[Thesis]

This paper introduces a straightforward technique for combining performance-based animation with a physical model in order to support complex interactions with objects in a real-time animated world. The core of the approach builds off of a conceptual framework which unifies kinematic playback of motion capture and dynamic motion synthesis. The proposed method integrates a real-time recording, which allows a human actor to provide an online performance, with dynamics-based control, in order to modify the motion data based on the physical conditions felt by the character – including external influences and the effects of balance. The system smoothly unifies the kinematic and dynamics systems by changing the influence of each in a semi-automatic fashion, both spatially within the body and temporally. Examples of rich interactions interleaved with intelligent response highlights power of the technique. The current system is intended for offline motion synthesis of rich, virtual interactions with real-time, reactive human-driven performance – although ultimately the general approach is likely to be beneficial for online applications such as electronic games.

Special thanks to Roby Atadero, Steve Suh, Muzaffer "Muzo" Akbay, and Victor Zordan.
[Thesis Document]:
- pdf [2.6MB]

[Videos]:
- submission video [33.5MB]
- live capture [36.9MB]