Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Multiscale Science and Engineering

https://doi.org/10.1007/s42493-021-00063-8

ORIGINAL RESEARCH

Mouse Models to Investigate the Influence of an Obstacle on Variables


that Create Pedestrian Bottlenecks
Somi Kim1 · Duyen Thi Hai Nguyen1,2 · Hyejin Oh3 · Junyoung Park1,2 

Received: 18 February 2021 / Revised: 17 March 2021 / Accepted: 30 March 2021


© Korean Multi-Scale Mechanics (KMSM) 2021

Abstract
When many people attempt to evacuate a place at the same time, a bottleneck occurs at the exit. Although there are many
experimental variables, such as obstacle size and distance from the exit, there is a lack of experimental studies on the
influence of such variables on escape tendency. Thus, this study was purposed to fill in this gap of knowledge by perform-
ing experiments that varied the exit angle and location of an obstacle. Because experiments with people who assume an
emergency situation may result in actual human casualties, the active avoidance method (i.e., electric shock) was applied to
mouse models in this study. Specifically, electric foot shock was used to induce panic in mice located in a designated wait-
ing section. Then, in a designated measurement section, the escape velocity, escape time, and escape trajectory length were
analyzed to compare the escape tendencies. It was found that the presence of an obstacle in front of the exit increased the
escape velocity, and decreased the escape time. The experimental results also revealed that there was an optimal location
for the obstacle relative to the exit.

Keywords  Rodent experiment · Obstacle · Crowd flow · Evacuation flow · Bottleneck

Introduction that occurred during the Hajj in the vicinity of Mecca, Saudi
Arabia [1]. In Madagascar, at least 15 people were killed
In the event of an indoor emergency situation, such as a in a stadium by people who created a stampede at the end
fire or an explosion, evacuation should be performed to of the Independence Day celebration on June 27, 2019 [2];
minimize human casualty. In the event that many people additionally, failure to control the crowd in a soccer sta-
are attempting to evacuate a place with limited entrances dium on September 9, 2018 left at least one person dead and
and exits, such as a stadium, movie theater, or concert hall, approximately 40 people injured [3]. In Angola, 76 casual-
they may push each other to escape first, creating a bot- ties, including 17 deaths, occurred on February 10, 2017
tleneck at exits. A bottleneck extends the evacuation time, as people attempted to enter a stadium before the start of
and cause stampedes that result in human casualties. Such the opening soccer match [4]. In Nigeria, at least 14 people
accidents are reported to occur every year. In September were killed in February 2019 as the crowd rushed toward the
2015, at least 717 pilgrims died as a result of a stampede gates after the president’s speech [5]. There have also been
stampedes in South Korea that have resulted in human casu-
alties. In Sangju in October 2005, 11 people were killed and
* Junyoung Park more than 100 people were injured during a stampede that
pcello@kumoh.ac.kr broke out as people rushed to enter the area of a concert held
1
Department of Mechanical Design Engineering, Kumoh by Munhwa Broadcasting Corporation (Korea) [6]. These
National Institute of Technology, 61 Daehak‑Ro, examples show that stampede-related accidents can occur
Gumi 39177, Gyeungbuk, Korea anywhere if a number of people suddenly feel the need to
2
Department of Aeronautic, Mechanical and Electrical attempt to escape from a congested space.
Convergence Engineering, Kumoh National Institute To preclude such accidents, interventions purposed to
of Technology, 61 Daehak‑Ro, Gumi 39177, Gyeungbuk, prevent pedestrian bottlenecks from occurring at exits are
Korea
required. For example, obstacles can be installed in front of
3
Department of Mechanical Engineering, Stony Brook exits to prevent a bottleneck from occurring in the event of
University, 100 Nicolls Rd, Stony Brook, NY 11794, USA

