贪心选择在地铁站内行人疏散中的应用

Greedy-based pedestrian evacuation planning applied in subway station

中文关键词  疏散策略  出口分配  半仿真模型  贪心选择策略

英文关键词  evacuation planning  exit distribution  semi-simulation model  greedy strategies

中文摘要

      大型室内场所空间布局较为复杂(如地铁站), 人群的高密度聚集往往存在一些潜在的风险. 本文在分析国内外人群快速疏散研究现状的基础上, 提出了一种基于贪心选择的行人疏散方法. 该方法以地铁站内复杂场景作为研究背景: 首先, 针对地铁站内的行人的行动轨迹难以获取问题, 本文利用地铁站内行人真实出站数据, 基于元胞自动机, 构建了行人疏散轨迹半仿真模型, 并利用实际流量数据优化该半仿真模型; 其次, 基于该轨迹模型, 为了满足高动态场景中的实时性, 采用复杂度较低的贪心选择策略分配最优疏散出口; 最后, 以杭州武林广场地铁站为例, 使用真实出站数据设计对比实验, 验证行人轨迹模型的有效性以及出口分配方法的性能. 结果表明, 本文所提出的行人轨迹模型能够较好的模拟行人的轨迹, 仿真中各出口疏散人数同真实出站数据拟合程度的可决系数R2达到了0.67. 相较于最短路径和最短时间出口分配方法, 本文所提出的方法在整体疏散效率上分别提高了27.2%和16.5%.

英文摘要

      The layout of the large indoor building is usually complicated (such as subway stations), and the dense crowd often brings up some potential risks. By reviewing and analyzing the previous evacuation studies, we propose a pedestrian evacuation planning based on a greedy strategy. This method is mainly applied in a complex subway station: Firstly, due to the pedestrian tracking data being difficult to obtain, we employ a cellular automata to analyze the real exit flow data of subway station, construct a semi-simulation model of pedestrian evacuation trajectory, and then continuously optimize the semi-simulation model with updated data. Secondly, based on the semi-simulation model, to meet the real-time requirement for high dynamic evacuation, we use a lightweight greedy strategy to optimize exit distribution. Finally, taking the Hangzhou Wulin square station as an example, we conduct a series of experiments to validate the semi-simulation model and test the performance of evacuation planning. The results show that the semi-simulation model can effectively simulate the pedestrian trajectory. The coefficient of determination R2 arrives 0.67. In addition, compared with the exit distribution methods based on the shortest path and shortest time, the greedy-based distribution improves the overall efficiency of evacuation planning by approximate 27.2% and 16.5%, respectively.