A Rough - Neuro Model for Classifying Opponent Behavior in Real Time Strategy Games

Real Time strategy games offer an environment where game AI is known to conduct actuality. One feature of realistic behavior in game AI is the ability to recognize the strategy of the opponent player. This is known as opponent modeling. In this paper, a classification Rough-Neuro hybrid model of the RTS opponent player behavior process is proposed. As a mean to achieve better game performance, reduction of the agent decision space and better high-level winning of real-time strategy games. The Rough-Neuro methodology allows the classification model to some extent simulate opponent behavior in playing RTS games. The methodology incorporates a two-stage hybrid mechanism. Rough sets for reduction of relevant attributes and artificial neural networks for classification opponent behavior during game playing. The proposed hybrid approach has been tested on an open source 3D RTS game called Glest. From our results we can deduce that the tactic may be successfully used for foretelling the demeanor of contender in the Glest game.