Development of O(logN) Forward Dynamics Algorithm for Large-Scale Structure-Varying Kinematic Chains and Its Application to Computer Graphics Production

点击翻译按钮获取中文摘要

The research results are summarized by the following four points :1. We developed an efficient forward dynamics computation algorithm for large-scale structure varying kinematic chains such as multiple human figures. The algorithm only requires O(N) computation for the forward dynamics of N articulated bodies. The computation time can be further reduced to O(logN) by applying parallel computation. The algorithm is as efficient as the most advanced forward dynamics algorithms and also applicable to structure-varying kinematic chains.2. We developed an inverse kinematics algorithm that computes a whole-body posture from the user-specified positions of a few links, and applied it to CG animation software package. The algorithm enables the users to create animations much easier than using existing CG software packages.3. We developed several methods for generating physically feasible motions of human figures by combining the above forward dynamics and the inverse kinematics algorithms. The users can generate physically natural motions through the simple interface.4. We developed an interface for game applications utilizing dynamics simulation. Our approach proved to be efficient for such motions as circus and acrobatic ones that are inherently difficult to be captured from human performance.