Research and Application of Public Safety Model Based on Artificial
Intelligence: Simulation of Emergency Evacuation in Zhuhai
Innovation Center Building
Zili Zhao
a
, Zhi’ao Zhang
b
, Xiupeng Zhang
c
and Jiaqi Wan
d
Zhuhai Institute of Urban Planning & Design, Zhuhai, Guangdong, China
Keywords: Artificial Intelligence, Agent, Steering, Public Safety, Simulation.
Abstract: Artificial intelligence has many applicable scenarios in the field of public security. Building safety
assessment is an important part of public safety, and emergency evacuation simulation is an important
means of building performance evaluation. Based on the evacuation simulation principle of artificial
intelligence agent and steering behavior theory, this study uses computer graphics simulation and game role
technology to simulate buildings Starting from the evacuation mode selection, parameter setting and
evacuation time sequence analysis of the simulation model, available safe egress time (ASET) and required
safe egress time (RSET) of the room, floor and the whole building are calculated to determine whether the
evacuation process in the building meets the safety standard. In order to make the simulation results more
scientific and reliable, three different simulation scenarios were set up in this study, namely: full staircase
mode, full elevator mode, and stair based elevator supplemented mode.
1 INTRODUCTION
Public safety is one of the important contents of
urban governance at this stage. As a new type of
application technology that has gradually matured in
recent years, artificial intelligence has been widely
used in the field of social public safety and has
produced very good results. Artificial Intelligence,
referred to as AI, is a technology for researching and
developing theories, methods, technologies and
application systems for simulating, extending and
expanding human intelligence. Artificial intelligence
has many applicable scenarios in the field of public
safety, and building safety assessment is an important
part of public safety. Since 1980, scholars have
constructed evacuation models to simulate the
evacuation behavior of people in buildings. Among
them, the widely used evacuation models are the
cellular automata model and the lattice gas model.
This research is based on the evacuation simulation
principle of artificial intelligence agent and the
modeling of Steering behavior theory, so that we can
simulate the evacuation of people in emergencies,
and simulate the evacuation process of room
evacuation, floor evacuation and arrival at the exit on
each floor. This model can simulate the escape path
and escape time of each individual in the event of a
disaster, and compare the simulation results with
industry standards to determine the reliability of the
evacuation model. Finally, based on the model
simulation results, the safety risk assessment of the
performance-based design of high-rise buildings is
carried out, and the rectification plan and suggestions
for the areas with potential safety hazards are
proposed. Therefore, the research has great practical
significance.
2 OVERVIEW, DATA SOURCES
AND RESEARCH METHODS
OF THE STUDIED AREAS
2.1 Model Simulation Principle
Artificial intelligence is mainly through deep
learning algorithms based on deep neural network
and convolutional neural network algorithms, which
achieves a certain degree of intelligent calculation
through induction, synthesis and other methods. This
research is based on the evacuation simulation
principle of the artificial intelligence agent and the
52
Zhao, Z., Zhang, Z., Zhang, X. and Wan, J.
Research and Application of Public Safety Model Based on Artificial Intelligence–Simulation of emergency evacuation in Zhuhai Innovation Center Building.
DOI: 10.5220/0011722300003607
In Proceedings of the 1st International Conference on Public Management, Digital Economy and Internet Technology (ICPDI 2022), pages 52-60
ISBN: 978-989-758-620-0
Copyright
c
2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)
steering behavior theory to model the movement of
pedestrians. The modeling is mainly composed of
three modules: a graphical user interface, a simulator
and a 3D result display. Steering mode. uses a
combination of path planning, guidance mechanism
and collision handling to control pedestrian
movement. If the distance between people and the
closest point exceed a certain threshold, the
algorithm will generate a new path to change the
walking trajectory of pedestrians.
