数据和模型混合驱动的虚拟人动作姿态快速生成与交互控制技术研究

项目名称： 数据和模型混合驱动的虚拟人动作姿态快速生成与交互控制技术研究

项目编号： No.61070141

项目类型： 面上项目

立项/批准年度： 2011

项目学科： 金属学与金属工艺

项目作者： 梁加红

作者单位： 中国人民解放军国防科学技术大学

项目金额： 10万元

中文摘要： 虚拟人特定动作姿态的快速生成与实时交互控制是各种虚拟人应用系统的公共支撑与关键核心技术，针对当前虚拟人特定动作姿态生成所面临的工作繁杂、所需的数据样本规模过大、可控性和适应性差等难题，本项目以虚拟人特定动作姿态的快速生成与实时交互控制为研究内容，深入探讨人体的动作生成机理，采取数据和模型混合驱动的策略，直接基于已有动作数据样本和仿真应用系统对虚拟人动作姿态的多个具体高层任务约束，依据优先级层次控制策略，在不同任务约束对应的冗余空间上开发具有任意多个优先级的动作姿态合成与交互控制算法，快速生成各种逼真的人体动作姿态，并对动作合成算法生成的各种人体动作姿态的逼真、自然程度进行定量评估，提高相关应用系统的开发效率、逼真度和交互性，为虚拟人仿真应用系统开发提供关键技术支撑。

中文关键词： 虚拟人；高层任务约束；冗余空间；优先级层次控制；数据和模型混合驱动

英文摘要： The quick generation and real-time interactive control of the virtual characters' custom motion posture is one of the key techniques of various virtual characters application systems. Currently the motion generation of virtual characters is encumbered by some difficulties such as tedious and time-consuming labor, large scale of sample data, poor controllability, bad adaptability etc. This project mainly studies the quick generation and real-time interactive control of the virtual characters' target motion posture, and explores the human's mechanism of the motion generation deeply with the hybrid data- and model-driven strategy. Based on the multiple high level task constraints of motion posture of virtual characters by recorded motion data samples and simulation application system, this project will adopt the priority level control strategy, present an motion posture integrated and interactive control algorithm, which is in the redundancy space of different task constraints and has multiple prorities, quickly generate various body motion postures and quantitatively evaluate the fidelity and naturalness of the motion postures generated by the motion integrated algorithm. Consequently, the project will enhance the developing efficiency, fidelity and interativity of the application system, and provide key techniques for the simulation application system of virtual characters.

英文关键词： virtual characters; high level task constraint; redundancy space; priority level control strategy; data- and model-driven