【作者】 黄希发；

【导师】 王科俊；

【作者基本信息】 哈尔滨工程大学 ， 模式识别与智能系统， 2010， 博士

【摘要】 由于大型场馆内的人员聚集带来的疏散效率降低和踩踏风险加大,人员疏散已成为在大型场馆内组织开展活动需要首先予以关注的基本问题。计算机仿真是开展人员疏散研究的主要手段,人员疏散微观仿真模型因其对个体属性、心理、行为等方面强大的描述能力,以及便于研究者对人与人、人与环境之间的关系进行刻画,成为该领域的研究热点。本文的研究目的就是通过人员疏散微观仿真模型的建立,重点研究疏散过程中人的决策机理,再现和解释现实的疏散情形,为大型场馆疏散设计和人群管理提供参考。基于人是公认的最佳“控制器”,本文提出将个体疏散行为看作控制问题进行建模,即将控制领域的两大思想(反馈和优化)引入人员疏散的研究中。反馈机制的引入能够有效降低对包括被控对象、外部环境等在内的控制系统建模精度要求,本文结合精细描述个体行为的Agent建模仿真方法学,以及粗糙描述外部环境的元胞自动机建模仿真方法学,对人这一复杂适应性行动主体分别从感知、决策和行动等环节进行建模,并对大型场馆这一复杂结构进行建模。论文按个体决策方式的不同提出五类个体疏散模型,并针对常见的遵循正常、拥挤、单步预走和多步预走等四类疏散模型的个体分别进行Agent建模。首先根据常识推断个体可能采取的疏散行为和整体的疏散情形,然后据此提出该类个体的疏散模型框架和具体的建模方法。在个体建模中,个体感知信息主要包括当前邻域内各网格到各出口的距离、视野范围内的人数和障碍及二者的密度,这是影响个体决策的最主要因素。个体正常疏散模型给出了个体依据上述影响因素寻优的过程;个体拥挤疏散模型在正常疏散模型基础上引入了个体的竞争能力、承受能力、承受极限等与拥挤相关的概念,并对由此产生的个体摔倒或伤亡等现象进行建模;个体单步和多步预走疏散模型在正常和拥挤疏散模型基础上引入了预走的概念,并由此引入了通过评判预走后的情形自适应调整疏散策略的建模思路。根据大型场馆的空间特点和个体的疏散模型,本文编程实现仿真平台,对每类疏散模型进行仿真分析,并对遵循不同疏散模式的人群的组合进行混合仿真。仿真结果表明,论文提出的建模框架、思路和方法具有可行性,模型具有很强的现实描述能力,能够再现和解释现实的疏散情形。更多还原

【Abstract】 Due to the high density of people gathered inside the same stadium or gymnasium, which cause the low efficiency of evacuation and the lurking risks of stampede, pedestrian evacuation has already become the number one concern in organizing any activity in the large-scale stadiums and gymnasiums.The purpose of this thesis paper is to study, on the basis of a micro computer simulation model of human evacuation, the decision-making mechanism of human being in the process of evacuation by reproducing and explaining an evacuation scenario in reality. By doing so, the paper aims at serving as a reference to the designing and management of large-scale stadiums and gymnasiums.Based on the universal acknowledgement of human-being as the best“controller”, this paper proposes to model the individual behavior of evacuation as a controlling process. Then the two major concepts in controlling can be introduced into the study of human evacuation, namely the concepts of feedback and optimization.The introduction of the feedback mechanism can effectively lower the requirements of accuracy in the modeling of the control system, which includes the modeling of the object being controlled and that of the external environment. This paper combines the simulation methodology of Agent modeling to accurately describe individual behaviors and the simulation methodology of Cellular Automata (CA) to roughly describe the external environment. Based on the combination, the paper constructs the model for human being, the principal of action with complicated adaptability, on the different stages of“feeling, decision-making and action”and constructs the model for the complex of large-scale stadiums and gymnasiums.Based on the difference of individual decision-making, this paper proposes five individual models of pedestrian evacuation. These models represent five categories of individual in the real life. It is fair to say these five groups of people almost represent all types of people in real life. In accordance with the five groups of people with five methods of evacuation, the paper constructs four evacuation models for the normal, crowded, mono-step and multi-step scenarios. To construct each and every model, the paper predicts the individual’s potential action and the general evacuation movement out of common sense. Then it proposes the framework of modeling and the specific modeling method. In the individual model, the instant information he/she obtains at a certain moment during the evacuation includes the distance between every neighboring grid to the exit, the number and intensity of people and obstacles within his/her instant sight, which are all critical factors for the individual’s decision-making process. In the model of normal evacuation, the individual seeks optimization of evacuation pathway on the basis of the afore-mentioned factors. In the model of crowded evacuation, the concepts of individual competitiveness, endurance capacity, endurance limit, crowdedness and other related concepts are added, based on which the sub-models for individual’s tumbling, being hurt or killed are consequently constructed. To construct the model of mono-step pre-movement and multi-step pre-movement, the concept of pre-movement is introduced into the previous two models, in which the individual’s adaptive adjustment to his/her evacuation strategy after making prejudgment of the consequence of pre-movement is also considered.Based on the different simulation models as previously mentioned, the paper constructs, using Java language, the simulation platform of PESS. The research analyzes the simulation result of the four different models at the four typical areas and also conducts mixed simulation for the mixed group of different types of people. The result shows that the modeling framework, principles and methods proposed in the paper are all feasible. The models have strong ability to describe, reproduce and explain the reality of evacuation.更多还原