In the Steering mode, the exit will not restrict the
flow of people, and a reasonable distance will be
maintained between evacuated individuals and
evacuated individuals. By combining computer
graphics simulation and simulation technology in the
field of game roles, graphical virtual exercises are
performed on each individual movement in multiple
groups. In order to observe the escape route of the
personnel, we delete all the external walls and other
structures in the model, leaving only the plane layout
structure diagram. By picking up the corresponding
room, and arranging the corresponding doors,
elevators, and stairs according to the specific location
of the model, an intelligent evacuation model for
people is formed. Finally, we set the relevant
parameters of the model, such as pedestrian gender,
height, walking speed, floor area, room area, etc. The
walking rate in the model is determined by the
density of the crowd in each room, and the flow of
people passing through the exit is determined by the
width of the exit.
2.2 Model Calculation Method
By analyzing the requirements, we set up the
physical scene, and the main functions can be
realized after the overall modeling structure and
personnel's behavior are determined. Therefore, the
agent-based system simulation modeling method can
be used. The process of personnel evacuation is
closely related to detection, alarm measures, and
characteristics of personnel escape behavior. The
necessary evacuation time is calculated based on the
sum of the alarm time, the evacuation pre-action time
of the personnel and the action time of the personnel
from the beginning of the evacuation to the safe
place.
321
TTTRSET ++=
(1)
In the formula: RSET is the required evacuation
time for personnel; T
1
is the detection and alarm
time, which refers to the time from the alarm in the
building to the detection of personnel; T
2
is the
personnel response time, which refers to the time
when personnel begin to evacuate after hearing the
alarm or discovering a danger signal and realizing the
threat of an accident; T
3
is the evacuation action
time, which refers to the time taken from the
beginning of the evacuation to the evacuation of all
internal personnel to a safe area.
=
TT
TT
DRI
T
max
0max
max
1
ln
μ
(2)
In the formula: DRI is the detector response
index; μ
max
= 0.197Q
1/3
H
1 /2
h
5/6
(r>0.15H); μ
max
=
0.946(Q /H)
1 /3
(h0.15H); Q is the heat release rate
of the flame; H is the height of the floor; h is the
height of the detector from the roof; T is the induced
temperature of the detector.
HST 4.0120
02
++=
(3)
In the formula: S
0
is the floor area and H is the
floor height.
)k
nm3
TTT +×=
(4)
v/
m
ST =
(5)
)/(
n
WvDPT =
(6)
In the formula: k is the safety factor, and the
model is set to k=1.5; T
m
is the time to walk to the
safety exit; T
n
is the time for people to pass through
the exit or passage; S is the distance from the initial
position to the evacuation safety exit; v is the
walking speed of people; P is the total number of
people who line up outside the exit or passage; D is
the density of people per unit area of queuing people
outside the exit or passage; W is the effective width
of the narrowest part of the exit or passage.
According to the asylum theory proposed by
Marchant, when a disaster occurs, the time sequence
of emergency evacuation time calculation is shown
in Figure 1. The reaction time of the model design
detection alarm device is calculated by the software
tool DETACT-QS module developed by NIST, and
T1 is about 60s. Personnel response time includes
identification time and reaction time, that is, the
response time after disaster identification to the start
of personnel evacuation. According to the
corresponding research calculation, T2 is 120s
.
Personnel evacuation action time refers to the time
required for the people in the building from the
beginning of the evacuation action to the end of the
evacuation, including the walking time and the time
for passing through the exit. T
m
is the time required
for the evacuees to take emergency actions after
responding; T
n
is the time required for the evacuees
to evacuate from the building to reach the safe area.
ASET (Available Safety Egress Time) is the time
that can be used to escape safely, also known as
"evacuation time allowed."
Research and Application of Public Safety Model Based on Artificial Intelligence–Simulation of emergency evacuation in Zhuhai
Innovation Center Building
53
Figure 1: Calculation of emergency evacuation time.
Figure 2: Basic structure diagram of the evacuation model.
