基于行人占据区域的泰森多边形密度统计方法及其模型研究

发布时间：2018-01-22 08:19

本文关键词： 行人动力学泰森多边形人际距离前向面积虚拟样点速度-密度关系行人流模型　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：行人动力学的研究一直是火灾等紧急情况下人员疏散研究的重要内容。人员疏散的实验研究和模型研究一直是行人动力学中主要的方向。研究者通过大量观测和可控实验,获得行人的运动行为特征,揭示人员运动规律,但是仍然缺乏足够的定量校准的实验数据。本文在总结了前人对于实验中全局密度与局部密度统计方法的基础上,基于泰森多边形方法,结合人际关系距离与视野的概念,提出了一种行人占据多边形的速度密度统计方法。并基于此方法提出了与行人样点属性不同的虚拟样点,来代替边界等障碍物。通过对平均密度平均速度的比较发现,本文方法得到的结果比传统统计方法得到的结果波动更小且更接近于连续变化、在密度低时比泰森多边形方法的结果更符合实际情况;而对于个人的速度密度散点图,本文方法的结果也比泰森多边形方法的结果更为集中,尤其是除去了密度小而速度低的不合理数据点。通过对行人前向120°视野范围内面积来计算行人的局部密度,并与泰森多边形、行人占据多边形计算的局部密度相比较,发现采用前向120°方法得到的密度能更好地符合行人速度密度基本关系。并基于此得到了不同的局部密度情况下行人的速度分布以及行人局部密度速度的关系,丰富了行人运动数据库。本文借鉴了交通流中的跟驰模型及行人运动方向变化机制,根据行人占据多边形的特点,提出了基于行人占据多边形的行人运动模型。利用该模型对单通道和存在瓶颈的单通道场景进行了模拟。模型中将前向120°视野范围分成了三部分——正前方、左前方、右前方,提出了两种方向确定方式:1)选取三部分中加权面积最大的方向,2)根据三部分加权面积的大小,以概率方式选取方向。同时速度大小也有两种确定方式:A)根据前向120°视野范围的面积得到局部密度来计算速度,B)根据选取方向对应面积得到局部密度来计算速度。两两组合共有四种方案。经过对单通道场景的模拟发现,1)B)组合的结果最符合实验结果,说明大部分行人在行走时都是以1)B)方式确定其运动方向和速度大小的。通过在边界上添加虚拟样点来代替边界以及形成瓶颈,对存在瓶颈的单通道场景进行了模拟,得到了拥堵和拱形分布现象。考虑到虚拟点的引入导致有效瓶颈宽度减小,我们对此瓶颈出的虚拟样点位置做出了优化,使之和实验场景相同,此时模型模拟的所有行人全部通过通道的时间和实验中的时间相差5.25%。
[Abstract]:The study of pedestrian dynamics has always been an important part of evacuation research in emergencies such as fire. The experimental study and model study of pedestrian evacuation have been the main directions of pedestrian dynamics. Controlled experiment. The characteristics of pedestrian movement behavior are obtained, and the movement law of personnel is revealed. However, there is still insufficient experimental data for quantitative calibration. This paper summarizes the previous statistical methods of global density and local density based on Tyson polygon method. Based on the concepts of interpersonal relationship distance and visual field, a new method of velocity density statistics of pedestrian occupation polygon is proposed, and based on this method, a virtual sample with different attributes from pedestrian sample is proposed. By comparing the average density average velocity, the results obtained by this method are smaller and more close to the continuous variation than those obtained by the traditional statistical method. When the density is low, the results of the method are more suitable than the results of the Tyson polygon method. The results of this method are more concentrated than that of Tyson polygon method for individual velocity density scatter plot. In particular, the local density of pedestrians is calculated by using the area within the range of 120 掳visual field for pedestrians, and the local density of pedestrians is calculated by eliminating the unreasonable data points with small density and low speed. The local density of pedestrians is calculated with Tyson polygon. The local density calculated by the pedestrian occupation polygon is compared. It is found that the density obtained by the forward 120 掳method can better accord with the basic relationship of pedestrian speed density. Based on this, the velocity distribution of pedestrians and the threshold of pedestrian local density velocities are obtained under different local densities. Yes. This paper draws lessons from the car-following model in traffic flow and the changing mechanism of pedestrian movement direction, according to the characteristics of pedestrian occupation polygon. A pedestrian motion model based on pedestrian occupation polygon is proposed. The model is used to simulate single channel and single channel scene with bottleneck. In the model, the forward 120 掳field of vision is divided into three parts-positive forward. Fang. In front of left and front of right, we put forward two ways of direction determination: 1) choose the direction of the largest weighted area in the three parts, and choose the size of the weighted area according to the size of the weighted area of the three parts. At the same time, there are two ways to determine the velocity: A.) the velocity is calculated by local density according to the area of forward 120 掳field of vision. B) according to the selected direction corresponding area to get the local density to calculate the speed. The pairwise combination has four schemes. Through the simulation of the single channel scene, it is found that the result of the combination is the most consistent with the experimental results. It shows that most pedestrians determine the direction of movement and the speed of the pedestrian in the way of 1 B. by adding virtual points to the boundary to replace the boundary and form bottlenecks. The single channel scene with bottleneck is simulated, and the phenomenon of congestion and arch distribution is obtained. Considering that the introduction of virtual point leads to the decrease of effective bottleneck width, we optimize the location of virtual sample point. The time of passing through the passageway of all the pedestrians simulated by the model is 5.25% different from the time in the experiment.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TU998.1

【参考文献】