RTS game.

ABSTRACTThis paper evaluates three different Evolutionary Algorithms (EAs) for generating non-player characters (NPCs) via Artificial Intelligence scripts for a Real Time Strategy (RTS) game. The first approach executes only a Genetic Algorithm (GA), while the second and third ones combine GA with a Dynamic Scripting approach. The Bos Wars game is used for testing these EAs, which are able to create and to evolve strategies of this RTS game coded as scripts in LUA language. In order to do this, the EA communicates with Bos Wars' engine by sending scripts, playing matches and capturing statistical data to evaluate its individuals.