Finding Optimal Rush Attacks in Real Time Strategy (RTS) Games

Abstract

What will you fell when play with an unchangeable AI in RTS game? Of cause, it is boring. You can beat them easily and that’s no fun. In this research, we will try to design an AI with learning-ability and return the fun to players. We firstly abstract a simple AI mode, and then implement a suitable learning method . Due to some developing problems, we simulate the system (ORTS). Finally, we establish a new learning system for RTS AI. It’s an advanced point system based on the conception of the evaluation system in commercial RTS game . Decision making processes could depend on the points of each unit. Point is calculated by unit information, current game states and “experience” and. The increase “experiences” lead the value to a precise number. These changes would affect some process and up to the whole game.