13
Vol.:(0123456789)
Multiscale Science and Engineering

an emergency situation. One may think that such obstacles studies on the respective influences of the presence of obsta-
will hinder escape; however, interestingly, they have been cles, the exit angle, and the distance to the exit on the escape
proven to be effective as a means to prevent bottlenecks, efficiency. Thus, in this study, the presence of an obstacle in
because they reduce pedestrian congestion [7]. Various front of the exit, as well as the exit angle, were investigated
related studies have been conducted, including simulated as factors that affect the escape efficiency; specifically, the
and actual experiments with pedestrians, silo experiments, respective influences of the distance between the obstacle
and simulations with animals. and exit, exit angle, and distance between the obstacle and
Frank et al. simulated the “social force model” to show the escape wall on the escape tendency of mice were experi-
that column-shaped obstacles can reduce evacuation time mentally investigated.
[8]. Haghani et al. emphasized that parameter correction is One of the biggest problems of research based on simu-
important for the accurate prediction of pedestrian flow by lation is the validation. Espeically, the parameters used in
simulating it in a bottleneck section using their corrected the simulation study should be chosen by the results based
parameters. They experimented with 114 people and found on the experiments like this study. The feasibilities of the
that an obstacle present in front of the exit was effective in simulation validation using experimental study is already
reducing the escape time when the exit width was narrow done by authors [14].
(i.e., 60–100 cm), and that the effect of the obstacle was
insignificant when the exit was wide (200 cm). They also
found that a longer obstacle was more effective in reducing Experiment
the escape time at the exit [9]. Zuriguel et al. found that the
pressure applied to the walls of a silo for grain and cement Background and Experimental Target
powder storage decreased as a result of installing an obstacle
directly above the exit, indicating an increase in the flow Because experiments with people who assume an emergency
rate [10]. situation may result in actual human casualties, rodent simu-
To date, there have been actual and simulated studies in lations were performed in this study. The brains of rodents
which obstacles were installed in front of the exit to mitigate include a hippocampus and an amygdala, which are known
pedestrian bottlenecking. Zuriguel et al. performed experi- to process fear and induce the subsequent response in a way
ments that entailed having a flock of sheep pass through that is very similar to that in the human brain; thus, it is
a narrow door [11], and Lin et al. conducted research on reasonable to use rodents to predict the escape tendency of
the escape efficiency afforded by an obstacle by placing humans in an experiment that induces panic [15]. There are
rodents in a competition to escape from the scented candle various methods that have been employed to induce panic
[12]. They reported that it is necessary to properly adjust in rodents, e.g., placing mice in a pool of water [16, 17], or
the obstacle size and distance to the exit to optimally reduce introducing the scent of a burning candle [12, 18]. Although
the evacuation time, and that an obstacle with a diameter as there are various other methods, the experimental environ-
much as 0.4 to 3 times the exit width that is positioned at a ment required to employ most of them is not easy to control;
distance of 1 to 2 times the exit width can increase the flow this makes them unsuitable for inducing the level of anxiety
rate. They also reported that the presence of an obstacle or fear that would result in a sudden need to quickly evacu-
in front of the exit did not always increase evacuation effi- ate. Thus, electric foot shock was used in this study [14].
ciency, and that improper obstacle placement may decrease The active avoidance test method entails using electric
evacuation efficiency. shock to induce panic in mice such that they would to want
Experiments with people were also performed. Shi et al. to evacuate an enclosed space [19]. In an active avoidance
found that the flow of pedestrians was significantly increased test, electric shock is applied to mice, allowing them to
by using obstacles to precisely adjust the structural charac- escape to a safe section within the enclosed space where
teristics of the exit pathway; however, they reported that they do not experience electric shock. After several training
obstacles may decrease the escape efficiency in non-emer- sessions, the rodents learn to escape to the safe section to
gency situations [13]. In experiments with up to 200 people, avoid the electric shock. It should be noted that the strength
Garcimartin et al. confirmed that obstacles do not necessar- of the electric shock is maintained at a level at which the
ily shorten the evacuation time; however, they found that rodents experience a tingling feeling, and there is no threat
they are helpful in preventing falls by suppressing collective to their life.
transversal rushes [7]. Regarding the animal model, a total of 48 C57BL/6N
The findings of the above-mentioned studies indicate mice with small inter-individual genetic differences were
that the installation of obstacles in front of the exit does not used (29 females and 19 males). Because environmental
guarantee desirable results. Pedestrian-focused studies have changes, such as breeding, may affect behavior analysis,
also been actively conducted, but there are few experimental males and females were separated and limited to four to five

