Home    Ph.D.    Publications    Contact    Affiliations   

Dynamically Coupled Particle Systems
for Geometric Modeling, Reconstruction, and Animation

David Tonnesen

1998

A thesis submitted for the degree of Doctor of Philosophy
Department of Computer Science, University of Toronto

Abstract

This dissertation presents a new technique, based on dynamically coupled particle systems, for creating and manipulating complex three dimensional shapes in a fluid like manner. The most novel feature of this approach to shape representation is the use of self organizing primitive elements. In the simplest case, these primitive elements, or particles, each posses state variables of position and mass, and the system of elements interact through pairwise potential energy functions. More complex systems include additional state variables combined with simple heuristics to create application specific behavior. The ability of these systems to self organize provides a representation technique which exhibits dynamically changing structure, an attribute not found in popular spline and polygonally based representations. To illustrate the usefulness of this approach it is applied to the following problems: free form shape modeling, computer assisted animation, and surface reconstruction. For free-form modeling the approach supports smoothness constraints similar to those inherent in the deformation energies of popular, elastic surface models. Unlike spline patches or parameterized surface models, the model does not attempt to enforce analytical continuity conditions, such as tangent or curvature continuity over the surface. Applied to computer assisted animation the approach computes the movement and deformation of models mimicking, at a rudimentary level, the physical behavior of flexible solids and fluids. Applied to surface reconstruction, these systems can infer surface structure from sparse data sets, without a prior knowledge of the surface structure or the topological genus. In summary, dynamically coupled particle systems provide an useful alternative to traditional shape representation and manipulation techniques.


The thesis is available as postscript and PDF files, uncompressed and compressed (using gzip).

PDF                   Postscript  
        
The thesis as a single document, gzipped. 3.5 MB 2.7 MB
        
Split into 12 separate files, tarred, and gzipped.          3.7 MB 3.9 MB
        
Abstract, table of contents, ...
91 KB 274 KB
Chapter 1 Introduction
246 KB 1274 KB
Chapter 2 Background
163 KB 306 KB
Chapter 3 Particle Volumes
255 KB 2091 KB
Chapter 4 Particle Surfaces
197 KB 399 KB
Chapter 5 Continuous Descriptions
179 KB 1362 KB
Chapter 6 Thermal Energy
168 KB 634 KB
Chapter 7 Implementation Issues
1358 KB 27914 KB
Chapter 8 Applications
1479 KB 10952 KB
Chapter 9 Conclusions
46 KB 239 KB
Appendices
225 KB 421 KB
References
104 KB 279 KB

- Dave. Nov-01