A difficult problem for computer graphics is the animation of models of living creatures. For high quality movies such as those done by Pixar, almost all animation is done by instrumenting articulated body models so that a human animator can easily move and position each joint of a model for each frame of a sequence. This is labor intensive but provides the most realistic motions due to the complex and subtle knowledge that humans can bring to bear on how things walk and move. Another way to generate behaviors is to use a simulation of the laws of physics to model muscle forces, articulated body dynamics, collision forces, and neural control, to produce an autonomous lifelike creature. Here is a tutorial on some of the physical modeling methods.
				The raptor model shown here uses a new hybrid form of the structurally recursive, forward dynamics algorithm for articulated multi body systems. The hybrid modifications allows for parts of the model to use forward dynamics (given the forces compute the accelerations) and other parts to use inverse dynamics (given the accelerations, compute the forces). The raptor can walk and run on uneven terrain while performing other body motions such as snarling and tail wagging. Here is a paper that presents the details of this method
	Another approach to physical modeling is the Lagrange Multiplier work of Andrew Witkin and David Baraff which was published in SIGGRAPH 1996. Once the problem of physical modeling is solved, an even more difficult problem is that of designing a brain for the creature that is able to drive the muscles to perform realistic and desireable actions. Even bipedal walking on uneven terrain is a research area that has thus far yielded only walking and running that is highly unstable and rather unrealistic and non adaptive. Once basic navigation is possible, brain models must deal with capabilities that, for example, marine gastropods evolved some hundreds of millions of years ago. Their simple nervous systems are divided into four main ganglions of neurons. Each of these four ganglions are dedicated to a behavioral mode. These four behavioral modes, sometimes called the four Fs of marine gastropod behavior, are feeding, fighting, fleeing, and copulation. Daytime soap operas have human actors with little more that this simple motivational structure. This basic behavioral framework underlies the nervous systems of all higher animals such as humans and other mammals.
	If we extend these four Fs with two additional motivational characteristics called helping and hindering, a far greater range of behavior can be modeled. In fact, good and evil can be given an extremely simple and elegant definition based on the four Fs and the two Hs. Each individual makes observations that give him knowledge about what other individuals are trying to do. The individual can then decide whether to help or hinder. A single control variable that varies the emphasis on selfishness, helping others, or hindering others can give behaviors ranging from self destructive altruism through simple selfishness to self destructive antagonistic behaviors. This can be viewed as a recognizable framework for defining good and evil. Here is a paper presenting this behavioral model of good and evil behaviors.
	Another approach to physical modeling is spring meshes. A collection of N point masses are interconnected with springs. F=ma for each point mass gives N equations. These equations can be rearranged to give AdX = G where A is an NxN matrix, dX is a vector of point mass velocities, and G is a vector of N vectors. A and G are assembled from positions X, velocities dX, and forces at time t and dX is computed for time t+dt. If you have Java and Java3D installed on your machine and are running an up to date browser you might try interacting with a spring mass box applet. The file-open will bring in the box. Check the info menu items. Here is the spring-mesh applet.