有限理性人群拥挤致灾机制与管控方法研究

Trampling disaster happened frequently at mass gatherings, posing a substantial threat to public safety. Studies on these disasters have indicated that when facing complicated environment, pedestrians show limited ability to access and analyze information. Thus their decision-making behaviors including route choice are not completely rational. Considering the fact that pedestrians in massive crowds have bounded rationality, we in the present study firstly design and perform individual decision-making experiments with focus on the influences of uncertainty on route-choice behavior. Based on prospect theory, boundedly rational route choice model would be proposed to describe individual decision making process when facing risk and uncertainty. Secondly, we will establish and improve a virtual reality collective decision-making behavior experiment platform. On this platform, real humans will be invited to participate in structured experiments. In these experiments, behaviors including route choice, herding will be analyzed by considering the influences of other individuals, groups and crowds because they share the same constraint conditions such as road resources. Various decision-making behaviors and patterns under different configurations of occupancy, view range, can be modeled and quantified in the framework of game theory. Day-to-day repeated route choice pattern and on-off route choice pattern would be investigated. With the insights into collective decision-making pattern, dynamic pedestrian assignment method can be formulized. Further, video recordings and other field survey data would be employed to empirical explore pedestrian activity pattern in massive gathering events. With the obtained activity pattern, decision pattern and previously obtained pedestrian movement pattern, a network-field coupled simulation model for boundedly rational pedestrian flow can be formulized. Model calibration and validation will then be performed. Case study and numerical analyses will be carried out to investigate the influence of boundedly rational decision-making on macroscopic spatiotemporal crowd distribution as well as on how forces cumulate and spread among individuals. Critical parameters will be identified and selected as indicators of crowd trampling risk. With CCTV surveillance cameras, pedestrian detection method will be developed to provide monitoring information on pedestrian flow dynamics. These information can then be used as inputs of the developed network-field coupled simulation model. Effects of different management methods, different intervention strategies and varied guidance provided can be simulated to predict and evaluate pedestrian flow state. Thus, the project is expected to provide insights not only in developing a safe and robust crowd management plan by the proposed boundedly rational crowd simulation model, but also in enhancing crowd safety by constructing a crowd risk evaluation system.

人群集聚活动中拥挤踩踏事故频发，对公共安全造成重大威胁。分析表明，面临复杂环境条件时，人群表现出有限的信息获取与分析能力，因此，其路径选择等决策行为并非完全理性。本项目拟设计并开展真人参与的可控实验，辅以事故资料收集，分析不确定性风险下个体决策行为规律，提出有限理性个体决策模型；开发完善群体决策虚拟实验平台，研究路网容量等约束条件下群体博弈过程，揭示收益冲突条件下群体决策模式，发展有限理性用户均衡模型，提出动态人群配流方法；最终结合人群迁移模式实测资料以及个体运动模式工作基础，建立基于智能体的“场—网”复合模型，研究事故案例，验证模型，探讨有限理性决策模式导致宏观人群时空分布不均、微观个体间作用力累积突变的临界条件，确定人群拥挤灾变风险判定指标，耦合视频监测技术，建立拥挤人群灾变风险预警系统，发展拥挤人群诱导与管控方法，为拥挤人群活动的规划、管理与应急处置提供基本方法、基础数据和科学依据。

人群集聚活动中拥挤踩踏事故频发，对公共安全造成重大威胁。分析表明，面临复杂环境条件时，人群表现出有限的信息获取与分析能力，因此，其路径选择等决策行为并非完全理性。.针对这一问题，本项目持续收集密集人群事故资料，设计并开展真人参与的现场疏散实验，分析不确定性风险下个体行为规律，提出有限理性个体决策模型；开发完善多人在线疏散虚拟实验平台，研究路径信息、社会信息影响下的群体决策模式，发展有限理性用户均衡模型，提出动态人群配流方法；最终结合实测资料、实验数据以及个体路径选择研究成果，建立基于智能体的“场—网”复合模型，研究事故案例，验证模型，探讨有限理性决策模式导致宏观人群时空分布不均、微观个体间运动行为突变的临界条件，确定人群拥挤灾变风险判定指标，耦合视频监测技术，建立拥挤人群灾变风险预警系统，针对地铁车站、大学校园等典型人员密集场所，结合疫情管控的实际需求，发展拥挤人群诱导与管控方法，为拥挤人群活动的规划、管理与应急处置提供基本方法、基础数据和科学依据。.项目执行期间，项目组与本领域内国内外知名研究团队开展了广泛的交流与合作，主办、协办领域内著名国际会议3次，组织并参加了一系列的学术活动，在国际国内核心学术期刊发表学术论文17篇，其中SCI检索13篇，EI检索2篇（不重复计算），受邀在5个国际/国内会议做邀请报告，申请并获批人员安全方面的发明专利4项、实用新型专利1项，软件著作权3项。培养2名博士研究生，17名硕士研究生，1人获“中国消防协会优秀博士学位论文”，建立并形成了稳定的研究团队，围绕行人流和人员疏散动力学开展深入的科学研究。.