Evolutive Autonomous Behaviors for Agents System in Serious Games

This article describes how to generate autonomous behavior to populate a virtual environment using Serious Games and Learning Classifier Systems. A serious game is a paradigm that simulates the real environment like a natural phenomenon. For example, people's behavior living an earthquake, fire, weather phenomenon or others. Into a serious game, the users are represented by virtual entities that have autonomous behavior taken from human's behavior. The principal interest to use serious games is that it's possible to obtain a tool with capabilities to predict, to plan and to train people involved in many natural phenomenons. The originality of this paper is that used a Learning Classifier Systems (LCS) inside in Serious games. It is possible to find a better simulation of the human's behavior into a real situation, using learning machine. The entities (agents) has autonomy and adaptability given by a genetic algorithm embedded in the LCS.