|
|
Principles vs Observation: How do people move? I update my talk slides
regularly every two or three years. The current version is based on my
research in 2010. If you ask me to give a talk about my research now,
presumably I will bring this slide set. slide
Abstract: The animation and simulation of human behavior is an important
issue in the context of computer animation, games, robotics, and virtual
environments. The study on human movements has revealed various principles
based on physics, biomechanics, physiology, and psychology. Many of
existing animation techniques rely on those principles, which maybe
described as mathematical equations, algorithms, or procedures. Another
stream of research, called data-driven animation, made use of human motion
data captured from live actors. The research on data-driven animation has
developed a variety of techniques to edit, manipulate, segment and splice
motion capture clips. The current trend of animation research is to combine
these two approaches to complement each other. Over the past few years, we
have explored several methods that addressed the problem of simulating
human behaviors in virtual environments. Each solution relies on different
principles of human movements and motion data captured at different scales.
We found that principles and observed data can interact with each other in
several ways. Sometimes, motion data drive physically-simulated bipeds that
walk, turn, and spin. Sometimes, physics principles guide interactive
motion editing to make a canned jump higher/wider and a spin longer. The
group/crowd behavior can be captured from video, analyzed, interpolated,
and re-synthesized to create a larger group/crowd of virtual humans for an
extended period of time. Sometimes, simply adding more flexibility to
motion data allows our animated characters to navigate highly-constrained,
cluttered environments interactively.
|
