Editing Dynamic Human Motions via Momentum and Force

Abstract

We present an integrated framework for interactive editing of the momentum and external forces in a motion capture sequence. Allowing user control of the momentum and forces provides a powerful and intuitive editing tool for dynamic motions. To make a higher jump, for example, the user simply increases the linear momentum in the vertical direction, while our system automatically calculates a motion that maintains both the same landing position and physical plausibility. Our key insight is using trajectory optimization based on normalized dynamics to simultaneously propagate momentum and force space changes. We demonstrate our approach with edits of long sequences of dynamic actions, including kicks, jumps, and spins.