Study on Emergency Evacuation and Intervention Behavior of Population with Spatial Knowledge Integrity


 The level of spatial knowledge integrity of a population is crucial for fire escape behavior. The use of appropriate interventions for people with different levels of spatial knowledge can effectively improve evacuation efficiency. However, different emergency situations also have different effects on evacuation behavior. In this paper, we combine spatial knowledge integrity, intervention behavior and emergency situations in a fire evacuation study. To complete this study, 128 participants were recruited using VR technology, classified into spatial knowledge completeness, and studied crowd evacuation through different intervention behaviors in different simulated emergency scenarios. The results of the study showed that participants with complete spatial knowledge had shorter evacuation distances and times. Secondly, leader interventions guided evacuation better for participants with incomplete spatial knowledge in low-hazard emergencies, while range interventions were better for participants with complete spatial knowledge. Thirdly, in high-risk emergencies, leader intervention was better than range intervention for evacuation, regardless of spatial knowledge completeness. Fourth, the interaction between spatial knowledge completeness and intervention behavior was significant, positively influencing the evacuation time and distance of participants.


Introduction
Fire is a great calamity. For example, the 1115 re in Shanghai killed 58 people 11 . Due to the rapid progress of urbanization, rapid development of commercial economy, the scale of a single building is constantly expanding, and the internal structure is becoming more and more complex. To improve the e ciency of evacuation and reduce casualties, the problem of evacuation in emergency has become a research hotpot.
The e ciency of evacuation in emergency depends on the completeness of spatial knowledge 19 .Having good spatial knowledge helps to reduce evacuation time and distance 23,35 .The in uence of spatial knowledge integrity on evacuation is mainly re ected in the following aspects: rst, the evacuation e ciency of participants with different spatial knowledge integrity is different. Participants with complete spatial knowledge were able to nd evacuation routes more quickly 23 .The participants with incomplete spatial knowledge cannot identify the environment, which leads to the di culty of evacuation 35 .Therefore, participants with complete spatial knowledge are more likely to reach the safe area than participants with incomplete spatial knowledge 16 .Secondly, people with different spatial knowledge have different evacuation characteristics, such as: people with complete spatial knowledge rely on autonomous evacuation 43 .People with incomplete spatial knowledge were evacuated with the surrounding crowd 19 .Thirdly, people with different levels of spatial knowledge tend to choose familiar evacuation routes 36 .People in emergency situations tend to choose familiar routes and exits rather than unfamiliar routes 1,34 .However, depending only on the spatial knowledge integrity of individuals to choose evacuation routes may lead to crowd congestion 46 , outlet blockage 5 , evacuation e ciency reduced 1, 34 and issues et al. Therefore, some scholars in the study of spatial knowledge based on the integrity of leaders, signs and crowd ow patterns and other factors to solve the existing problems in crowd evacuation 19,23 .However, the impact of spatial knowledge integrity on population evacuation under different dangerous emergency degrees is unclear.
Effective intervention can improve the e ciency of crowd evacuation. Intervention refers to the behavior that the leader himself or the leader provides the security information to the crowd in the emergency situation through the equipment 42,46 .There are two kinds of intervention: leader intervention and scope intervention 42,46 .Leader intervention is the behavior of trained leaders directly entering the site to provide evacuation information and guide people from dangerous areas 24,27 .Scope intervention refers to the act of providing people with security information such as routes and exits through broadcasting 21 and marking 24 .The scenarios for leader intervention are as follows: rst, people cannot accurately access environmental information. Because the vision of the people is limited by the smoke or the lack of illumination, the leaders' intervention becomes the main guiding behavior 7 . Secondly, the information acquired by personnel is uncertain. When a re occurs in the tunnel, the warning information obtained by the personnel through the FM radio is uncertain, which leads to the invalidation of the scope intervention guidance 4 . The scope of intervention is as follows: rst, the disabled escape. Disabled people with visual impairment are sensitive to LED signs in res and can be effectively identi ed for escape 3 . Secondly, the route or exit is blocked. In the event of a re or blockage at the exit, a dynamic identi cation system will display a warning message, which will be recognized by the evacuee, who will change the direction of movement and nd a new evacuation route 9 . Thirdly, there are not enough leaders. In large public places, such as stadiums, due to the lack of leaders, people mainly by identifying signs to evacuate 23 . Although the effect of intervention behavior on evacuation in emergency has been studied, the effect of intervention behavior on evacuation in different dangerous emergency degree is not clear.
The degree of spatial knowledge integrity is an important factor affecting intervention behavior. For people with incomplete knowledge of space: rst, the leader intervenes. Leaders directly into the site to conduct emotional evacuation and provide evacuation routes 42 , the experiment proved to be effective in guiding the evacuation of personnel 5,10 .But the number of leaders should not be too many, generally 10% of the total number of people for the best 26,27 . Second, scope intervention. The leader intervenes all the individuals in the space by means of information diffusion, such as broadcasting, and asks them to evacuate in order according to the signs 42 .For example, in large public places, the crowd for the environment is mostly unfamiliar, in case of emergency mainly follow the signs to evacuate 23 .For those with a complete knowledge of space: rst, the intervention can enhance their familiarity with the route. Be familiar with route blocking and be able to navigate other routes 27,46 .Secondly, when the re occurs, the smoke will affect people's line of sight, in different visibility, light or sound can effectively guide the evacuation 37 .Finally, with the development of intelligence, the intelligent guiding evacuation system and signs developed for crowd evacuation also have good results 2 .However, the effect of the interaction of spatial knowledge integrity and intervention behavior on crowd evacuation is still unclear.

