【作者】 王恒；

【导师】 李枫；

【作者基本信息】 同济大学 ， 工程（专业学位）， 2022， 博士

【摘要】 轨道交通以其运量大、速度快、准点率高、节能环保、对地面交通干扰小等优势,已成为我国各大城市、特大城市公共交通系统的骨干。截止2021年底,在中国内地已有46个城市开通轨道交通,运营车站达到5940座。地铁车站作为大规模乘客进出和集聚的中心节点,不仅自身空间相对封闭、结构复杂,而且常常布设在地下或高架,使得紧急情况下救援十分困难。因此对乘客应急疏散能力提出了更高要求。地铁车站乘客应急疏散能力,一般是基于各类应急疏散模型通过模拟仿真加以评估认定,但由于底层的地铁车站乘客应急疏散的心理和行为数据较难获取,同时一些影响疏散的因素也难以量化,使得现有的疏散模型往往缺少实证数据支持与验证,相应的仿真评估结果的可信性也无法保障。因此,理清底层的地铁车站乘客应急疏散时的心理与行为的量化特征显得尤为重要。基于虚拟现实技术和情景假设试验的地铁乘客应急疏散心理与行为的量化分析研究,是提高现有乘客应急疏散模型预测的准确性的基础,具有重要的理论意义和实践应用价值。本文主要研究工作和相应成果如下:首先,在确定主要的疏散影响因素的情况下,针对地铁车站乘客急疏散行为数据较难获取的情况,采用两种试验方法收集地铁车站乘客应急疏散的心理与行为数据,为后续研究提供基础的实证数据。首先利用SP-off-RP情境假设问卷调查方法收集数据,其次利用虚拟现实技术的“沉浸式”体验的优势,搭建地铁车站火灾虚拟现实平台采集数据。其次,基于SP-off-RP情境假设问卷数据和地铁车站火灾虚拟现实平台数据,运用离散选择模型中的条件Logit模型、随机参数Logit模型分析影响因素对乘客疏散决策的影响权重,量化这些因素对乘客疏散决策行为的影响。同时,引入心理学中的“艾森克人格问卷”,量化乘客的人格特征,定量研究了乘客人格特征与疏散影响因素之间的关系。再次,基于以上两种试验数据,运用随机参数Logit模型分别量化分析乘客对相关疏散影响因素的偏好异质性,从定量化角度阐述了乘客在疏散时对影响因素的感知偏差;基于以上乘客疏散时对影响因素的权重和偏好异质性的感知量化,提出了运用潜类别Logit模型,并结合EM算法,对疏散人群进行量化分类。结果显示,将疏散人群分为两种类别时模型达到最优,即“距离避远者”人群和“可见性偏好者”人群。最后,根据人群分类结果,通过Anylogic软件仿真,分析应急疏散时不同类别人群的关键疏散指标的差异,并通过仿真与疏散试验进行对比,验证了这种基于群体分类的疏散仿真模型的合理性。综上所述,论文通过SP-off-RP情境假设问卷和地铁车站火灾虚拟现实平台采集乘客应急疏散心理与行为数据,为修正疏散仿真模型提供实证数据,也扩充了国内疏散行为数据集;论文量化分析了影响因素对乘客疏散影响的权重问题,并定量化证实了乘客在疏散时对影响因素的感知偏差即存在偏好异质性,并通过这种偏好异质性将疏散人群进行有效分类;区别于以往的地铁车站疏散仿真中将人群不分类或者只进行简单分类,本文提出了基于试验数据的差异化人群分类的疏散仿真方法。最后,对本次研究之不足及进一步的研究方向进行了简要的讨论。更多还原

【Abstract】 Rail transit has become the backbone of the public transport system in major cities in China because of its advantages such as large traffic volume,fast speed,high punctuality rate,energy conservation and environmental protection,and small interference to ground traffic.A total of 46 cities in mainland China have opened urban rail transit,with 5940 operating stations by the end of 2021.As the central node for large-scale passenger access and gathering,subway stations are not only relatively closed in space and complex in structure,but also often arranged underground or elevated,making it very difficult to rescue in an emergency.Therefore,higher requirements are put forward for passenger emergency evacuation capability.Subway station passenger emergency evacuation capability is generally evaluated and identified through simulation based on various emergency evacuation models.However,it is difficult to obtain the psychological and behavioral data of emergency evacuation of passengers in subway stations,and some factors affecting evacuation are also difficult to quantify.The existing evacuation models are often lack of empirical data to support and verificate,and the credibility of the corresponding simulation evaluation results can not be guaranteed.Therefore,it is particularly important to clarify the quantitative characteristics of passengers’ psychology and behavior during emergency evacuation in subway stations.The quantitative analysis of subway passengers’ emergency evacuation psychology and behavior based on virtual reality technology and scenario hypothesis test is the basis to improve the accuracy of existing passenger emergency evacuation models.It has important theoretical significance and practical application value.The main research work and corresponding results of this article are as follows:Firstly,In the case of determining the main influencing factors of evacuation,two test methods are used to collect the psychological and behavioral data of subway station passengers’ emergency evacuation,so as to provide basic empirical data for subsequent research.Firstly,the SP off RP scenario hypothesis questionnaire method is used to collect data.Secondly,the virtual reality platform of subway station fire is built to collect data by taking advantage of the "immersive" experience of virtual reality technology.Secondly,based on the data from SP off RP scenario hypothesis questionnaire and the subway station fire virtual reality platform,the conditional Logit model and random parameter Logit model are used to analyze the influence weight of influencing factors on pedestrian evacuation decision-making,and quantify the impact of these factors on pedestrian evacuation decision-making behavior.At the same time,the "Eysenck Personality Questionnaire" in psychology is introduced to quantify the personality characteristics of passengers.The relationship between passengers’ personality characteristics and evacuation influencing factors is quantitatively studied.Thirdly,based on the above two kinds of test data,the random parameter Logit model is used to quantitatively analyze the heterogeneity of passengers’ preference for relevant evacuation influencing factors,and the perception bias of passengers on the influencing factors during evacuation is described quantitatively.A potential category Logit model combined with EM algorithm is proposed to quantitatively classify the evacuees based on the above perceptual quantification of the weight of influencing factors and preference heterogeneity during passenger evacuation.The results show that the model is optimal when the evacuees are divided into two categories,namely,the "distance avoidance" crowd and the "visibility preference" crowd.Finally,according to the crowd classification results,the differences of key evacuation indicators of different categories of crowd during emergency evacuation are analyzed through the simulation of anylogic software.The rationality of the evacuation simulation model based on group classification is verified through the comparison between simulation and evacuation test.In summary,the psychological and behavioral data of passengers’ emergency evacuation collected through the SP off RP scenario hypothesis questionnaire and the subway station fire virtual reality platform can provide empirical data for modifying the evacuation simulation model and expand the domestic evacuation behavior data set.The article analyzes the weight of the impact factors on pedestrian evacuation quantitatively,and confirms that there is preference heterogeneity in the perception bias of pedestrians on the impact factors during evacuation.Evacuees can be effectively classified through this preference heterogeneity.This article proposes an evacuation simulation method based on the differentiated crowd classification of experimental data,which is different from the previous subway station evacuation simulation in which the crowd is not classified or simply classified.Finally,the deficiency of this study and the further research direction are briefly discussed.更多还原