Generic Classifiers System and Learning Behaviours in Virtual Worlds

Abstract

Animating entities in a virtual world is a complex problem. Many are solved using some scripting engines but programmers must spend a lot of time designing and implementing them. The use of learning engines tends to ease the work of the programmer. Learning classifiers systems (LCS) mix learning and evolution to generate adaptive behaviours. The extension of LCS to a polymorphic structure simplifies the rule coding without sacrificing performances. We present in this paper this generic structure and two applications of virtual reality using those systems to produce individual and group behaviours.