Page 4/18
The degree of urgency of different risks is another important factor in uencing intervention. The evacuation behavior is different in different dangerous emergency degree, such as: In Low Dangerous Emergency, people tend to scatter to escape 44 .In high-risk emergencies, people tend to congregate, often following others to escape, causing crowds, blocked exits and ine cient evacuations 7 .The intervention behavior is different for different emergency degree. For example, in low risk emergencies, re information is provided to people through smoke detector and sound alarm systems to improve evacuation e ciency 6,8 .In high-risk emergencies, the ability to move people is signi cantly reduced, and leaders can more effectively provide evacuation guidance and reduce casualty rates 6,8 .However, in real life, the participants' spatial knowledge integrity, intervention behavior and the degree of danger and urgency are integrated into one, for example, subway, shopping mall re in the environment of people with different spatial knowledge integrity 19,32 , studying the intervention behavior of this emergency has realistic signi cance.
To test the behavior of crowd evacuation in re emergency, there are four methods in the previous research: First, computer simulation. For example, by building a model to simulate the dynamic behavior of crowd evacuation, study the e ciency of crowd evacuation 15,25,29 .Or include in uencing factors to the underlying model, such as the effect of leaders on population evacuation 36 ,Effect of emotional contagion on crowd evacuation 22 .Second, the questionnaire survey. Investigate crowd behavior and characteristics, as well as crowd self-organization phenomena during evacuation, such as route and exit choice 14 38,39 .However, these methods have some limitations, such as lack of real population data or incomplete data, high cost, di cult to set up a controlled experimental environment, and so on. So virtual reality (VR)-based immersive virtual environments (Ives) offer an alternative to crowd evacuation studies. Virtual reality is "A real or simulated environment in which the perceiver experiences telepathy" 37 .The evacuation behavior observed in virtual reality environments is qualitatively comparable to real-world behavior 17 .
IVE is an effective research tool with reasonable ecological validity, which can arouse human emergency behavior. It can control and manipulate key variables according to experimental design, and collect reliable qualitative and quantitative behavior data 18,47 .In this paper, the method based on IVE is used to solve three problems in this study. First, what is the effect of spatial knowledge integrity on evacuation

Participants
The participants in this study were 128 students or social participants from Sichuan Normal University, China (average age =20.2, ranging from 18 to 22 years; 64 men and 64 women). They had normal or corrected to normal vision (Visual acuity above 1.0) and made no difference in color identi cation (noncolor-blind color). Participants were required to sign an informed consent form prior to the start of the experiment. Following this, participants were asked to complete the pre-experiment questionnaire, including questions about participant basic information, positive effects, and negative impact PANAS scale 40 . Each subject was paid 50 CNY. The study was approved by the Ethics Committee of Sichuan Normal University. All experimental procedures were carried out in accordance with the Declaration of Helsinki.

Instruments
This experiment was carried out at the Sichuan Normal University. We used the VR system, and the integrated resolution of the binocular display was (1080 (horizontal) × 1200 (vertical)) pixels. We used 3D Studio Max software to model and render IVE (immersive virtual environments) and then import it into the platform of the Unity3D gaming engine. The Unity3D particle system was also used to visualize the spread of re and smoke in cases of re emergency experiments, using headphones to provide emergency broadcasting and re sound.
During the experiment, participants maintained a standing position, interacting with the IVE using the operating lever. Speci cally, they manipulate the lever to move in the IVE at a constant speed of 1.2m/s, moving forward, backward, left, or right at their own speed while keeping the body still.