2.3 Basic Structure of the Model
The model structure mainly includes basic building
conditions setting, evacuation density setting, and
overall evacuation characteristics setting of evacuees,
stairwell situation setting, and elevator room situation
setting. Calculate the evacuation time and the number
of evacuees from time t = 0 to the next time interval,
then judge whether the number of people arriving at
the designated safe place is equal to the total number
of people at the initial evacuation, judge whether the
available evacuation time (ASET) of evacuees is
greater than the necessary evacuation time (RSET),
that is, ASET > RSET, and judge whether the design
scheme needs to be adjusted after evaluation, If the
design scheme meets the requirements of safe
evacuation of all personnel, the simulation results
will be output.
2.4 Model Parameter Setting
Taking into account the behavioral habits of people
in emergencies, and "congestion" is an important
factor affecting the evacuation time, this article is
based on the Steering mode and the basic theory of
SFPE behavior, based on traffic, pedestrians will
follow the planned path The exit direction is moving
forward. During the emergency evacuation, if the
movement of pedestrians is affected by the
surrounding environment and other pedestrians, the
personnel will automatically move to the nearest exit
to re-plan the route, but the queue must comply with
the SFPE assumption. Firstly, establish a 3D space
simulation model of the building, and then set
various constraint parameters for each evacuated
person, mainly including the number of people, the
height of the person, the shoulder width of the
person, the density of the person, the distance from
the person to the nearest exit, floor area, house area,
walking Speed, effective width at the narrowest point
of exit or passage, number of elevators, etc. Among
them, the calculation of the number of evacuated
persons is based on the GB50016-2014 "Code for
Fire Protection Design of Buildings" (2019 revised
edition), and finally the escape route and time of
emergency evacuees are generated.
3 EMPIRICAL ANALYSIS
3.1 Introduction to the Evacuation
Scene
Zhuhai Planning Science and Innovation Center
Building serves new industrial projects such as
planning and engineering design consulting. Zhuhai
Planning Science and Innovation Center Building
serves new industrial projects such as planning and
engineering design consulting. The population
density in the building during working hours is
relatively high, and the safety requirements for
evacuation are extremely high. The author conducts
modeling and analysis of buildings in order to
ICPDI 2022 - International Conference on Public Management, Digital Economy and Internet Technology
54
simulate the escape route and time-consuming of
personnel in emergencies. This research selects
Zhuhai Planning Science and Technology
Innovation Center office building as the object of
empirical analysis, which has important research
significance. The design prototype of Zhuhai
Planning Science and Innovation Center Building
has 25 floors, 2 underground floors and 23 floors
above ground. The floor heights range from 3.5m to
6.6m, and the total height is 99.2m. Among them,
B1~B2, F1.5~F5 are parking lots. F6~F22 floors are
dominated by office areas, supplemented by a small
amount of exhibition, catering and leisure areas.The
simulated floors of the model are F14~F22, of which
the total number of simulated people is 600, of
which 375 are male employees, accounting for
62.5% of the total number. The specific functions
and personnel distribution of each floor are shown in
Table 1:
3.2 Building a Simulation Model
Combined with the CAD plan of the Zhuhai Planning
Science and Innovation Center Building, the Steering
mode is used to establish an emergency evacuation
simulation 3D model. Combined with the CAD plan
of the Zhuhai Planning Science and Innovation
Center Building, the Steering mode is used to
establish an emergency evacuation simulation 3D
model. For the transparency and visibility of the
model, so as to observe the escape route of personnel,
delete all external walls and other structures in the
model, leaving only the plan layout structure
diagram. By picking up the corresponding room, and
arranging the corresponding doors, elevators, and
stairs according to the specific location of the model,
a simulation model of intelligent evacuation of
people is formed.
Table 1: The specific functions and personnel distribution of the building.
