PostDoc: Parallel Search Algorithms for Automating the Animation of Human Motion

9704694哈金斯，杰西卡K.佐治亚理工学院CISE博士后研究助理实验CS：并行搜索算法自动化人体运动动画（该奖项支持CES助理南希波拉德。）逼真的人体运动模拟仍然是一个难以实现的目标。尽管最近的系统已经展示了诸如跑步、骑自行车、跳水和跳马等动态任务，但大多数当前系统的一个特定缺陷是它们不能自动适应模拟场景的变化。这项工作的目标是开发有效的并行算法，以适应人类运动的生物力学模拟，以适应任务条件的变化以及移动角色的大小和比例的变化。这项研究将开发缩放定律和搜索程序，以自动适应各种情况的模拟。由于建模和模拟人体运动的复杂性，搜索技术将是计算密集型的。作为解决这种复杂性的一种方法，并行搜索算法将在不同的硬件配置上实现和测试，以深入了解各种架构的有效性，以创建逼真的模拟。这项研究将为开发低成本的并行架构提供必要的数据，这些架构将使复杂的人体模拟能力广泛应用于从生物力学和人体工程学到教育和娱乐等领域的动画师和研究人员。

9704694 Hodgins, Jessica K. Georgia Institute of Technology CISE Postdoctoral Research Associates in Experimental CS: Parallel Search Algorithms for Automating the Animation of Human Motion (This award supports CES associate Nancy Pollard.) Realistic simulation of human motion remains an elusive goal. Although recent systems have demonstrated dynamic tasks such as running, cycling, diving, and vaulting, one specific flaw in most current systems is that they cannot automatically adapt to changes in the simulated scenario. The goal of this work is to develop efficient parallel algorithms for adapting biomechanical simulations of human motion to changes in task conditions and to changes in the size and proportions of a moving character. This research will develop scaling laws and search procedures to automate adaptation of simulations to a range of situations. The search techniques will be computationally intensive due to the complexity of modeling and simulating human motion. As one means of addressing this complexity, parallel search algorithms will be implemented and tested on different hardware configurations to gain insight into the effectiveness of various architectures for creating realistic simulations. This research should contribute necessary data for the development of low-cost parallel architectures that will make sophisticated human simulation capabilities widely available to animators and researchers in fields ranging from biomechanics and ergonomics to education and entertainment.