Efficient synthesis of physically valid human motion

摘要: Optimization is a promising way to generate new animations from minimal amount of input data. Physically based optimization techniques, however, are difficult scale complex animated characters, in part because evaluating and differentiating physical quantities becomes prohibitively slow. Traditional approaches often require optimizing or constraining parameters involving joint torques; obtaining first derivatives for these generally an O(D2) process, where D the number degrees freedom character. In this paper, we describe set objective functions constraints that lead linear time analytical derivatives. The surprising finding includes on validity, such as ground contact constraints. Considering only results fast per-iteration computation times problem appears well more characters. We show qualities squash-and-stretch expected physically result our approach. Our animation system particularly useful synthesizing highly dynamic motions, examples swinging leaping motions characters having 7 22 freedom.