Floor Department
Floor area (
)
Number/
Person
Male/
Person
Female/
Person
Planning number/
Person
14 Water Affairs Branch 2550.94 61 45 16 63
15
Municipal Affairs Branch 1
2414.68
45 37 8 45
Engineering Technology Center 12 9 3 72
16
Municipal Affairs Branch 2
2397.62
50 42 8 50
Garden Design Branch 35 21 14 63
17
Building branch 1
2414.68
45 31 14 45
Project Management Center
47
18
Building branch 2
2397.62
47 32 15 50
Transportation Branch 46 36 10 63
19
Planning branch 1
2414.68
51 21 30 45
Political Research Center 24 13 11 47
20
Planning branch 2
2384.62
50 21 29 50
information Center 15 9 6 51
21
Production Management
Department
2456.94
17 4 13 23
Human Resources Department 9 0 9 15
Chief Engineer Office 19 10 9 20
22
General Department
2370.19
12 7 5 31
Finance Department 8 3 5 15
Academy Party Committee 3 1 2 4
Figure 3: Floor plan design.
Research and Application of Public Safety Model Based on Artificial Intelligence–Simulation of emergency evacuation in Zhuhai
Innovation Center Building
55
Figure 4:
Top view of floor model.
Figure 5: Perspective view of model structure.
Figure 6: Simulation model diagram.
ICPDI 2022 - International Conference on Public Management, Digital Economy and Internet Technology
56
3.3 Model Parameter Setting
Typically, the evacuation rate increases staff
personnel decreases the density, the greater the flow
density, the slower the movement of personnel, and
vice versa. Foreign research shows: when the general
population density is less than 0.54 people/m
2
, the
speed of the crowd on the level ground can reach
70m/min without being crowded, and the speed of
going down the stairs can reach 48-63m/min. On the
contrary, when the population density exceeds 3.8
people/m
2
, the crowd will be very crowded and
basically unable to move. It is generally believed that
the relationship between personnel density and
moving speed can be described as a linear
relationship within the range of 0.5 to 3.5 persons/m
2
.
In normal use, the elevator can be used as the
main tool for emergency evacuation. The authors set
the experimental elevator parameters based on the
relevant data of the ThyssenKrupp elevator meta200
model. The maximum operating speed is 3m/s, the
maximum lifting height is 150m, and the maximum
load is 1600kg. The maximum number of people
transported by the elevator is about 21 people. The
acceleration is 1.2m/s2. The maximum speed is 7s
when the elevator doors open and close. The person's
body width will affect their comfortable distance, and
the evacuation speed will affect the person's
evacuation time. The body width is set according to
"Chinese Adult Human Body Size (GB10000-88)",
and the specific parameter settings are shown in
Table 2.
3.4 Simulation Process
According to the given conditions, the simulated
number of people on each floor is set according to
the existing statistics of each department. Among
them, the number of people evacuated from the
office room on each floor, the evacuation speed and
time can all be obtained according to the model. In
order to make the simulation results more scientific
and reliable, this research has established three
different scenarios: full stairs mode, full elevator
mode, and mode with stairs as the main and elevators
as a supplementary. In the evacuation process of the
three modes, the fast-or-slow effect is followed, that
is, the number of emergency evacuation continues to
increase with the passage of time, the evacuation rate
is faster at the beginning of the evacuation, but the
evacuation rate is relatively slow in the later period
of the evacuation. After the model is built, the author
combines computer graphics simulation and
simulation technology in the field of game characters
to perform graphical virtual exercises on individual
movements in multiple groups. The simulation
results are shown in Figure 7 and Figure 8. In
addition to analyzing the personnel's escape time, it
can also analyze the bottleneck of the escape channel
based on the real-time distribution of the heat map of
the personnel, and help designers improve the safety
performance of the building.
Table 2: Personnel parameter setting.
Gender Average speed/(m·s-1)
Velocity distribution
threshold (m·s-1)
Body width/cm Height/cm
Male 1.35 Normal distribution, [1.15,1.55] 45.58 167.1
Female 1.15 Normal distribution, [0.95,1.35] 44.58 155.8
Figure 7: The overall effect of simulation.
Research and Application of Public Safety Model Based on Artificial Intelligence–Simulation of emergency evacuation in Zhuhai
Innovation Center Building
57
Figure 8: Local effect diagram of simulation.
