The thesis is organized as follows:
In Chapter 
we review previous work upon which our
research draws. At its lowest level of abstraction, our work is an
instance of physics-based graphics modeling. Therefore we first survey
physics-based modeling where we discuss two basic approaches: the
constraint-based approach and the motion synthesis approach. The
modeling and control technique we employ in producing realistic fish
locomotion pertains to the latter. At a higher level of abstraction,
our research is an instance of advanced behavioral animation. We
survey prior behavioral animation work and describe related previous
perception models developed for the purposes of animation. Then we
proceed to discuss the design of action selection mechanisms and
review some related previous work in ALife/Animat research.
In subsequent chapters, we describe in detail the animation system
that we have developed. In Chapter we begin by
presenting a functional overview of the artificial fish model.
In Chapter we describe the biomechanical model
and how it achieves muscle-based hydrodynamic locomotion. Next, we
develop a numerical simulation of the equations of
motion. Subsequently, the motor control of the physics-based
artificial fish, derived from piscine biomechanical principles, is
presented. This includes the construction of the muscle motor
controllers as well as the pectoral fin motor controllers.
Chapter describes our approach to constructing
geometric display models that capture the form and appearance of a
variety of artificial fishes. An interactive, deformable contour tool
is developed and applied to map realistic textures over the fish
bodies. Finally we explain how the geometric display models are
coupled to the biomechanical fish model to yield realistic animation.
In Chapter we present the perception model
employed within the artificial fish. In particular, we describe the
modeling of the perceptual attention mechanism and the use of motor
preferences for generating compromised actions. We present concrete
examples of how perception guided behaviors are synthesized. Possible
extensions of our perception model are also discussed.
In Chapter we discuss behavior modeling in the
artificial fish. We describe the internal motivations and action
selection, which is carried out by an intention generator and a set of
behavior routines that are explained in detail. Results are presented
to illustrate the various behaviors achieved in three varieties of
artificial fishes: pacifists, prey, and predators. We then analyze
the properties of the action selection mechanism that we have designed
and possible extensions are suggested.
Chapter discusses the modeling of the marine
environment of the artificial fishes. In particular, we describe the
physics-based modeling of seaweeds, food particles and water currents.
In Chapter we present the user interface that we
have designed to facilitate the use of our animation system.
In Chapter we describe the animation results that we
have achieved to date.
In Chapter we review the contributions of the
thesis, and list possible directions of future research.
| Xiaoyuan Tu | January 1996 |