Virtual display
This study used the at oor of a virtual large shopping mall displaying a variety of different commercial shops. The mall is 50 meters long, 40 meters wide and 4 meters high. As can be seen from the scene diagram of Figure 1, there are two entrances, and two emergency exits in the mall. In this large virtual shopping mall, there are 16 shops, all of which are in an open state. Furthermore, participants were not allowed to cross the walls of the room.
As shown in Figure 1, there is a promotion point symbol (marked by red police ag), with a re point is the promotion point, as an experiment of re emergency. This paper simulates two experimental scenarios, each where re and smoke spread across the 3mx4m area of the ground, which can be seen from anywhere in the mall ( re and smoke in Figure 2)

Experimental process
The study adopted the 2(spatial knowledge integrity: complete spatial knowledge, incomplete spatial knowledge) X 2(risk emergency: low emergency, high emergency) X 2(intervention behavior: leader intervention, scope intervention) model among the participants. In this experiment, subjects were divided into two large groups with equal numbers and equal male to female proportions, each divided into four more groups, respectively A, B, C, D, E, F, G, H. The rst group was asked to complete two tasks namely promotion and evacuation, and the second group only required evacuation. The promotion point task is to be familiar with the mall environment along the speci ed route ( Figure 3) and needs to nd the correct one from 6 maps, consisting of 5 interference maps and one correct map, formed on the correct map by adding or deleting some building elements (e. g., outlet and store location) or changing the location of existing building elements. The evacuation mission is an emergency escape mission at the promotion point, (there will be 1 minute of adaptation time before performing the mission). The screen appears: "The re occurs, please evacuate as soon as possible". Evacuation mission began and participants performed emergency evacuation by selecting different evacuation routes. (Figure 4) The experiment was conducted in low-risk and high-risk emergencies, with the low-risk emergency in the low re and smoke environment with 70% visibility in the smoke scenario and the re and smoke environment, with 10% visibility in the smoke scenario ( re and smoke in Figure 2). In addition, crowd evacuation is guided through leader intervention or scope intervention. Leader intervention refers to the leader directly entering the mall to provide emotional comfort and evacuation guide to participants.
Scope intervention is to broadcast reminder at each route from the re point to the exit "do not panic, orderly evacuation according to the evacuation sign".

Experimental validation
The experimental process is shown in Figure 5. The gure shows that upon arrival at the laboratory, participants were required to sign a consent form before starting the experiment (Figure 3).Following this, participants were asked to complete the pre-experiment questionnaire, including questions about participant basic information, positive effects, and negative impact PANAS scale 40 .Next, participants were asked to read the instruction manual on how to use IVE navigation in the system. They then need to perform the training tasks. They will wear HMD, immersion in an empty open space to simply demonstrate IVE, to be familiar with the operation of VR devices and immersion in the virtual environment.
Next, participants in the rst group were asked to understand instructions for the promotional point task and then perform the promotional point task as instructed ( Figure 1).Participants were not told whether the map they chose correct after the end of the promotional point task, however, only participants who chose the correct map were considered the evacuation outcome valid.
All participants were then asked to perform evacuation missions at the virtual mall and the re required escape to the exit as soon as possible. After reading the instructions for the evacuation task, subjects needed to perform the evacuation task in an immersive environment. (There will be a minute of adaptation before the mission), the screen appears: "The re occurs, please evacuate as soon as possible". The evacuation mission began, and virtual res and smoke emerged. There will be different emergencies and interventions for different groups. Two A, B groups in the A, B, C, D of the rst group would be placed in a low-hazardous emergency and two C, D groups in a highly hazardous emergency. For A for leader intervention, B for scope intervention, C for leader intervention and D for scope intervention. For the two E, F groups in the E, F, G, H of the second group that were placed in a low hazardous emergency, the two G, H groups were placed in a highly hazardous emergency. For E for leader intervention, F for scope intervention, G for leader intervention and H for scope intervention.
At the end of the experiment, participants were asked to ll out the post-experiment questionnaire PANAS 40 . In addition, the following questions should be answered (5 points): the vividness of the test scene, the walking di culty of the planned route, and the vertigo.

Results
Route selection (using one-way ANOVA) (Table 1) The evacuation route choice of participants in different scenarios was summarized, and the results were analyzed by one-way ANOVA. And it turns out, both the intervention behavior (F=21.847, P=7.472E-06<0.05) and the degree of danger emergency (F=4.105, P=0.0450<0.05) signi cantly in uenced the participants' choice of evacuation route, the impact on evacuation routes is independent, and the impact of interventions on route selection is greater than that of hazardous emergencies.
However, spatial knowledge integrity (F=2.170, P=0.143>0.05) did not signi cantly affect the participants' route choice, which is not in line with our conventional cognition, and the study of its in uencing factors has practical signi cance.
Route selection in Table 1 shows that participants with incomplete spatial knowledge have more evacuation route choices than participants with complete spatial knowledge; different intervention behaviors in the re can also affect participants to choose different routes for evacuation.

