† Corresponding author. E-mail:
Project supported by the National Natural Science Foundation of China (Grant No. 61233001) and the Fundamental Research Funds for the Central Universities, China (Grant No. 2017JBM014).
This paper focuses on the simulation analysis of stripe formation and dynamic features of intersecting pedestrian flows. The intersecting flows consist of two streams of pedestrians and each pedestrian stream has a desired walking direction. The model adopted in the simulations is the social force model, which can reproduce the self-organization phenomena successfully. Three scenarios of different cross angles are established. The simulations confirm the empirical observations that there is a stripe formation when two streams of pedestrians intersect and the direction of the stripes is perpendicular to the sum of the directional vectors of the two streams. It can be concluded from the numerical simulation results that smaller cross angle results in higher mean speed and lower level of speed fluctuation. Moreover, the detailed pictures of pedestrians' moving behavior at intersections are given as well.
Thousands of people died and more were injured in crowd stampedes over the past 100 years around the world.[1] Cities are densely populated nowadays and the collective behaviors seem to occur more frequently, such as traveling by subway, shopping in malls, and watching sports games in stadia. Hence, how to guarantee the safety and comfort of pedestrian motion and improve the efficiency of public space has become an urgent problem to be solved.
Pedestrian dynamics refers to the study of how and where crowds form and move, which usually studies the dynamic features of pedestrian flows. The study on pedestrian dynamics has attracted considerable attention of researchers in fields of physics, transportation, and more. Some good pedestrian models have been put forward with the advance of research. In general, pedestrian models are divided into three categories: (i) macroscopic models, where pedestrians are studied as a group without considering the interactions among pedestrians and pedestrian flows are analogous to gases and fluids in some sense;[2] (ii) microscopic models, where every pedestrian is treated as an individual and the interactions among pedestrians are considered;[3] (iii) mesoscopic models, which inherit from both macroscopic models and microscopic models.[4] The overview of some representative pedestrian models can be found in Refs. [5]–[7]. Pedestrian models can describe the individual and group behaviors and reproduce the empirical scenarios, so they can help researchers to analyze the walking behaviors and dynamic features of pedestrian flows. In the previous investigations, arch-like blockings in bottlenecks, lane formation in bidirectional flows, and other self-organization phenomena have been observed and studied by various approaches.[1,8] Furthermore, the escape panic[9,10] and the emergency evacuation process under normal conditions and from a smoke-filled or afire room[11–15] have been studied by using multiple pedestrian models.
Recently, intersecting flows become one of the topics of most concern because they are practically unavoidable and likely to cause clogging and accidents. The stripe formation is observed when two pedestrian flows intersect, but no stable pattern exists when three or more pedestrian flows intersect.[1,8] Zhang et al. proposed a two-dimensional model based on the cellular automata model in order to simulate pedestrians’ crossing behaviors and the simulation results showed that this model can be applied successfully.[16] Saegusa et al. studied the intersecting flows under a periodic boundary condition by using the lattice gas model, and found that the overshooting in the fundamental diagram is attributed to the unstable congestion.[17] Guo et al. put forward a new microscopic model which is similar to the discrete choice model, and conducted several real pedestrian experiments with different intersecting angles to calibrate it.[18] Besides, pedestrians’ avoidance behaviors in crossing flow were investigated in Ref. [19].
On the basis of the previous studies, little research has been reported about the dynamic features of intersecting flows with different cross angles, so we aim to find the direction of the stripes, the influence of the cross angles on dynamic features, and the moving details of pedestrians at intersections in this paper. The model we choose is the social force model, which is a continuous microscopic model. It is universally accepted that self-organization phenomena are the consequence of interactions, which are considered in the social force model. By using the social force model, we can reproduce the stripes formation successfully. To reflect the effect of the cross angles, we set up several simulation scenarios in which only the cross angles are different, and then investigate the mean speeds and the fluctuations of speed. By comparing the mean speeds and the standard deviations, we can obtain whether the cross angles affect the motion of pedestrians and what the effects are. As for the moving details, we adopt the approach that picks some individuals and then snapshots them for each second in a time period when they are at intersections, then we find the common regularities.
The organization of this paper is as follows. In Section
Based on socio-psychological studies that both socio-psychology and physical forces affect the motion of pedestrians, Helbing et al. introduced the social force model to investigate the pedestrian movement, which is particularly suited for describing the occurrence of pressure in crowds.[9]
The social force model considers the influence of the neighboring pedestrians and the environment on individuals’ movement in crowds. Pedestrians are driven by: the desired force
Generally, pedestrians have certain destinations, so each pedestrian has a desired velocity and a desired direction. In practice, the actual speed and the desired speed are different due to the interference of obstacles and other neighboring pedestrians. The desired force expresses pedestrians’ willingness to achieve the desired velocity. Pedestrian i likes to move with a certain desired speed
Pedestrian i prefers to keep a distance from other pedestrian j, so there exists an interaction force
When dij is smaller than rij, the pedestrians touch each other and compressions emerge. In this situation, there exist two extra forces. The term kg(rij − dij)
The interaction force between pedestrian i and wall W is treated analogously as
The subway stations, malls, stadia, or other public places are likely to show intersecting pedestrian flows. Intersections of pedestrian flows are one of the major problems that affect the safety and efficiency of pedestrian movement and impossible to avoid in practice, so it is meaningful to know how people behave at intersections and the effect of the cross angles.
