Desenvolvimento de um agente inteligente para o MMORPG tibia

Abstract

Artificial Intelligence has always been used in designing of automated agents for playing games such as Chess, Go, Defense ofthe Ancients 2, Snake Game, billiard and many others. In this work, we present the development and performance evaluation of a reactive agent for the RPG Game Tibia. The intelligent agent is built using a combination of AI techniques such as graph search algorithm A* and computer vision tools like template matching. Using four algorithms to get global position of player in game, handle its health and mana, target monsters and walk through the game, we managed to develop a fully automated Tibia agent based in raw input image. We evaluated the performance of the agent in three different scenarios doing ten sessions of fifteen minutes and five sessions of one hour, collecting and analyzing metrics such as XP Gain, Supplies Usage and Balance. The simulation results show that the developed agent is able to play the game consistently according to in-game metrics.