Primary Contributions

This thesis contributes both to the field of computer graphics and to the field of artificial life. It leverages the synergy between these two fields for the realistic modeling, simulation, and animation of animals. Our contributions have been published in the computer graphics literature [Tu and Terzopoulos1994a, Tu and Terzopoulos1994b] and in the artificial life literature [Terzopoulos, Tu and Grzeszczuk1994a, Terzopoulos, Tu and Grzeszczuk1994b].

In computer graphics, our work is the first to combine within a unified framework extensive physics-based graphics models, locomotion control, and higher-level behavioral models for animation. In the context of artificial life, we develop animats of unprecedented realism and sophistication. Our life-like animations of fish in their habitat demonstrate a functional model that captures the interplay of physics, locomotion, perception and behavior in animals. The behaviors that our artificial animals emulate range from reflexive behaviors to motivational behaviors, and from complex individual behaviors to elaborate group behavior. It is important to appreciate that our goal is not to attempt to replicate the complete behavioral repertoire of any one specific fish species but, rather, to develop a generic behavioral model suited to the animation of various species of fishes.

The main contributions of this thesis in more detail are as follows:

• We develop an animation framework that, with minimal intervention from the animator, can achieve the intricacy of motion evident in certain natural ecosystems. This framework encompasses realistic appearance, movement, and behavior of individual animals, as well as the patterns of behavior evident in groups of animals. In addition, unlike many other animation systems, our framework has promise for interactive graphical applications, such as virtual reality. Our paradigm is validated by a physics-based, virtual marine world inhabited by a variety of realistic artificial fishes. In particular, we have developed:
  1. An efficient, physics-based graphical fish model:
     • We introduce the first graphical fish model to yield life-like aquatic motions without keyframing. Our physics-based fish model captures the streamlined shape, the muscular structure, and the general biomechanical properties of natural fishes.
     • We have constructed a set of motor controllers that effectively control the muscles of the artificial fish to generate realistic fish locomotion.

  2. A perception model: The perception model simulates essential visual abilities and limitations. It is equipped with a perceptual attention mechanism which is lacking in previous perception models for animation. This perceptual attention mechanism is essential for the realistic modeling of behavior.
  3. A behavior model:
     • A model of the internal motivations of an animal which comprises the innate characteristics of an animal and its dynamic mental state.
     • A set of behavior routines that implement a range of individual and group piscine behaviors that are common across many species, such as collision avoidance (in the presence of both static and moving obstacles), foraging, wandering, searching for comfortable niches, fleeing, schooling and mating.
     • An ``intention generator'' that arbitrates among different behaviors and controls the perceptual attention mechanism.
• This thesis provides a new experimental environment for research in related disciplines, such as computer vision and robotics. For example, artificial fishes have been used to design an active computer vision system and evaluate its performance [Terzopoulos and Rabie1995] and they have been employed to develop algorithms for learning locomotion and other motor skills [Grzeszczuk and Terzopoulos1995]. Artificial fishes are virtual robots situated in a continuously dynamic 3D virtual world. They offer a much broader range of perceptual and animation capabilities, lower cost, and higher reliability than can be expected from present-day physical robots used in hardware vision [Terzopoulos1995]. For at least these reasons, artificial fishes in their dynamic world can serve as a proving ground for theories that profess to account for the sensorimotor competence of animals.

Xiaoyuan Tu

January 1996