Abstract: In virtual environment, the method of using real-time driving virtual human with motion capture to simulate the assembly and disassembly for complex products had a high requirement for virtual human's driving accuracy. On the basis of analyzing the influencing factors of virtual human's real-time driving accuracy, the solution and key technology were proposed systematically. Aiming at the problem that the driving accuracy was influenced by size difference between real human and virtual human, the virtual human real-time motion control model by using dynamic constraints was built, and human driving with individuation was realized. Aiming at the problem that the driving accuracy was influenced by driving noise, a noise filtering rule was established based on the body joints' limit angles and the ranges of the joints angles between current frame and previous frame. A noise compensation model based on Grey System Theory (GST) was introduced to realize online compensation for filterable missing information. An automobile engine's virtual assembly was taken as an example to analyze the real-time driving accuracy of virtual human, and the result showed that the proposed solution could meet the driving accuracy and real-time demand.

Key words: virtual human, motion control, motion capture, driving accuracy, simulation, automobile engine

摘要： 为了提高复杂产品装配、拆卸操作仿真中虚拟人的驱动精度且满足交互实时性,在分析影响虚拟人实时驱动精度因素的基础上，系统地提出虚拟操作仿真环境中满足虚拟人实时驱动精度的解决方案及其关键技术。针对真人与虚拟人的尺寸差异影响驱动精度的问题，提出基于关节旋转信息,采用动态约束方式建立虚拟人实时运动控制模型，实现了个性化人体的驱动；针对实时驱动过程中驱动噪声导致人体动作失真进而影响驱动精度的问题，提出以人体关节极限角度和设置前后关键帧关节角度变化幅值为判断准则，对驱动噪声进行过滤；随后引入灰色系统理论建立驱动噪声补偿模型，实现对被过滤掉的缺失数据进行在线补偿。以某型发动机的虚拟装配为例，分析交互装配过程中虚拟人的实时驱动精度，结果表明上述方案满足驱动精度要求和交互实时性要求。

关键词: 虚拟人, 运动控制, 运动捕获, 驱动精度, 仿真, 发动机