Artificial Intelligence Berbasis Pengetahuan Pemain untuk Real Time Tactic Game Menggunakan Knowledge Based Artificial Neural Networks
Solution for the complexity problem of Artificial Intelligence (AI) on Real Time Tactic (RTT) game as one of the real-world simulation game requiresefective control system.Control system is in the form of a representative agent as a human player in anticipating the changes of the game states. In this thesis, control agent is built by applying a Knowledge Based System (KBS) based on the knowledge base related to human player action for goal achievement of the game. The construction of KBS inference is divided into two phases, which isdetermination of the case inlinguistic format of human players from the numerical values of complex states in the game, and selection of the appropriate tactic when a series of cases occur.Existanceof knowledge that is not deterministic as the basis of the inference process, requires adaptability of the agent through weighting system of knowledge and learning. KBS is mapped to Knowledge Base Artificial Neural Networks (KBANN) using certainty factor (CF) based back propagation as learning method. Inference is limited to the process of achieving main goal through controlling of attack and defense. The design of AI system is implemented in RTT Game œThe Cursed through the shared interfaces of SPRING AI Game Engine. Testing against other static AI shows the ability of adaptation to changes in circumstances and improved quality control of the game by 0.017745641. These results fit expectations of human players who expect an improvement of playing quality in each session through the selection of appropriate goal achievement action.