Walking behavior of pedestrian social groups on stairs: A field study

Highlights

• A field study was conducted to explore group walking behavior on stairs.

• Average distances for groups of different sizes are relatively stable.

• Groups can form a flexible structure even under the effect of stair step constraint.

• Members belonging to the same group seek to move at similar speed and offset angles.

• Pedestrian characteristics can impact on group walking speed and dissimilarity.

Abstract

There is growing consensus that most pedestrians in a crowd walk in groups, and these groups have an influence on crowd dynamics whenever under natural or emergency situations. Despite this fact, however, empirical data on group walking behavior on stairs are limited. This paper aims to explore how group members interact with each other and how groups organize on stairs. In order to achieve these, a field study was conducted on stairs in a campus of a university. A total of 105 pedestrian groups descending the stairs were selected as observation subjects. After obtaining their trajectories by the optimal flow algorithm, interpersonal distances and angles between group members, average speed, space-time diagrams offset angles and walking dissimilarity on stairs were calculated and analyzed. Results demonstrate that average distances for groups of different sizes are relatively stable. Groups with different sizes on stairs can form different walking patterns. Members belonging to the same group always seek to move at similar speed and offset angles, and can adjust their motion during the descending process, in order to ensure comfortable group walking and coherence. It is discovered that group size has a significant negative effect on average group speed. Gender and social relationships have little influence on group walking speed and dissimilarity values, except in dyads. These findings have implications for group walking modelling and safe pedestrian facility design.

Introduction

With the increasing number of mass events or emergency situations in recent years, how to ensure crowd safety has drawn significant interest and attention. Human behavior and evacuation dynamics are becoming important research objectives. In fact, social interactions among pedestrians play a role in the complex dynamics of crowd behavior. It is necessary to investigate them for a deep understanding of crowd dynamics.

Pedestrian social groups (von Krüchten and Schadschneider, 2017) can be always observed in our daily life. Persons may choose to conduct activities (such as shopping, working or travelling) with their family members, friends, colleagues, etc. Here, pedestrians in a small group are those who have social relationships, and walk together (Moussaïd et al., 2010). The walking patterns of groups include walking side-by-side, the ‘V’ shape in triads, ‘U’ shape in four-person groups and ‘river-like’ formation (Moussaïd et al., 2010, Karamouzas and Overmars, 2012). These walking patterns are affected by crowd density (Moussaïd et al., 2010, Zanlungo et al., 2015, Zanlungo et al., 2017, Zanlungo and Kanda, 2015). It was reported that 70% of pedestrians walked in a group consisting of two to four members (Moussaïd et al., 2010). Pedestrians in a group always seek to adjust their speed, movement direction and interpersonal distances in order to maintain the group formation (Zaki and Sayed, 2018). Considering the large number of pedestrian social groups in a crowd, researchers begin to focus on small pedestrian group behavior.

Both models and experiments have been used to investigate pedestrian social groups (Cheng et al., 2014). Models, such as agent-based models (Fang et al., 2016, Bandini et al., 2014, Gorrini et al., 2018), game-theory models (Huang et al., 2015, Wang et al., 2015), cellular based models (Pereira et al., 2017, You et al., 2016) and physical based models (Singh et al., 2009, Zhang et al., 2018), have been modified by involving grouping behavior. In agent-based models, pedestrians are endowed with different characteristics, and their decision making can be affected by personal desires (Fu et al., 2016). However, the models are usually computationally expensive than others. In game-theory models, pedestrians evaluate all available chances, and choose one to ensure that its utility is maximum. Everyone’s final utility payoffs will rely on the options selected by all pedestrians (Zheng et al., 2009). The game-theory models have been employed to simulate the exit selection behavior. Cellular based models simulate pedestrian movement in discrete space and time according to transition probabilities and several simple rules. They are the most widely adopted to study pedestrian dynamics, especially in complex scenarios (Fu et al., 2016). Physical based models involve optimal acceleration and physical forces among pedestrians, such as the magnetic model (Okazaki and Matsushita, 1993), social force model (Helbing et al., 2000) and centrifugal model (Yu et al., 2005). These models are more suitable to simulate simple scenarios. With all these types of models above, several interesting phenomena and results have been generated. For example, it was found that pedestrian transverse movement during evacuation processes could be blocked by vertical grouping (Wang et al., 2013). Suitable leaders and stable group formation were helpful in reducing collision and evacuation time in a multi-room office with obstacles (Zhang et al., 2018). With the leader-follower evacuation method, more ordering and efficiency displayed during evacuation processes (Dong et al., 2014). It was also reported that the V-like walking pattern of groups was beneficial to social communication among group members. However, this pattern could reduce the flow (Moussaïd et al., 2010).

