【作者】 沙云飞；

【导师】 史其信；

【作者基本信息】 清华大学 ， 土木工程， 2008， 博士

【摘要】 我国文化、体育方面的大型活动日益增多,同时,火车站、机场等大型城市枢纽的交通压力也日益增大。由于在建筑设计、行人组织、紧急事件管理等方面的应用价值,人群疏散的特点和规律日益受到学术界的关注。行人行为模型是描述行人的交通行为过程的模型,是进行不同时空尺度的行人仿真、人群疏散仿真的算法核心。本文通过“行人运动实测研究-行为特点总结-行为建模-仿真建模-模型验证-模型应用”的思路,对行人运动和人群疏散进行建模和仿真研究。论文回顾了行人行为和人群疏散的研究现状,系统地总结了基于物理力的模型、基于元胞自动机的模型的基本原理、假定条件、核心方法和发展沿革,指出了现有模型的不足,通过实测研究行人的运动规律,提出了基于精细网格划分、人工势能场的微观行人仿真模型。通过视频采集行人运动,使用视频处理和坐标转换技术得到真实行走空间内的行人运动轨迹,通过对行人运动轨迹进行分析,可得到行人运动参数。在分析了行人的行为特点、行人运动参数之间的关系和影响行人行为的因素的基础上,论文提出了基于精细网格划分、人工势能场的邻域决策模型来描述行人的行为,给出了在考虑多种因素影响下的单人决策模型。针对从单人决策到集体决策的问题、叠加势能场的定义和计算方法等问题,给出了解决方案。精细网格划分是指对行走空间进行10cm×10cm元胞划分,相对于传统的离散模型,这种描述方法加强了对建筑物空间的描述能力,能够更精确地描述行人的体型和位姿。采用人工势能场机制,能够考虑到达目的地的最短路、障碍物、其他行人以及特殊事件对行人的影响。该模型的优点是能够将基于全场景的宏观层面上的路径规划决策和基于邻域的微观层面上的单步决策有机地结合在一起。并且,采用人工势能场机制描述各种因素对行人行为的影响,可以灵活地增加考虑的因素,有利于模型的扩展。论文在基于精细网格划分、人工势能场的行人行为模型的基础上,提出了人群疏散的仿真系统框架。仿真流程主要包括四个步骤:场景建模、势能场生成、仿真运行、结果输出与分析。在Microsoft Visual Studio 2005的开发环境下使用Visual C++语言进行开发,编写完成了初步的人群疏散仿真软件EVA 1.0,对模型和仿真软件的应用进行了说明,通过仿真算例证明行人行为模型和疏散仿真软件具有良好的应用效果。更多还原

【Abstract】 The cultural and physical activities are more and more in China. The traffic pressure of train station and airport are increasing. The flow pattern and characteristics of crowd evacuation is attracting increasing research efforts because of its important implications in the fields such as architecture design and emergency management. The pedestrian behavior model describes the process in which pedestrians make decisions in traveling. This model is a core algorithm, based on which we simulate pedestrian traffic and evacuation in different temporal and spatial dimensions. This dissertation aims at modeling and simulating the movement and evacuation of pedestrians, by following the processes of observing pedestrian movements, analyzing data, modeling behaviors, building simulation model, validating the model and applying in practice.First, this dissertation reviews existing literature on pedestrian modeling and evacuation studies. It systematically summarized the basic theories, assumptions, core algorithms, and evolution of models based on physical theories and cellular automata, respectively, and pointed out the insufficiencies of existing models. This study analyzed the characteristics of pedestrian behavior and factors affecting their behavior, using video data collection, coordinate transformation, and analysis of pedestrian trajectories. The major influential factors considered in this dissertation include the destination, obstacles, other pedestrians, and special events. Based on findings from this analysis, this dissertation proposed a behavior model based artificially potential field and fine grid division. The fine grid division method divides pedestrian space into 10cm by 10cm grids, which enhanced our ability to describe the space in buildings, as well as pedestrians themselves compared to traditional disaggregate models. By adopting the method of artificial potential field, this paper combined the macroscopic route planning in the strategic decision-making and microscopic decisions (e.g. avoiding other pedestrians) in the field. This research method could consider different factors with great flexibility, which provides significant potentials for further expansion. This dissertation finally established a four-step framework for simulating pedestrian evacuation based on the behavior model, including space modeling, potential field construction, simulation, and interpretation of outputs. A software package, EVA 1.0, was developed to simulate evacuation, using Visual C++ under the environment of Microsoft Visual Studio 2005. The simulation experiments reveal that both the behavior model and the simulation software package are effective in field studies.更多还原