GPU-based Motion Blending for Motion Graphs

Abstract

Character animation, bringing virtual characters to life, plays an important role in modern filmmaking, video games and virtual reality simulations. Recently, the motion graph has been widely used in the motion synthesis of character animations. An effective motion graph should take into account both the diversity of motions and the time cost of graph construction and search. However, motion graph approach becomes time consuming as graph size increases. Due to the fact that the most time consuming step in motion synthesis is the blending operations, which builds the whole motion graph. Sometimes it takes hours to build a graph. A compromise between visual quality and performance needs to be found. But the efficient use of modern programmable graphics hardware offers potential to increase both visual quality and performance. A major aspect lies in the architecture of modern graphics hardware, known as graphics processing units (GPUs). In this paper, we propose the GPU-based motion blending algorithm for motion graphs.