Another approach to animating articulated figures is the simulation of the physics underlying the motion. These methods promise the realistic reproduction of physical phenomenon. Principles such as slow in slow out, timing, anticipation and arcs are addressed by these systems automatically.
Witkin and Kass, [12] introduce constraints and motion criteria that allow the software to create a physically based motion that meets the animator's higher level instructions. They refer to some of the animation principles that will always be the artist's domain and others that can be achieved by simulation (personality and appeal). The software in their opinion can do the physics based motions. One thing that this approach does not address is the exaggeration commonly found in film animation.
In [13] Cohen proposes a system that solves a complex set of nonlinear equations representing the movements of the figure through space over time. Each degree of freedom in the figure traces out a spacetime curve that can be affected by constraints. He creates a system to "define, modify and guide constrained spacetime problems." This system breaks the larger problem of constrained optimization into smaller windows of spacetime. This allows faster solution of the optimization problem and more interactivity with the animator.
The animator can "sketch the motion" of the figure by setting some key positions at different times. However, the animator is somewhat insolated from the simulation and can only specify constraints using mathematical equations.
A later work by Liu, Gortler and Cohen, [14] uses a hierarchical wavelet representation of the generalized DOFs that provides the ability to add detailed motion only where needed. This simplifies the calculations and the systems of equations converge more quickly.
Raibert and Hodgins[15] discuss the simulation of motion in actuated systems. This is a very high level specification of motion translated down to particular torques on the figure's joints. They characterize the higher level control instructions as suggestions to the lower level simulation of the joint torques and inertial forces. The simulation must apply the proper joint torques and use existing kinetic energy in the system to move according to the higher level suggestion.
These methods suffer from the problem of very large and complex systems of nonlinear equations. They have been limited to very simple figures.
There are some great time savings to the animator to be found in these methods. However, as Lasseter[2] and Johnston[1] point out, purely realistic motion can sometimes be uninteresting and perhaps even confusing to the viewer. The artist should have enough control to place the proper pauses and anticipations into the animation. These features allow the audience time to grasp the significance of the motion before moving on to the next thought.
| Table of Contents | < Forward and Inverse Kinematic Control | Motion Capture and Editing > |