大规模行人流实验和基于大数据的人群疏散仿真

为了减少人群踩踏伤害事故，保障公共安全，有必要研究真实环境中大规模人群的运动规律和应急疏散动力学，新兴的大数据收集与分析技术为此提供了新契机。本项目拟组织大规模行人流实验，建立基于背景场的微观连续模型，据此研究行人自组织现象，并与实验结果进行对比。在广场、地铁站、道路拥挤地段等场所进行大量行人交通实测，积累海量数据，从中提取不同类型行人的运动特征；以节假日群众聚散、大型群体活动为研究对象，从大数据中提取相关信息。此外，还将通过数值模拟产生与真实大数据性质类似的虚拟大数据。基于大数据重构特定区域的人群分布，提取OD信息，模拟真实人群的聚集、拥堵和疏散过程。利用模拟结果预测易发生踩踏事故的区域，提出应对预案，为有效疏散密集人群和减少踩踏事故的发生提供理论依据和可行方案。

In order to reduce stampede accidents, more attention should be paid on large-scale crowd movements in the real environment. Techniques of collecting and analyzing big data offer the new possibility of such kind of research. In the project, large-scale experiments on pedestrian flows will be conducted. A kind of more realistic microscopic continuous model is proposed, which is based on the floor field and calibrated by the experimental data. Then the model is used to investigate self-organization phenomena of pedestrians and its numerical simulations are compared with experimental results. Extensive observations on pedestrian flows in plazas, subway stations, places of crowd gathering, etc., are performed, related big data are collected and movement features of different types of pedestrians are analyzed. With consideration of these features, the presented model for pedestrian flows is able to depict the movements of the crowd in reality. We take holidays and large-scale group events as objects and extract information related to these activities from big data. In addition, virtual big data can be generated by numerical simulation, which is similar to real big data. Then the distribution of pedestrians is reconstructed by big data and the OD information is applied as well. The aggregation, congestion and dissipation of the real crowd can be simulated by the presented model. Possible stampedes can be predicted by using these simulation results, hence early preventive measures can be taken in time. This research provides theoretical and practical basis to evacuate the crowd and reduce the occurrence of stampedes.