Dissertation Proposal (in progress - I really have no idea what I'm doing here...)
- Thesis - In this work, we explore the use of interactive evolutionary algorithms for motion design in computer animation applications. We have developed some interesting enhancements to the standard interactive evolutionary algorithms including a new method of interaction with spatio-temporal data and a modified crossover algorithm that allows this interaction method to influence reproduction in a highly controllable manner. We present several applications of our method to illustrate the new enhancements and also to provide proof of their utility.
- Justification
- Background
- Outline
Crowd Paper Skeleton (in progress)
- Abstract - We present a novel method for intuitively and efficiently designing the motion of mid-sized crowds using evolutionary algorithms with interactive determination of fitness provided by the user. We aim to fill the current gap in authoring strategies for crowd animation that exists between crowd scenes that are small enough to be keyframed and large-scale, particle-based crowd simulations that are typically used for epic battle scenes. By allowing the designer to initialize the behavior of each group and to guide the evolutionary process by which our method tunes the large set of variables in a crowd system, we hope to achieve a balanced level of interaction that provides both ease-of-use and a refined level of control.
- Rough Outline
- Intro - problem statement and why it's important
- Background - here's what others have done and this is the niche we want to fill
- Method - basic crowd framework (FSM + steering primitive blending), our evolutionary algorithm, interesting interaction techniques
- Results - some example output, perhaps a subject test (depending on the publishing venue), computation times + memory usage
- Conclusion - Here's what we've accomplished (in a nutshell)