13
Multiscale Science and Engineering

mice were per cage to minimize such changes. The experi- circuit in the event that they are too close, and to prevent the
ments were performed in accordance with Internal Animal mice from falling in between the rods in the event that the
Care and Use Committee (IACUC) protocol to minimize spacing is too wide. An electric stimulator was connected
stress on the rodents (authorized No. UNISTIACUC-20-27). to the stainless steel rods in parallel to enable the electric
shock. The safe section, i.e., the section where electric shock
Experimental Setup could be avoided, was fabricated by using black acrylic, with
sawdust laid on the floor.
Figure 1 shows a schematic of the experimental setup. The Two types of entrance restrictions were applied, i.e., a full
setup consisted of a waiting section, measurement section, partition, and a partition with an opening. The dimensions
and safe section. The walls of the waiting section and meas- of the exit to the safe section were 5 × 5 cm. These dimen-
urement section were acrylic, but the floor was composed sions were set in consideration of the 2–3-cm width of the
of stainless steel rods to enable electric shock application abdomen of an adult mouse to allow two mice to escape at
to the mice. The 2-mm-diameter stainless steel rods were the same time. However, the exit angle and distance from
spaced 1 cm apart to prevent the feces from creating a short the exit to the obstacle were variable.

Fig. 1  Experimental settings. a Schematic of experimental setup (W = 40 cm, L = 170 cm, L’ = 100 cm, H = 5 cm), and b schematic floor plan. X
and θ are variables

13
Multiscale Science and Engineering

Figure 2 shows the device used to apply electric shock. blocked the waiting section was removed to allow the mice
An isolated square-wave stimulator (PHIPPS & BIRD, NO. to escape to the safe section by passing through the meas-
7092-611/120) was used. A Sony FDR-AX700 camcorder urement section. The experiment was concluded once all of
was used to obtain video for experimental analysis. An in- the mice reached the safe section. The entire process, which
house program, referred to as the MouseTracker program, corresponds to the span of time beginning at the moment
was developed and used to enable video analysis. This pro- when the first mouse escaped from the waiting section to
gram was designed to allow the operator to select x and y the measurement section, and ending at the moment when
positions in each frame, thus enabling measurement of the the last mouse reached the safe section, was recorded using
instantaneous velocity, average velocity, escape time, and the camcorder. The recorded video was transferred to a com-
escape trajectory of the mice. Figure 3 is a screenshot that puter and analyzed using the MouseTracker program.
provides an example of how the program was used to ana- The experiments were performed under the conditions of
lyzing the video. three values each for the exit angle and distance between the
exit and obstacle (Table 1).
Experimental Method The exit angle (15°, 45°, and 75°) and distance of the
obstacle from the exit were varied to investigate their respec-
To begin the experiment, the mice were placed in the wait- tive influences on escape tendency. The effects of having an
ing section. Electricity was applied to the stainless steel rods obstacle were also investigated by performing experiments
via the isolated square-wave stimulator that was connected with and without the obstacle. If there is an obstacle, it may
in parallel; this resulted in an electric shock being applied not be possible to escape if the distance between the obsta-
to the feet of the mice standing on the rods in the waiting cle and the wall at the exit angle of 75°, the most difficult
section. The mice were surprised and appeared to panic to escape, is too narrow. Therefore, the experiment was
when they experienced the electric shock. When it was con- conducted with a minimum distance of 15 cm for smooth
firmed that all of the mice were panicked, the partition that escape by adjusting the distance between the exit wall and the

Fig. 2  Electric shock device and stimulation output. a Isolated square-wave stimulator, and b example of electrical stimulation output

Fig. 3  Screenshot showing an example of video analysis with the MouseTracker program

13
Multiscale Science and Engineering

Table 1  Experimental conditions

Exit angle (°) 15, 45, 75


Distance between the exit 15, 40, 60
and obstacle (cm)
Measurement section
entrance restrictions

