Snap-Together Motion: Assembling Run-Time Animations

摘要

Many virtual environments and games must be populated with synthetic characters. These characters must move with sufficient realism, so as not to destroy the visual quality of the experience, yet be responsive, controllable, and efficient to simulate. In our paper we introduce an approach to character motion called Snap-Together Motion that addresses these demands. Our methodology allows quality motions to be synthesized in a controllable manner with little run-time overhead. A corpus of motion capture data is processed into a set of short clips that can be "snapped together" (concatenated) into seamless streams of motions at run time. This process is guided by an author who identifies (with help from our system) character poses that appear frequently in the corpus. Each such pose serves as a jump point at which any motion that enters can be followed by any motion that leaves, as shown in Figures 1 and 2. The result is a simple graph structure that allows clips to be connected into longer motions.