基于姿态库和PIK算法的虚拟人全身姿态生成

当前,虚拟人的全身关键帧姿态主要通过手工方式设计生成,耗时费力,将以姿态样本为基础的数据驱动方法和以直观的高层任务约束为基础的目标驱动方法相结合,实现了虚拟人全身姿态的自动交互生成。首先构建姿态样本库,依据特定任务约束,对全身关节进行分组并提出了参数化姿态拼接方法用于虚拟人全身姿态的估计生成,随后以估计值为初始姿态,利用带优先级的逆向运动学（PIK）算法对所有任务约束按优先级顺序进行IK迭代求解,实现对全身姿态的进一步优化生成。在姿态迭代求解过程中,将静态平衡约束和其它任务约束分层单独求解,解决了平衡约束优先级难以设置的问题并可确保姿态结果的静态平衡性。实验结果表明,通过将姿态样本库和PIK算法相结合,既克服了已有基于数据样本的逆向运动学算法只能生成与原有姿态样本差别不大的姿态的限制,又可解决PIK算法的初值选取问题,相应地减少了PIK算法所需的迭代求解次数,提高了求解计算效率,并可适度保证最终姿态结果的逼真自然性。

Whole Body Posture Generation of Virtual Characters by Combining Posture Examples and Prioritized Inverse Kinematics

Currently,the key-frame whole body posture of virtual characters is mainly designed by hand,which is time-consuming and hard.The combination of data-driven method based on posture samples and goal-driven method based on intuitive and high-level task constraints was proposed to realize the whole body posture generation automatically and interactively.First,according to the set of the high level constraints,the whole body was divided into some partial joint groups,whose postures were then generated from a sin...