基于人群密度的虚拟人群行为建模及仿真技术研究

项目名称： 基于人群密度的虚拟人群行为建模及仿真技术研究

项目编号： No.61300101

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 孙立博

作者单位： 东南大学

项目金额： 22万元

中文摘要： 虚拟人群仿真技术能够提供人群运动的分析和演示工具，在安全科学、体育仿真、影视动画、军事训练等领域有着广泛的应用前景，因而逐渐成为计算机图形学及其相关领域的研究热点，受到了众多学者的关注。本项目旨在反映人群密度对感知、规划和行为的影响，主要研究要点包括：(1) 研究虚拟人群获取环境信息的方法，并在此基础上使用Q学习方法研究人群密度变化、环境动态改变、个体间发生信息交流时虚拟人实时感知的方法。(2) 针对动态复杂环境，研究基于人群密度的虚拟人群全局路径规划算法；针对未知环境情况，研究基于机器学习的导航算法。(3) 研究个体的行为决策和推理机制，着重模拟虚拟人与环境的交互行为和闲置行为，以体现个体行为的多样性；并在此基础上研究群体的信息交互机制，以体现群体行为的社会性。(4) 构建虚拟人群行为控制框架，实现对人群行为的自主控制；并从面向用户交互的角度，实现用户对人群行为的高层控制。

中文关键词： 虚拟人群仿真；人群密度；感知模型；路径规划；机器学习

英文摘要： Crowd simulation can provide a tool for analyzing and displaying the crowd movement and has a wide application in safety science, sports simulation, movie-TV animation and military training. Therefore, crowd simulation becomes a hot research topic in computer graphics and related research areas and attracts great attention of many scholars. This project aims at reflecting the influence of the crowd density on the perception, the path planning and the behaviors of the crowd and as a result, the main research aspects include: (1) Study how the virtual crowd obtains environment information and study a real-time perception approach for the agent based on Q learning when the crowd density and the environment change dynamically and the information exchange happens between the agents. (2) The global path planning algorithm based on crowd density is studied in dynamic and complex environment and meanwhile, the navigation algorithm based on machine learning is studied in unknown environment. (3) Study the mechanism of behavior selection and behavior reasoning for each agent and simulate the interaction behavior and the idling behavior to show the diversity of individual behavior and furthermore, study the information exchange mechanism between the groups to show social and interactional aspects of crowd behaviors. (4) Co

英文关键词： crowd simulation；crowd density；perception model；path planning；machine learning