AUTOMATIC MOTION SYNTHESIS FOR 3D MASS-SPRING MODELS

We describe how to synthesize, automatically, motion controllers for l ocomotive tasks involving animated characters modeled as 3D mass-sprin g lattices. The motion controllers determine an actuation sequence bas ed on elapsed time, not physical state. Actuation is represented econo mically with a small, predefined set of global lattice deformations, a nd a stochastic search is used to determine effective values for the c ontroller parameters. Our algorithm generates controllers that produce attractive, visually plausible motion for simple locomotive tasks in under 1 h on a standard workstation, which is more than an order of ma gnitude faster than comparable approaches to motion synthesis for 3D a rticulated linkage models.