Background

There have been three main approaches to the problem of specifying figure motion in computer graphics. They are kinematic control, dynamic control and motion capture.

Kinematic control is based strictly on motion. No attempt is made to simulate physical properties such as mass and force. The earliest kinematic method used in animation has come to be referred to as forward kinematics. This refers to the process in which each joint of the figure is individually moved by the animator to achieve a given pose much as an algorithm would compose the rotations of the individual joints to calculate the position of the last link or end effector. Computer graphics has also borrowed the idea of inverse kinematics from robotics. In this approach, the end effector position is determined first and then the joint angles of the figure are determined by inverting the joint angle/position relationship.

Dynamic control uses mass and force properties to guide the animation. Some workers have set up very elaborate control systems that mimic the human walking, running or jumping muscle control functions. Others have applied optimization methods to the very large systems of nonlinear equations that represent the motion of a figure through space. These methods allow the solver to animate the figure based on physical constraints such as gravity and obstacles.

This is the process of instrumenting an actor with motion sensors or optical targets and then recording the translations and rotations of their limbs as they act. This method has been applied in many films, commercials and games. Recently, several workers have proposed methods of editing and blending existing motion capture data sets to produce new animations.

How do these methods compare when measured against the amount of control given to the artist? To answer this question, we should first look at the traditional methods used to create animated figure motion.