Figure 9: The effect of the flow of people on the floor.
Figure 10: Heat map of real-time dynamic distribution of personnel.
Table 3: Evacuation time for three different scenarios.
Pattern t
yp
e T1/s T2/s T3/s RSET/min ASET/min
Full stairs simulation mode 60 120 666.5 14.11 (24, 48)
Full elevator simulation mode 60 120 569.5 12.49 (24, 48)
A model with stairs as the main
and elevators as a supplement
60 120 462.5 10.71 (24, 48)
3.5 Comparative Analysis of Different
Scenarios
In the process of emergency evacuation, the
traditional evacuation method is that when the safety
of the elevator cannot be guaranteed, people can only
choose to use the stairs to evacuate, and cannot use
the elevator. Therefore, adopting the full-stairs
evacuation mode is the lowest guarantee for a smooth
evacuation of people. However, as the main tool for
vertical transportation in buildings, elevators are also
one of the main means of escape for people in the
middle and high-rise buildings. Therefore, in order to
make the simulation results more scientific and
reliable, this study established three different
simulation scenarios, namely: full stairs mode, full
elevator mode, and stairs as the main elevator
auxiliary mode. After simulation, the required
evacuation time (RSET) of emergency evacuation
personnel under three different scenarios was
calculated. Finally, the simulation results are
compared with industry standards to determine the
reliability of the evacuation model. If the available
evacuation time (ASET) of personnel is greater than
the required evacuation time (RSET), the building
design plan is feasible, otherwise the design plan
needs to be adjusted until it meets the safe evacuation
of all personnel. See Table 3 for details.
ICPDI 2022 - International Conference on Public Management, Digital Economy and Internet Technology
58
Figure 11: Simulation of three different scenarios.
Figure 12: Simulation comparison analysis of three different modes.
After analysis, it can be found that the full staircase
mode takes 14.11mins, the full elevator mode takes
12.49mins, and the staircase main elevator takes
10.71mins for the auxiliary evacuation mode. By
comparative analysis, it takes 14.11mins for the full
staircase mode, 12.49mins for the full elevator mode,
and 10.71mins for the staircase main elevator and
auxiliary evacuation mode. In an emergency, the first
reaction of a person is to choose the stairway closest
to him as the first escape exit. Judging from the
behavioral characteristics of the evacuees, people
showed a clear herd mentality when they chose to
take the elevator, and few people would take the
elevator without queuing up. Therefore, people
waiting to take the elevator during the emergency
evacuation process are the main factors affecting the
evacuation time. Secondly, the congestion, detention
and the turning design in the stairs are also the key
factors affecting the evacuation time.
4 CONCLUSIONS AND
RECOMMENDATIONS
(1) The evacuation simulation model based on agent
technology and Steering mode reflects the evacuation
Research and Application of Public Safety Model Based on Artificial Intelligence–Simulation of emergency evacuation in Zhuhai
Innovation Center Building
59
behavior of the evacuated people more objectively
and truly. But there are still many uncertain factors in
the evacuation process, such as the physical
condition of the personnel, stressful psychological
factors, evacuation methods as well as the emergency
rescue situation. In the actual process, people show
obvious herd mentality when choosing stairways or
taking elevators to escape. Therefore, regular
emergency evacuation drills can reduce the fear of
people in the event of a disaster, so that the crowd
can escape in an orderly manner in the event of a
disaster, shorten the evacuation time greatly.
(2) In the process of emergency evacuation, the
evacuees will choose the closest stairway as the first
escape exit, so the stairwell and stair turns are most
likely to cause congestion. Reducing the gathering of
people between different floors at the stairway can
greatly increase the evacuation rate. For example,
each stairwell can be equipped with a monitoring
system, so that the control room commander can
dynamically monitor in real time, and open the
completed or about to complete evacuation of the
stairwell to other floors. So that the waiting personnel
can obtain a new escape route, which can improve
the evacuation efficiency to a certain extent.