Real-life experiment is unexpected and dangerous, so there is an urgent need to set up a reliable model to simulate the pedestrian motion. The parameters in the social force model are specified in Table
We study the intersecting flows which consist of two components of pedestrians moving into an area whose size is 50 m × 50 m. The schematic diagram of the simulation scenarios is shown in Fig.
In each scenario, the initial speed of pedestrians on both sides is 0. On each side, the distance between pedestrians’ initial positions and the entrance is 1 m, which is designed in order to allow the pedestrians to enter the area with their normal walking state. The width of pedestrians’ initial position line is 4 m, that is, the initial positions are (−1 m, 13–17 m), (13–17 m, −1 m) in scenario 1, (−1 m, 23–27 m), (23–27 m, −1 m) in scenario 2, and (−1 m, 13–17 m), (23–27 m, −1 m) in scenario 3. Also, the initial number of pedestrians and the inflow rate of the two streams are exactly the same in the three scenarios. The initial number is 50, and the inflow rate is set to 5 p/s. The positions and speeds of pedestrians are updated after each time step, and time step ΔT is set to 0.01 s.
The parameters of the social force model in these scenarios, namely, the desired speed, reaction time, and body size, are identical. According to Ref. [20], the parameters are set as follows. Reaction time τ satisfies a uniform distribution between 0.2 s and 0.5 s with interval Δτ = 0.05 s. The desired speed is subject to a normal distribution, the mean of which is 1.35 m/s and the standard deviation is 0.2. The radius uniformly distributes between 0.2 m and 0.3 m with interval Δr = 0.025 m. The parameters are presented in Table
The simulation time of each scenario is 150 s, and each scenario runs 10 times.
To ensure the accurate analysis of dynamic characteristics at intersections, only the data collected when the two streams of pedestrians intersect completely are used, so we define a statistical area of size 10 m × 10 m, which is the central dashed square shown in Fig.
When two wide pedestrian flows intersect, stripe formation can be observed. This can be understood as a self-organized behavior, which means that the patterns are not planned or organized artificially, instead, the self-organization phenomena are considered as the consequence of interactions among pedestrians. What is more, the empirical investigations showed that the stripes are perpendicular to the sum of the directional vectors of the two pedestrian flows,[1] for example, the stripes are expected to have an angle of 45° in the case of a perpendicular intersection.
We try to find the collective pattern of intersecting flows in our simulations. As shown in Fig.
Figures
The conclusion drawn from our simulations is identical to the empirical investigations, that is, the stripes are perpendicular to the sum of the directional vectors of the two pedestrian flows.
The performance of pedestrian flows is characterized by the mean speed and the fluctuation of speed. When the simulation is conducted once, we obtain a mean speed which is calculated from 40 s to 150 s. After running simulations of each scenario for 10 times, we obtain 10 mean speeds for each scenario, the results are shown in Table
As for the fluctuation of speed, we choose one from 10 simulations of each scenario as the example. The fluctuation diagrams are shown in Fig.
By analyzing Tables
These results indicate that greater cross angle leads to stronger interactions among pedestrians in two directions, which means more conflicts at intersections. In order to get a precise and direct comprehension of the results, we draw a sample force diagram in which the cross angle is 120° (see Fig.
Moreover, because of the dramatic speed changes in scenario 3, we infer preliminarily from experience that the fluctuation of speed is related to the process of congestion to some extent and a higher level of fluctuation is dangerous when the pedestrian density is large, while this inference should be verified by future work.
In order to know how pedestrians move at intersections, we use different colors to mark several individual pedestrians, then we choose a time period and take snapshots for each second, the pictures are shown in Figs.
Figure
Figure
Figure
The above observations and analysis lead us to the following conclusions. (i) Pedestrians always move forward with neighboring pedestrians who have the same desired direction, and these pedestrians form a group in the form of a stripe. (ii) As time goes by, it is possible that some other pedestrians who are in the same direction with the stripe and not far from the stripe will approach the stripe gradually (see Fig.
As we all know, pedestrians moving in two different directions occupy the same space when these pedestrians meet at intersections, so they are not able to choose the direct way to go forward. By forming stripes, they are allowed to penetrate each other with less conflicts and can continue walking to their destinations without stopping. In this way, the stripe formation is an optimal way to pass the intersection.
By using the social force model, we have investigated the direction of self-organized stripes, presented and analyzed the dynamic features of the intersecting flows with three typical cross angles, and found the way that pedestrians move at intersections. We have confirmed the empirical observations in simulations that pedestrians form stripes in a self-organized way when two pedestrian streams intersect and the direction of the stripes is perpendicular to the sum of the directional vectors of the two pedestrian flows. The simulation results show that greater cross angle results in more conflicts which are characterized by low mean speed and higher level of fluctuation. Greater cross angle is more likely to cause blockings and have adverse effects on the passing of pedestrians. What is more, we found that pedestrians move with stripes and can extend along the stripes. This phenomenon is regarded as an optimal way to pass the intersections quickly.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] |