Social groups in pedestrian crowds as physical and cognitive entities: Extent of modeling and motion prediction

Most pedestrian crowds are composed of social groups that are typically formed by dyads (two members), triads (three members), or larger groups. Depending on the context, social groups may make up half or even more of the membership of the crowd. Therefore, understanding their motion is crucial for predicting crowd dynamics. The presence of social groups modifies crowd behavior. When the proportion and size of groups are known, crowd motion (e.g., the “flow” of passengers collectively moving inside a train station) could become predictable. However, a bidirectional flow experiment performed in 2010 revealed that the presence of groups could lead to partially surprising results because crowds composed of small social groups moved more smoothly than those composed of individuals (singletons). Results were partially disregarded because of statistical insignificance. A subsequent experiment in 2015 with latest tracking techniques resulted in similar results and investigated the cause of the superior flow in the presence of groups. The results revealed that when groups arrange themselves in certain shapes, their partially “obstructing” nature (in a counterintuitive manner) facilitates lane formation, which benefits overall crowd motion. Because the arrangement of a dyad, i.e., whether both members walk next to each other or in a front–back alignment, is partially linked to the coordination (or the lack thereof) between both members, predicting such a mechanism is difficult. Simulation results from a commercial software program confirmed that predicting the dynamics of social groups is not trivial; however, at the macroscopic scale, some general trends are depicted at least from a qualitative perspective. This study revealed that, whenever possible, several crowd composition patterns should be considered when planning crowd events or drafting safety guidelines for pedestrian facilities. Depending on the context, crowds composed of individuals may move smoother than social groups do, and the worst-case scenario should be used for determining safety margins. Thus, we revealed that predicting the motion of crowds composed of social groups is difficult because the microscopic organization within the group determines overall crowd dynamics. Although this internal organization may result in counterintuitively efficient group structures, the occurrence of such conditions depends on several variables, which renders crowd control in social groups complex, requiring close monitoring especially at high densities.