Invoking Behavioral Patterns on Autonomous Virtual Agents by Fuzzy Cognitive Maps and Video Games Techniques using Modern Tools

In order to increase the game characters realism and its arti cial environments, recent research has focused on improving the autonomous adaptation in real-time of characters to particular situations. This work resume an experimental process for creating Autonomous Virtual Agents (AVAs). Our work shows the main phases where an appropriate cognitive modeling leads to suitable responses from external stimuli and internal desires. As rst proof of this, an undersea world is built and used to test the perception capabilities of arti cial sh and sharks. This behavioral system is based on MoCAMG cognitive architecture following a methodology that combines common video-game techniques with non classical methods of AI. The proposal highlights how simple Fuzzy Cognitive Maps allows to get high-level control and generate simple behavioral patterns equivalent to habits through realistic perception capabilities.