Fig. 5  Escape velocity and escape time as functions of the exit angle


obstacles. Figure 4a provides an example of how the experi- (blue: no obstacle; red: with obstacle)
ment proceeded under the condition that the full partition was
used. Because the mice exhibited a tendency to pass through Figure 5 shows the results of the experiment performed
the measurement section along the walls, which allowed them with 48 mice. The blue bars indicate the average escape
to escape without being affected by the obstacle, an entrance times, which were consistently shorter than those corre-
restriction (Fig. 4b) was applied to enable analysis of how sponding to the trials in which an obstacle was in place.
the presence of an obstacle influences escape tendency. The The line graph results show that the average escape velocity
operating conditions for the isolated square-wave stimulator was higher for the trials with an obstacle than those without
were fixed as follows: an amplitude of 45 V was applied at a an obstacle. These results indicate that installing an obstacle
frequency of 400 Hz for a duration of 1 ms. in front of the exit increased the escape velocity and reduced
the escape time.
Figure  6 shows the escape trajectories of the mice
Experimental Results and Discussion according to the escape angle, and Fig. 7 shows the aver-
age escape trajectories. In Fig. 6, the red solid lines repre-
Effects of the Presence of an Obstacle sent the average trajectories. The experimental equipment
is symmetrical with respect to the center exit. Since the
To examine the effects of the presence of an obstacle in front average trajectory of the mice simply showed a straight
of the exit on the escape tendency of the mice, the results of line in the center, the average trajectory was obtained
experiments with an obstacle placed 15 cm in front of the after all the trajectories on the left were transformed sym-
exit were compared to those without an obstacle (Fig. 5). metrically to the right. Then, the average trajectory on the

Fig. 4  Screenshot showing the moment of escape from the waiting section to the measurement section in the case of a no entrance restriction
and b an entrance restriction

13
Multiscale Science and Engineering

Fig. 6  Complete escape trajectory results obtained under the conditions of no obstacle (left) and an obstacle at X = 15 cm (right); a exit angle:
15°, b exit angle: 45°, and c exit angle: 75°

right side is also drawed in symmetry on the left side. The It can be seen in Fig. 8 that the mice tended to escape in
results presented in Figs. 6 and 7 indicate that the aver- a more orderly manner when the exit angle was 45° than
age escape trajectory was not significantly affected by the when it was 15°. In the case of the 75° exit angle, the obsta-
presence of an obstacle. Thus, although the presence of cle appeared to have hindered escape because the escape
an obstacle did not significantly affect the escape trajec- path was already very narrow. These observations indicate
tory, its presence corresponded to more desirable escape that escape efficiency is affected by the exit angle, the dis-
velocity and escape time results. The reason for this is tance between the obstacle and exit walls, and the distance
believed to be that the obstacle in front of the exit prevent between the exit and the obstacle (X).
a bottleneck from occurring.
It can be ascertained from Table 2 that, under the condi- Obstacle Location
tion of an obstacle placed at 15 cm in front of the exit, the
average escape trajectory length was shorter for an exit angle Figure 9a, b show the measured escape velocity and escape
of 75° than for that of 15°. Because the average escape tra- time results for an obstacle positioned at X = 15, 40, and
jectory length was shorter, the escape time was also expected 60 cm.
to be shorter. However, the escape time associated with an Figure 9 shows that the escape velocity increased and
exit angle of 75° was longer than those associated with exit the escape time decreased as the distance between the exit
angles of 15° and 45°, as shown in Fig. 5. Additionally, there and obstacle was increased from 15 to 40 cm, regardless of
was no significant difference in escape velocity between the the exit angle. Furthermore, the results presented in Table 3
75° and 45° exit angles. These results indicate that, under the show that the average escape trajectory length observed
condition of an exit angle of 75°, the presence of an obstacle at X = 40 cm was shorter than that observed at X = 15 cm
in front of the exit may hinder escape. for all exit angles. Alternatively, a comparison of the

