Computer Science

Graduate Research Forum Details

Evolutionary Locomotion Syntheses

Abstract

Digital animation was originally created by laboriously hand-animating key frames. Hand-animation of physically-accurate motion is an intractable task, yet it is easy for the human eye to spot motion that does not appear physically plausible. Hand animation of articulated figures, especially locomotion, is difficult due to the hierarchical nature of the jointed systems. Motion Capture technology affords the input of hierarchical motion data, but creates fixed animation sequences that are difficult to modify at animation time. It is possible to automatically generate character animation without the use of Motion Capture data by utilizing optimization criteria and high-level animation goals. Evolutionary Algorithms have been used to teach characters how to achieve high-level goals such as locomotion. Evolutionary locomotion affords autonomy for animated characters, but requires a complex infrastructure to specify and generate animation sequences. This talk will provide a review of published work and a discussion of the current state of the art in automatic generation of animated character locomotion.