Analyzing stripes in crossing pedestrian flows using temporal matrices and a geometric model

Abstract

Understanding pattern formation in crossing pedestrian flows is essential for analyzing and managing high-density crowd dynamics in urban environments. This study presents two complementary methodological approaches to detect and characterize stripe formations, an emergent structure observed when two pedestrian groups cross at various angles. First, we propose a matrix-based method that utilizes time-resolved trajectory data to determine the relative crossing order of pedestrians from opposing groups. By identifying points of minimal spatial separation between individuals and analyzing associated time differences, we construct a crossing matrix that captures the sequence and composition of stripes. Second, we introduce a geometric model based on elliptical approximations of pedestrian groups, enabling analytical prediction of two key macroscopic quantities: the number of stripes and the interaction time between groups. The model captures how these quantities vary with the crossing angle and shows strong agreement with experimental data. Further analysis reveals that group elongation during crossing correlates with the vertical cross-section of the elliptical shape. These methods provide effective tools for analyzing large-scale movement datasets, informing the design of public spaces, and calibrating mechanistic models. The study also presents hypotheses about pattern transitions in continuous pedestrian streams, suggesting promising directions for future research on collective motion under varying flow geometries and densities.