13
Multiscale Science and Engineering

Entrance Restriction

All of the results presented above corresponded to no


entrance restriction. As previously mentioned, because the
mice were able to escape along the walls from the beginning,
obstacle existence and placement did not play a significant
role until they approached the exit at the same time. Because
the influence of such congestion can be partially removed
by applying an entrance restriction, experiments with an
entrance restriction were also performed. The entrance
restriction is controlled by the partition existing between
the Waiting section and the Measurement section. (Refer to
Fig. 1a and Table 1) The entrance size of the restricted parti-
tion was 5 cm in consideration of the fact that width of the
abdomen of an adult mouse was 2–3 cm in the same way as
the exit size. Since the experimental equipment is symmetri-
cal, a square hole is located at the center of the partition.
Figure 10 shows the results according to the presence or
absence of entrance restrictions when there was no obstacle.
When there was entrance restriction, the bar graph shows
that the escape time decreased and the line graph shows
that the escape velocity also increased at 45° and 75°. This
indicates that cases with entrance restriction bring more
favorable results for escape than those without entrance
restriction, because significant congestion occurs near the
exit. If there is an entrance restriction, the escape time and
velocity varies noticeably as the exit angle changes. It may
be related with “the congestion at the exit”. Namely, the
Fig. 7  Averaged escape trajectory results; a exit angle: 15°, b exit escape time and velocity are dependent on the exit angle in
angle: 45°, and c exit angle: 75° the case of entrance restriction. And they are not in the case
of no entrance restriction.
Figure  11 shows how the existence of an obstacle at
Table 2  Average escape trajectory lengths for various exit angles X = 15 cm affected the escape time and escape velocity when
Exit angle Case: No obstacle Case: there was an entrance restriction. It can be ascertained from
X = 15 cm this figure that the existence of an obstacle tended to shorten
obstacle the escape time, but that the difference in escape times
15° 125.5 cm 134.4 cm
between the present/absent obstacle conditions decreased
45° 123.3 cm 124.7 cm
as the exit angle was increased from 15° to 75°. Regarding
75° 121.9 cm 127.5 cm
the escape velocity results, under the condition of an exit
angle of 15°, the escape velocity tended to be higher when
there was an obstacle; however, it was comparatively lower
when the exit angle was set at 75°. At an exit angle of 45°,
average escape trajectory lengths at X = 40 cm and X = 60 cm the average escape velocities were roughly equal, indicat-
revealed similar average escape trajectory lengths, even ing that the presence of an obstacle had minimal influence;
though the escape velocity decreased and the escape time however, the escape time was much shorter when there was
increased at X = 60 cm. These results indicate that there is an an obstacle.
optimal obstacle position. The reason for this is believed to It can be ascertained from the results presented in
be related to the congestion that was observed near the exit. Table 4 that the presence of an obstacle did not signifi-
Specifically, the optimal distance from the exit seems to be cantly affect the average escape trajectory length. Thus,
close to 40 cm, because the associated conditions resulted the results obtained under the condition of an entrance
in less congestion, whereas more congestion was observed restriction revealed that, when the exit angle was set to 45°,
in the cases of X = 15 cm and 60 cm, even though there was even though the escape velocities were roughly the same,
improvement as compared to the trials without an obstacle. the escape time in the presence of an obstacle tended to be

13
Multiscale Science and Engineering

Fig. 8  Screenshots showing mice evacuating the measurement section; a exit angle: 15°, b exit angle: 45°, and c exit angle: 75°

Fig. 9  Escape time (a) and escape velocity (b) as functions of the distance between the exit and obstacle for various exit angles

Table 3  Average escape trajectory length results for different obstacle when there was an obstacle. Furthermore, the escape time
distances was similar regardless of the presence of an obstacle. This
Exit angle Case: X = 15 cm Case: X = 40 cm Case: X = 60 cm is because the time delay caused by the bottleneck at the exit
when there was no obstacle nearly matched the time required
15° 134.4 cm 117.2 cm 120.5 cm for the mice to slowly line up in front of the obstacle to pass
45° 124.7 cm 118.4 cm 118.6 cm through the exit.
75° 127.5 cm 116.9 cm 116.3 cm

Conclusion
shorter; this is because the obstacle prevented a bottleneck
from occurring at the exit. When the exit angle was set to In this study, experiments were performed to investigate
75°, the escape trajectory length was not influenced by the how an obstacle positioned in front of the exit may prevent
presence of an obstacle, but the escape velocity was lower stampedes from occurring in the event of an emergency. The

