Scientists at Brown University have developed a new model to predict the behavior of people in a crowd based on optical and other sensory data. The study is published in Proceedings of the Royal Society.

Experiments conducted by researchers at Brown University showed that each pedestrian in a crowd controls the direction and speed of their movement based on two visual variables.

First, people try to move in such a way that their neighbors remain motionless in their perception. For this, the speed and direction of movement of those who walk nearby must be the same.

Secondly, people prefer that the field of view does not change during movement. Such changes can occur when neighbors move closer or further away. Therefore, pedestrians in the crowd unconsciously try to fix the distance between themselves and their neighbors.

The scientists also found that the participants in the experiment most of all reacted to the movements of their nearest neighbors. Changes in the behavior of those who walked at a distance did not have a strong effect. This is due to two effects, scientists say: the laws of optics and the principles of occlusion. The movements of a distant object seem less pronounced to us, and pedestrians who walk at a distance are partially covered by the backs of their neighbors. This means that it becomes more difficult to track and predict their actions.

To explore individual movement trajectories, scientists used virtual reality. Study participants in a large open room wore VR headsets that showed animated people. The experimenters controlled the movements of virtual characters in the crowd. For example, some people could turn in a different direction while everyone else kept going straight.

The participants in the experiment were asked to move with the crowd, and the scientists tracked how changes in the behavior of virtual characters affected the trajectory of an individual’s movement.

Based on the data obtained, the researchers built a model that successfully predicts how each individual person in the crowd will move. According to the researchers, the effectiveness of the model has been proven both in virtual reality and for analyzing the movement of people in a real crowd.

“This is the first time we have used sensory data to analyze coordinated movements,” says William Warren, professor of cognitive, linguistic and psychological sciences at Brown University and one of the co-authors of the study. “The model takes into account what people in the crowd see, so we can make more accurate predictions about how the whole group will behave.”

Scientists note that crowd movement prediction models have a wide range of applications. They can be used to plan public spaces, transport infrastructure, escape routes and emergency plans.