(3) Under a certain floor height, the evacuation
mode supplemented by stairs as the main elevator is
more time-saving than the evacuation mode of full
stairs or full elevators. Therefore, in the actual
disaster escape, if the safe use of elevators can be
ensured, choosing the evacuation mode with stairs as
the main elevator as the auxiliary will greatly
improve the escape efficiency of middle and high-
rise buildings.
ACKNOWLEDGMENT
This work is funded by Supported by Major State
Basic Research Development Program (Item
Number: 2016YFB0502300).
REFERENCES
Alizadeh R.A dynamic cellular automaton model for
evacuation process with obstacles [J]. Safety Science,
2011, 49(2): 315 -323.
Bryan J.L. Behavior in fire: the development and maturity
of a scholarly study area. Fire and Materials,1999,
23(6): 249-253.
B N. Guide for smoke management system in malls, atria
and large area [M]. Quincy: National Fire Protection
Association, 2000.
CAPPUCCIO J. Pathfinder: A computer-based timed
egress simulation [J]. Journal of Fire Protection
Engineering, 2000, 8(1): 11-12.
DANG Huisen, ZHAO Yuning. Simulation of Crowd
Evacuation Based on Pathfinder [J]. China Public
Security. Academy Edition, 2011, 25(4): 46-49.
Fang Xiaoyu, Yang Zhenshan. Pathfinder based
simulation of crowd evacuation in high-rise building
[J]. Journal of Bohai University (Natural Science
Edition), 2016, 37(2): 177-183.
FANG Z M, SONG W G, ZHANG J, et al. Experiment
and modeling of exit-selecting behaviors during a
building evacuation [J]. Physica A, 2010, 389(4): 815 -
824.
Fukamachi M, Nagatani T. Sidle effect on pedestrian
counter flow [J]. Physica A, 2007, 377(1): 269 -278.
Guo Jun. A typical application of BIM in the whole life
cycle of building in China [J]. Architectural Skills,
2011(1): 95.
Helbing. Social force model for pedestrian dynamics.
Physical Review E, 1995, 51(4): 4282-4286.
HELBING D, FARKAS I, VICSEK T. Simulating
dynamical features of escape panic [J]. Nature, 2000,
407(6803): 487-90.
JIN Run- guo, MAO Long, YUE zeng. Application of
FDS and Pathfinder in Building Fires and Evacuation
[J]. Industrial Safety and Environmental Protection,
2009, 35(8): 44-50.
MARCHANT E W. Modeling fire safety and risk [A].
David Canter Fires and human behavior [C]. New
York: John Wiley and Sons, 1980, 302.
OLENICK S M, CARPENTER D J. An Updated
International Survey of Computer Models for Fire and
Smoke [J]. Journal of Fire Protection Engineering,
2003, 13(2): 87-110.
PAULSEN R L. Human Behavior and Fires: An
Introduction [J]. Fire Technology, 1984, 20(2):15.
SAILENDRA D, SHAH A. Assessment of emergency
escape routes for a building using Pathfinder: A Case
Study [J]. International Journal of Engineering
Sciences & Research Technology, 2015, 4(6): 200-
207.
Shields T J, Boyce K E. A study of evacuation from large
retail stores [J]. FIRE SAFETY JOURNAL. 2000,
35(1): 25-49.
Stephen • Lucci, Danny • Ke Peike. Artificial Intelligence
(Second Edition) [M]. People Post Press, 2018(10).
WU Shuigen, YOU Yulin. Simulation of Emergency
Evacuation in High-rise Residential Buildings Based
on BIM and Pathfinder [J]. Structural Engineers, 2017,
33(4): 83-89.
YU J F, FANG Y T, WANG J. Study on the effect of
exit's position changes on evacuation in unit rooms
with two exits [J]. Journal of Applied Fire Science,
2012, 22(1):1-19.
ICPDI 2022 - International Conference on Public Management, Digital Economy and Internet Technology
60