耦合行人博弈行为的人群安全疏散策略研究

The game behaviors analysis of crowd evacuation in the large-scale scenarios is advantageous to propose effective inductive strategies. A plenty of evacuation strategies researches focus on limiting the crowd flux and the passing time, instead of adjusting the congest situations before the crowding appears. In fact, the pedestrians, who take part in the large-scales activity, could cognitive and bear that the congestion degree in activity scene is larger than the ones in normal locale. In this project, we focus our research on investigating the relationships between game behaviors and the pushing interactions among pedestrians in the high crowded; the relationships could be the inherent influence mechanism to incur the stampede. Finally, based on the internal mechanism of pushing and game behaviors, we design the safety evacuation strategies. On the one hand, to analyze game behaviors among pedestrians how to influence the pushing forces when the crowd is in the overcrowding situation, we propose a cost potential field cellular automata model with game factors and a pushing force field, in order to simulate the evacuation dynamics, and investigate the microscopic interactions among pedestrians resulted from the game behaviors. On the another hand, using this proposed model, we detect and forecast the pushing forces among the crowd instantaneously, before the pedestrian who keep still for a certain time will change his or her mode blindly, based on the globally optimal way, we design that the managers will get involved and guide the pedestrian to decrease the experienced pushing forces from non-cooperative mode, in detail, the pedestrians who could be faced with the large pushing forces should change the cooperative or non-cooperative situation, and follow the rational paths. In this project, we design the pedestrian safety measures with considering self-organization learning behaviors and game behaviors among pedestrians, and verify the feasibility of these moving strategies according to the simulations using the proposed model and the analysis of field records.

在大型集散场地人群安全疏散问题中考虑行人博弈行为的研究工作有助于提出有效诱导策略。已有策略研究的重点以控制行人流量和出入时间为主，对人群集聚前的动态控制不够充分。参与大型活动的人群对现场的拥挤现象均有一定认知，且具备比日常行走时更大的心理忍受力，本项目将研究高密集度人群中个体博弈心理行为与挤压承受力之间的变化对人群挤压踩踏事故发生的影响作用机制，用于提出安全通行策略。一方面建立耦合行人博弈行为特征的费用势函数场元胞自动机模型，模拟人群疏散动态，探究行人心理相互影响下的微观挤压作用机理。另一方面监测模型参数实时分析行人承受推挤作用力强度，在行人无法承受巨大作用力而盲目改变博弈状态前，以群体最优为主导，合理引导行人通过改变合作或对抗的状态，减少对抗引发的挤压作用，继续向目的地前进。本项目以人群的自组织学习、博弈行为作为行人安全疏散设计的前提，利用数值模拟结合现场视频数据分析提出安全疏散策略。

本项目考虑到不同场景高密集度人群疏散中的安全隐患，通过建立耦合行人博弈行为的加细网格势函数场元胞自动机模型，模拟人群疏散动态，研究高密度人群安全隐患的控制方法。主要内容与重要结果数据包括三个方面。第一，建立势函数场元胞自动机模型，用于模拟单出口、双向通道、含瓶颈出口、含斜坡通道人群疏散。模拟分析得到：单出口场景中设计出口宽度时并非越宽越好，要考虑场景尺寸与容量；相向行走的两组人群在中垂线附近相遇后出现位置抢占冲突，但行人进入通道的随机方式破坏系统的对称性，在入流密度接近0.056ped/s/m时，人群出现自组织成行现象而减少冲突，有利于提高通行流量；含瓶颈出口通道场景研究中侧重考虑聚集挤压作用效果，得到流量的第二波峰；含斜坡通道疏散的初始密度大于3ped/m^2、倾斜角大于20度时，倒地人数达到极限值；在考虑邻居结对行走方式后，可减少倒地人数近90%。第二，建立耦合博弈行为的势函数场元胞自动机模型，应用于模拟单出口人群疏散。采用囚徒困境博弈机制解决位置抢占冲突，采取演化博弈中的学习机制或自我判断更新行人博弈状态。模拟得到：行人博弈参数取值接近1时，行人倾向于选择自我判断，采用合作状态参与博弈冲突，有利于减少拥堵时间；相反，行人缺少经验，表现出学习他人的从众心理行为，一定程度上阻碍人群安全疏散。第三，在模型中考虑细化网格，通过设计参数与变量，量化安全疏散方法。在模型中加细行人占据的离散空间网格，允许邻居行人共用边界小网格，再现挤压作用效果。随着场景长度增加，可模拟的最大密度接近9.375ped/m^2。通过设置密度阈值（如9ped/m^2）并预测每个位置达到密度阈值的累计时长，设计高密度区域预警机制，及时疏导，避免挤压踩踏事故。本项目通过量化行人个体微观心理与模拟宏观疏散动态，利用量化参数设计安全疏散策略，为实际人群疏散指导的应用推广提供必要的理论依据。