Guided by gaze: Prioritization strategy when navigating through a virtual crowd can be assessed through gaze activity

Abstract

Modelling crowd behavior is essential for the management of mass events and pedestrian traffic. Current microscopic approaches consider the individual's behavior to predict the effect of individual actions in local interactions on the collective scale of the crowd motion. Recent developments in the use of virtual reality as an experimental tool have offered an opportunity to extend the understanding of these interactions in controlled and repeatable settings. Nevertheless, based on kinematics alone, it remains difficult to tease out how these interactions unfold. Therefore, we tested the hypothesis that gaze activity provides additional information about pedestrian interactions. Using an eye tracker, we recorded the participant's gaze behavior whilst navigating through a virtual crowd. Results revealed that gaze was consistently attracted to virtual walkers with the smallest values of distance at closest approach (DCA) and time to closest approach (TtCA), indicating a higher risk of collision. Moreover, virtual walkers gazed upon before an avoidance maneuver was initiated had a high risk of collision and were typically avoided in the subsequent avoidance maneuver. We argue that humans navigate through crowds by selecting only few interactions and that gaze reveals how a walker prioritizes these interactions. Moreover, we pose that combining kinematic and gaze data provides new opportunities for studying how interactions are selected by pedestrians walking through crowded dynamic environments.

Studying Gaze Behaviour during Collision Avoidance with a Virtual Walker: Influence of the Virtual Reality Setup

“…We can observe in Figure 11 (top and bottom right) an overall decrease of Angle GW with a simultaneous increase of Angle GT , showing that participants tended to look progressively more towards the walker and less towards the target in the first part of the interaction. As the walker is perceived with a possible risk of collision, this pattern is in correlation with previous work [22,28]. However, we observed that the amplitude of these two variations was smaller in the real condition compared to the VR conditions, suggesting that peripheral vision might have played an important role in the real condition.…”

“…In the future we want to look deeper into these effects, such as evaluating the impact on the crossing order, or the relationship between gaze behaviour and body motions. As VR also offers the possibility to explore complex or even inextricable situations while preserving qualitative human behaviours, we also want to explore further such complex scenarios in the future, e.g., by considering situations such as walking in a crowd, such as [28], while manipulating different factors. In line with previous study, participants' gaze activity provides precious insights about individual behaviours in crowds, the nature of information acquired from the environment to achieve crowd behaviours, and the understanding of complex interactions in such contexts.…”

“…they focus on following rather than on avoiding. More recently, Meerhoff et al [28] studied gaze while walking through a crowd in a virtual environment: gaze was more attracted towards the VH with the highest risk of collision, suggesting a close relation between gaze behaviour and trajectory adaptations. In a different context, VR devices equipped with an eye-tracker have also been used to study human locomotion.…”

Collision Avoidance With Multiple Walkers: Sequential or Simultaneous Interactions?

“…Future work could extend this measure by looking at how the magnitude of DG and the final gap crossing behavior relates to body-scaled characteristics such as the shoulder to aperture ratio. In addition to examining I12 and I13, future work could for example use gaze (Meerhoff et al, 2018 ) to also focus on I23 - the interaction between the grouped walkers - to tease apart whether pedestrians in a crowd form a coordinated strategy to let others pass between them. Depending on how the interactions between W2 & W3 (i.e., within a group) are perceived, the decision to go through or around are strongly affected (Bruneau et al, 2015 ).…”

From Crowd Simulation to Robot Navigation in Crowds

“…Each agent is equipped with a synthetic vision component that computes two visual cues, i.e. the distance and time of closest approaches with an other agent, that have been shown in [38] to be relevant in the regulation of human locomotion. These cues are used to estimate a collision risk which is then minimized by the control strategy.…”