Adaptive behavior for cooperation: a virtual reality application

We present a behavioral system based on artificial life for animating actors in a virtual reality application. Through a virtual soccer game, we show how a set of autonomous players (called agents) can cooperate and communicate to perform common tasks. The user is immersed in the game. He/she interacts with the other agents and is integrated in the cooperation and communication systems. Every entity reads in real-time by using a classifiers system which is composed of a set of binary rules and a reward system. The originality of such method is the ability to build a behavior (by emergence) without initial knowledge. The analysis of the simulation gives interesting results: after convergence, the global behavior of the teams produces coherent movements. Moreover, the introduction of disturbances does not affect the performances of the classifiers system.