Cellular automaton simulation of counter flow with paired pedestrians

Hui Xiong, Xuedong Guo, Wuhong Wang, Huachun Tan, Heng WeiInternational Journal of Computational Intelligence Systems 2011 · DOI

Abstract

Knowledge on pedestrian behavior is the basis to build decision support system for crowd evacuation management in emergency. In this paper, the impact of paired walking behavior on pedestrian counter flow in a channel is studied. The pedestrian walking behaviors are simulated by the cellular automaton model and the pedestrians are classified as single right walker, single left walker, paired right walker, and paired left walker. Single walker can move forward, leftward, rightward or stand still. The paired pedestrians are considered as a combined unit similar to the single walker in terms of route choice and they can move to the same direction simultaneously. It is found that flow and velocity decrease with increase of the paired rate in case of stable density. Simulation results reveal the phase transitions in terms of density from free flow to the unstable flow and from the unstable flow to the congestion flow. However, the critical densities of phase transition are unaffected by the channel size.

Introduction: The Social Dimension of Crowd Dynamics

Most microscopic pedestrian models treat individuals as independent "particles" moving through space. However, in real-world scenarios—like shopping malls or emergency evacuations—people often move in groups or pairs. This paper, published in the International Journal of Computational Intelligence Systems, identifies a gap in understanding how these social units affect "counter-flow," a common situation where two streams of people move against each other in a confined channel. Understanding the transition from a smooth "free flow" to a "jammed" state is critical for architects and safety engineers designing evacuation routes and public spaces.

Methodology: The Cellular Automaton Model

The study utilizes a Cellular Automaton (CA) model, a discrete computational system where a grid represents the walking space. The grid is a 2D system of $0.4 \times 0.4$ meter cells, the typical space occupied by one person in a crowd. Pedestrians are classified into four types: single right-walkers, single left-walkers, paired right-walkers, and paired left-walkers.

At each time step (approximately 0.4 seconds), an agent can move forward, side-step, or stay still based on the occupancy of adjacent cells. The core innovation of this model is the treatment of paired pedestrians as a combined unit. They occupy two adjacent cells and must move to the same direction simultaneously. This adds a layer of "rigidity" to the flow that isn't present in individual-only simulations.

Figure 6: Movement rules and conflict scenarios for paired pedestrians in the CA grid.

The simulation uses periodic boundaries, meaning a pedestrian exiting the right side of the hallway immediately re-enters from the left, keeping the total number of walkers constant to observe long-term steady states.

Results: The Impact of Pairing on Flow and Density

The most significant insight is the "pairing penalty": pairs act as larger, less flexible obstacles. As the "paired rate" (the percentage of people walking in pairs) increases, the mean velocity and total flow of the system drop sharply.

In a crowd of individuals, the system can handle a certain density before jamming. When pairs are introduced, this "critical density" shifts. A density that allows for free flow with individuals might result in a total standstill if those same people are grouped into pairs. For example, at a fixed density of 0.16, the system remains in a fluent state when all walkers are individuals. However, once the paired rate reaches 0.75, the system enters "serious congestion."

Discussion: Channel Size and Phase Transitions

The model clearly identifies three states: Free Flow (stable high velocity), Unstable Flow (velocity begins to drop), and Congestion Flow (velocity nears zero). The researchers tested hallway widths from 40 to 80 cells. While a wider hallway allows more total people to pass (higher absolute flow), it does not change the critical density at which the phase transition to congestion occurs.

Furthermore, in larger channels, conflicts between opposing streams are more distributed. This can lead to a slight decrease in average velocity at certain densities because "lane formation"—a natural phenomenon where people follow those in front of them to avoid head-on conflicts—becomes less efficient in wider spaces.

Conclusion and Future Work

The study demonstrates that social bonds significantly impact crowd dynamics by introducing rigidity into the flow. However, the authors acknowledge several simplifications. Currently, all pedestrians move at the same maximum speed, and pairs never split up. In reality, pairs might temporarily separate to navigate tight gaps. Future models could include dynamic social grouping and a distribution of walking speeds to better reflect real-world complexity.

Summary updated

Abstract

Knowledge on pedestrian behavior is the basis to build decision support system for crowd evacuation management in emergency. In this paper, the impact of paired walking behavior on pedestrian counter flow in a channel is studied. The pedestrian walking behaviors are simulated by the cellular automaton model and the pedestrians are classified as single right walker, single left walker, paired right walker, and paired left walker. Single walker can move forward, leftward, rightward or stand still. The paired pedestrians are considered as a combined unit similar to the single walker in terms of route choice and they can move to the same direction simultaneously. It is found that flow and velocity decrease with increase of the paired rate in case of stable density. Simulation results reveal the phase transitions in terms of density from free flow to the unstable flow and from the unstable flow to the congestion flow. However, the critical densities of phase transition are unaffected by the channel size.
BibTeX
@article{xiong2011cellular,
  author = {Hui Xiong and Xuedong Guo and Wuhong Wang and Huachun Tan and Heng Wei},
  title = {{Cellular automaton simulation of counter flow with paired pedestrians}},
  journal = {International Journal of Computational Intelligence Systems},
  year = {2011},
  doi = {10.2991/ijcis.2011.4.6.36},
}
Figure