LA PHYLOGENESE POUR LA CREATION DE CREATURES ARTIFICIELLES

Abstract

Computer animation of human locomotion has become popular in recent years because of the desire to use human beings as synthetic actors in three-dimensional simulation environments. Researchers have proposed various motion control mechanisms to simulate human-like figure locomotion. However, most of the animation systems based on these control mechanisms are only suitable for animating human walking on flat ground, without obstacles. The main purpose of this study is to provide a new solution to the important problems of walking in various environments. In this dissertation, we present research into building an animation system which is capable to simulate human walking on varied terrain. For that we adopted an approach of order which is biologically inspired and is based on artificial neural networks (ANNs) theory. Another approach follows the paradigm of the artificial evolution, i.e the use of evolutionary algorithms (EAs), to optimize a given controller, who is a recurrent neural network (RNN). Anthropomorphic walking for a simulated humanoid creature has been realized by means of artificial evolution of a recurrent neural network (RNN). The approach has been investigated through full rigid-body dynamics simulations in 3D of the creature with 15 degrees of freedom. The model of controller (RNN) undergoes structural modifications on the level of its weights by using an evolutionary algorithm. Stable, natural gait patterns were obtained, with a maximum walking speed of around 0,94 m/s.