13
Multiscale Science and Engineering

shorter in the cases of X = 40 cm and X = 60 cm as com-


pared to the case of X = 15 cm. A comparison between the
X = 40 cm and X = 60 cm results revealed that the condi-
tions created by X = 60 cm (i.e., obstacle placement closer to
the entrance) resulted in a lower escape velocity and longer
escape time. Additionally, although it would seem that the
obstacle would hinder escape, the escape time was found to
be much shorter when an obstacle was present, indicating
that the presence of an obstacle creates conditions that are
more favorable for a safe escape.
Finally, the application of an entrance restriction was
found to comparatively reduce the escape time and increase
Fig. 10  Effects of entrance restriction application. Escape velocity the escape velocity, demonstrating its ability to improve
and escape time as functions of the exit angle in the case of no obsta-
evacuation outcomes. However, unlike the results obtained
cle (green: no entrance restriction; orange: entrance restriction)
from the experiments with no entrance restriction, the
escape trajectory and escape time were shorter in the case
of X = 15 cm than in the cases of X = 40 cm and X = 60 cm.
Specifically, in the cases of X = 40 cm and X = 60 cm, when
there was no entrance restriction, the mice were able to
escape along the walls, with the obstacle in the center hav-
ing minimal influence on their escape route. However, when
the entrance restriction was applied, none of the mice were
able to avoid the obstacle in the measurement section. Thus,
although the application of an entrance restriction may seem
to be counterintuitive, it was found to prevent the congestion
at the exit that was observed when there was no entrance
restriction. The effects of obstacles and entrance restric-
tions on escape efficiency may be topics for future research;
Fig. 11  Effects of entrance restriction and the existence of an obsta- however, it is predicted that the positive effects of entrance
cle. Escape velocity and escape time as functions of the exit angle
restrictions will reduce as the number of mice increases.
(blue: no obstacle; gold: with obstacle)
Acknowledgements  We would like to acknowledge the financial
support from the National Research Foundation of Korea (NRF)
Table 4  Average escape trajectory length results obtained under the
[NRF-2018R1A2B 2004207] and the MSIT(Ministry of Science
condition of an entrance restriction
and ICT), Korea, under the Grand Information Technology Research
Exit angle Case: no obstacle Case: Center support program(IITP-2021-2020-0-01612) supervised by the
X = 15 cm IITP(Institute for Information & communications Technology Planning
obstacle & Evaluation).

15° 115.6 cm 112.7 cm Funding  This work was supported by the National Research Foun-
45° 113.0 cm 115.0 cm dation of Korea (Grant number NRF-2018R1A2B 2004207) and the
75° 111.1 cm 111.0 cm Ministry of Science and ICT of South Korea under the Grand Infor-
mation Technology Research Center support program (Grant number
IITP-2021-2020-0-01612) supervised by the Institute for Information
& Communications Technology Planning & Evaluation.
experimental results revealed that positioning an obstacle
in front of the exit tended to yield a more desirable escape Declarations 
velocity and escape time, regardless of the exit angle. The
experiments were performed by varying the exit angle, and Conflict of interest  Not applicable.
it was found that the application of an exit angle of 75° in the
presence of an obstacle (X = 15 cm) hindered escape because
the distance between the obstacle and the walls near the exit
was too small. Additionally, an exit angle of 45° was found References
to maximize escape efficiency in all cases.
1. BBC NEWS, Hajj stampede: At least 717 killed in Saudi Arabia.
In the case of the obstacle position-varying experiments, (BBC NEWS, 2015), https://​www.​bbc.​com/​news/​world-​middle-​
the escape trajectory and escape time were found to be east-​34346​449. Accessed 24 September 2015

