Research and Application of Public Safety Model Based on Artificial

Intelligence: Simulation of Emergency Evacuation in Zhuhai

Innovation Center Building

Zili Zhao

, Zhi’ao Zhang

, Xiupeng Zhang

and Jiaqi Wan

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

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

 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

is the detection and alarm

time, which refers to the time from the alarm in the

building to the detection of personnel; T

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

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.













−

DRI

max

0max

max

(2)

In the formula: DRI is the detector response

index; μ

max

= 0.197Q

1/3

1 /2

5/6

(r>0.15H); μ

max

0.946(Q /H)

1 /3

(h≦0.15H); 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

++=

(3)

In the formula: S

is the floor area and H is the

floor height.

nm3

TTT +×= （

(4)

ST =

(5)

)/(

WvDPT ∗∗=

(6)

In the formula: k is the safety factor, and the

model is set to k=1.5; T

is the time to walk to the

safety exit; T

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

is the time required

for the evacuees to take emergency actions after

responding; T

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

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

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

Municipal Affairs Branch 1

2414.68

45 37 8 45

Engineering Technology Center 12 9 3 72

Municipal Affairs Branch 2

2397.62

50 42 8 50

Garden Design Branch 35 21 14 63

Building branch 1

2414.68

45 31 14 45

Project Management Center

Building branch 2

2397.62

47 32 15 50

Transportation Branch 46 36 10 63

Planning branch 1

2414.68

51 21 30 45

Political Research Center 24 13 11 47

Planning branch 2

2384.62

50 21 29 50

information Center 15 9 6 51

Production Management

Department

2456.94

17 4 13 23

Human Resources Department 9 0 9 15

Chief Engineer Office 19 10 9 20

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

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

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

, 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

, 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

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

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

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

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

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