Anyone who looks up at platforms 3 and 4 at Eindhoven Central Station will spot them high up in the roof structure: small black devices that reduce passersby to anonymous, moving circular shapes. From 2021 to 2024, they continuously recorded how people hurried to catch their trains and how they exited toward the station’s main hall. They did so during the busy evening rush hour as well as on early Sunday mornings.

Data from the devices was analyzed within the research group Computational Multiscale Transport Phenomena, led by professor Federico Toschi. The group in the Department of Applied Physics and Science Education has long focused on modeling pedestrian flows in crowded environments, he explains. 

“When we zoom out far enough on a crowd, we can use mathematical and physics-based models to systematically analyze movement within a network. With that perspective, and together with rail infrastructure manager ProRail, we studied how people behave on the platform.”

To track people over longer periods on routes with changes in elevation—such as stairs and escalators—new sensor technology was essential, Toschi says. 

“This allows us, as one of the first research groups worldwide, to characterize human behavior in public spaces with high accuracy and over extended periods of time.”

One leader, many followers

That is how postdoc Ziqi Wang identified something striking in the movement patterns of about one hundred thousand passengers exiting trains, which she analyzed using a mathematical method. In an email—Wang is currently in China for a few weeks—she explains her findings. 

“When you step off a train, you often end up following the person in front of you. In the platform footage, we observed a remarkable number of clusters consisting of two people moving in the same direction and at the same speed: a leader and a follower.”

Wang examined three train door zones near a kiosk on the platform. “Even when someone chose the longer walking route past the kiosk, strangers followed that person.” 

This often triggers what Wang describes as a choice avalanche, a pattern she also identified in the data. “Followers are then followed by others, and those followers in turn attract more followers. That’s how we see a sudden shift in walking routes emerge—even when it’s not the most efficient path.”