Experiments are helpful in collecting real movement data of pedestrian groups and providing input for model validation. Many controlled experiments (Okada et al., 2012, Chen et al., 2017, Cłapa et al., 2015, Fu et al., 2019) and field studies (Zanlungo et al., 2015, Jazwinski and Walcheski, 2011, Wei et al., 2015) have been conducted to explore group walking behavior on escalators, the effect of pedestrian load or emergency signage on group evacuation, the influences of group sizes, children and density on group locomotion in flat public places, etc. Controlled experiments were primarily performed on stairs (Ma et al., 2017), or in circular arenas (Dyer et al., 2009), corridors (Crociani et al., 2017), rooms (von Krüchten and Schadschneider, 2017) and underground bus concourses (Cuesta et al., 2016). Group behavior primarily during evacuation processes or in bi-directional flow was investigated. It was highlighted that group behavior had a negative influence on the stairwell evacuation performance under normal visibility conditions, but induced an increase in descent speed under impaired visibility conditions (Ma et al., 2017, Zhang et al., 2018). The presence of groups in counter flow could result in fragmented lanes, and reduce specific flow (Crociani et al., 2017).

In field studies, the influences of cultural differences, group size, the number of accompanying children, gender, group composition, etc. on group behavior and crowd dynamics in different public environments were observed. It was demonstrated that both the group size and accompanying children could independently predict pedestrian walking speed in a shopping mall, although neither of them noticeably impacted on movement trajectories (Jazwinski and Walcheski, 2011). In comparison with individual pedestrians on sidewalks, wide-sidewalks and tourist precincts, larger group sizes induced an evident decrease in walking speed (Rastogi et al., 2010). Male groups walked faster than female groups on sidewalks, shopping streets, the seafront, etc. (Costa, 2010). The presence of groups could reduce the capacity of the infrastructure, such as a bridge (Duives et al., 2014). Observation data showed that when dyads crossed non-signalized intersections, there would be a leader who crossed first and then followed by the other one (Gorrini et al., 2017).

From previous research, we can discover that although group behavior has been investigated in different scenarios by experiments, few focused on the movement of social groups on stairs, especially through the field studies. Most research concerning group behavior on stairs was performed under evacuation situations by controlled experiments. Social interactions between group members and microscopic characteristics of group movement on stairs under normal situations were rarely involved. It is well-known that stairs represent the primary egress component for buildings (Ma et al., 2012, Huo et al., 2016). Pedestrian movement on stairs is more complicated than that in corridors or rooms (Ma et al., 2017). There is a need to systematically investigate how pedestrian group members interact with the others and what the feature of their organization is on stairs. These will be useful for a better understanding of group behavior in different scenarios, and for crowd management.

This paper aims to explore intragroup organization and microscopic movement characteristics (i.e., speed, space-time diagrams and offset angles) on stairs, and to compare the difference in social interactions among group members on stairs with that in/on passageways, sidewalks, etc. The influences of pedestrian characteristics (e.g., gender and social relationships) on group movement are further discussed. Theseare performed through a field study on stairs outside a building located in a university. The remainder of this paper is organized in the following: Section 2 presents the detailed experiment setup; Section 3 introduces experimental results and discussions; Finally, conclusions are reported in Section 4.