Solving inverse kinematics using kohonen network for 3D human walking

Abstract

Character animation in simulated virtual environments like computer games has progressed rapidly over the past few years. A wide variety of techniques are used in the process of creating 3D computer animation to build virtual worlds in which characters and objects move and interact. A wide-spread method for animating these articulated figures is kinematics. Inverse kinematics technique is the main concern in this project that refers to the problem of specifying the angles values for each joint when the end-point of the character articulation known. There are many different conventional methods to solve this problem, and this attention transformed the concept into a well-known animation alternative. However this study focuses on Kohonen network as an alternative way to solve the inverse kinematics in for walking animation for a simple human model. Kohonen learning algorithm captures two essential aspects of the map formation namely, competition and cooperation between neurons to produce a set of joint angles values when the position is defined. The ability of this artificial neural technique is proven by examining the training process and the result shows the effectiveness of Kohonen in finding the relationship and the continuity of the neurons. As highlighted by walking animation example, the Kohonen draws its good generalization capabilities through the topological order based of only a few reference vectors as a supervisor.