Motion Synthesis

I’ve worked four years at the Games and Virtual Worlds Group at Utrecht University as a PhD student, primarily focussed on motion synthesis techniques for walking and reaching. Basically, motion synthesis is the process of combining multiple motions/animations into new motions. Animations recorded for a specific situation can then be reused for new environments or tasks. For a more thorough overview of this work and references, see the Publications page.

Motion splicing is the technique of stitching the upper-body motion of an animation on top of the lower-body motion of another animation. This reduces the need for additional motion capture recordings. This problem has been investigated, but with the assumption that both the lower-body and upper-body motion are known in advance. I developed a solution that allowed the upper-body motion to come from a space of motions. A problem I defined as motion space splicing. This work has been published in Computer Graphics Forum. I’ve also written a survey on motion splicing, published in IEEE Computer Graphics and Applications.
This video contains additional upper-body and lower-body examples.
I did research on motion graphs, in particular evaluation of posture distance metrics. Furthermore, I developed a simple improvement to motion graphs, called path abstractions that result in higher-quality animations.
Path abstractions reduce the pelvis trajectory distortion, still, simply providing a path to animation system is an under-specification of the task we want to character to perform. Therefore, I explored step-driven animation systems. Footsteps are more descriptive than a single path and may provide the animator with more control.

Comments are closed.