13
Multiscale Science and Engineering

2. BBC NEWS, Madagascar stadium crush kills 16 during national 11. I. Zuriguel, D.R. Parisi, R.C. Hidalgo, C. Lozano, A. Janda, P.A.
celebrations. (BBS NEWS, 2019), https://​www.​thest​adium​busin​ Gago, J.P. Peralta, L.M. Ferrer, L.A. Pugnaloni, E. Clément, D.
ess.c​ om/2​ 019/0​ 6/2​ 7/s​ tampe​ de-m
​ adaga​ scar-s​ tadiu​ m-l​ eaves-m​ any-​ Maza, I. Pagonabarraga, A. Garcimartín, Sci. Rep. (2014). https://​
dead/. Accessed 27 June 2019 doi.​org/​10.​1038/​srep0​7324
3. BBC NEWS, Deadly stadium stampede at Madagascar v Senegal 12. P. Lin, J. Ma, T.Y. Liu, T. Ran, Y.L. Si, F.Y. Wu, G.Y. Wang, Phys.
match. (BBC News, 2018), https://​www.​bbc.​com/​news/​world-​ A (2017). https://​doi.​org/​10.​1016/j.​physa.​2017.​04.​021
africa-​45465​970. Accessed 9 September 2018 13. X. Shi, Z. Ye, N. Shiwakoti, D. Tang, J. Lin, Phys. A (2019).
4. The Guardian, More than a dozen fans killed in stampede at https://​doi.​org/​10.​1016/j.​physa.​2019.​01.​086
Angolan football match. (The Guardian, 2017), https://w ​ ww.t​ hegu​ 14. H. Oh, J. Park, Sci. Rep. (2017). https:// ​ d oi. ​ o rg/ ​ 1 0. ​ 1 038/​
ardian.​com/​world/​2017/​feb/​10/​17-​fans-​killed-​stamp​ede-​footb​all-​ s41598-​017-​14007-6
match-​angola. Accessed 10 February 2017 15. J.F. Cryan, A. Holmes, Nat. Rev. Drug Discov. (2005). https://d​ oi.​
5. BBC NEWS, Nigeria stampede: Deaths at Buhari rally in Port org/​10.​1038/​nrd18​25
Harcourt. (BBC NEWS, 2019), https://w ​ ww.b​ bc.c​ om/n​ ews/w ​ orld-​ 16. C. Saloma, G.J. Perez, G. Tapang, M. Lim, C.P. Saloma, in Proc.
africa-​47220​825. Accessed 13 February 2019 Natl. Acad. Sci. USA (2003). https://​doi.​org/​10.​1073/​pnas.​20319​
6. BBC NEWS, S Korea concert stampede kills 11. (BBC NEWS, 12100
2005), http://​n ews.​b bc.​c o.​u k/2/​h i/​a sia-​p acif ​i c/​4 3046​9 8.​stm. 17. R. Morris, J. Neurosci. Methods (1984). https://​doi.​org/​10.​1016/​
Accessed 3 October 2005 0165-​0270(84)​90007-4
7. A. Garcimartín, D. Maza, J.M. Pastor, D.R. Parisi, C. Martín- 18. P. Lin, J. Ma, T. Liu, T. Ran, Y. Si, T. Li, Phys. A (2016). https://​
Gómez, I. Zuriguel, New J. Phys. (2018). https://​doi.​org/​10.​1088/​ doi.​org/​10.​1016/j.​physa.​2016.​02.​017
1367-​2630/​aaf4ca 19. H. Kim, Brain Functional Animal Behavior Test for Brain Neu-
8. G.A. Frank, C.O. Dorso, Phys. A (2011). https://d​ oi.o​ rg/1​ 0.1​ 016/j.​ rologists (Sigma Press, 2006), pp. 36–41
physa.​2011.​01.​015
9. M. Haghani, M. Sarvi, Simulating pedestrian flow through narrow Publisher’s Note Springer Nature remains neutral with regard to
exits. Appl. Phys. Lett. (2019). https://​doi.​org/​10.​1016/j.​physl​eta.​ jurisdictional claims in published maps and institutional affiliations.
2018.​10.​029
10. I. Zuriguel, A. Janda, A. Garcimartín, C. Lozano, R. Arévalo,
D. Maza, Phys. Rev. Lett. (2011). https://​doi.​org/​10.​1103/​PhysR​
evLett.​107.​278001

13

You might also like