Completeness of spatial knowledge
In this study, we assessed participant evacuation performance by evacuation distance (m) and evacuation time (seconds). The mean values of the two dependent variables are shown in Figure 6.This paper on the evacuation distance and time of evacuation task 2(spatial knowledge integrity: complete spatial knowledge, incomplete spatial knowledge) X 2(dangerous emergency: low emergency, high emergency) X 2(intervention behavior: leader intervention, scope intervention) experiment. The results show that the degree of spatial knowledge integrity has signi cant effect on evacuation distance (F=7.924, p=0.006<0.05). Participants with complete spatial knowledge had shorter evacuation distance than those with incomplete spatial knowledge(35.1m vs 39.8m). The results also reveal that the degree of spatial knowledge integrity has a signi cant effect on evacuation time, F=3.842, p=0.005<0.05, This indicates that the evacuation time of participants with complete spatial knowledge is shorter than that of participants with incomplete spatial knowledge (47.9s vs 55.5s).

Intervention behavior
In Figure 7, This study shows that different interventions have a signi cant impact on the evacuation distance(F=104.568,p=3.575E-18<0.05).The results showed that participants in the leader intervention had shorter evacuation distance than those in the scope intervention 30m vs 43.3m).The different intervention behavior also has the remarkable in uence on the evacuation time, F=72.922, p=4.077E-14<0.05, this indicated that the participants' emergency evacuation time under leader intervention was shorter than that under scope intervention (35.9s VS 64s)

Spatial knowledge integrity and intervention in behavioral interactions
The results show that the interaction of spatial knowledge integrity and intervention behavior has a signi cant effect on evacuation time (F=30.027,P=2.277E-7<0.05) and evacuation distance (F=5.230,P=0.024<0.050), in which the participants can escape more quickly and improve the evacuation e ciency.

Emergency degree
This paper also conducts for emergency evacuation under different danger emergency 2(spatial knowledge integrity: spatial knowledge, spatial knowledge) X 2(dangerous emergency: low emergency, high emergency) X 2(intervention behavior: leader intervention, scope intervention) experiment. The results show that the degree of danger emergency has no signi cant effect on evacuation distance (F=2.721, p=0.102>0.05) Results were shown in Figure 8, participants in low-risk emergencies had longer evacuation distances than those in high-risk emergencies(38.9 m vs 36.1 m) .The degree of danger emergency also has a signi cant effect on the evacuation time, F=14.408,p=0.0002<0.05, This suggests that participants in a low-risk emergency were evacuated for longer than those in a high-risk emergency(59s vs 44.5s).

Subjective evaluation
Two participants in the study failed to select the correct sketch and their data were excluded from the results. The following analysis considered a total of 128 participants. Independent sample tests of potential confounding factors showed no signi cant differences among groups.

Discussion
In uence of the completeness of spatial knowledge on re emergency evacuation In summary, the results of this study show that the participants' evacuation route choice is determined by the degree of spatial knowledge integrity. Speci cally, participants in an emergency evacuation tend to take familiar routes, and participants with complete spatial knowledge are more likely to seek out and choose the best routes 11,35 . (Table 1 Evacuation routes) The vast majority of participants (77) with complete spatial knowledge were evacuated via the shortest route (Line 2). Because participants with complete spatial knowledge understand the whole environment, they can use analytical decision systems to make informed path-nding decisions 21,33 .For the participants with incomplete spatial knowledge, the choice of evacuation route is more random and unpredictable than the participants with complete spatial knowledge 35 (Table 1 Evacuation routes). Participants with complete spatial knowledge are therefore able to make optimal decisions and nd the best routes to reduce hazards when they occur.
In addition, the evacuation e ciency of participants is also determined by the degree of spatial knowledge integrity. The results show (Figure 6) that the participants with incomplete spatial knowledge had longer evacuation distance and time in the evacuation task. Under the condition of incomplete spatial knowledge, the evacuation time (55.5s) and the evacuation distance (39.8m) were longer than the participants' evacuation time (47.9s) and the evacuation distance (35.1m).Under the condition of incomplete spatial knowledge, the participants need more time to perceive the environment and hesitate to make the route decision, which makes the escape more di cult 35 .And having complete spatial knowledge reduces hesitation, reduces di culty, and makes evacuation easier to complete 12,35 .So people need to pay more attention to the surrounding information in daily life, remember the signi cant markers, to achieve the space and self-integration state.
Effect of the intervention behavior on the re emergency evacuation According to the above results, the leader intervention is more effective for crowd evacuation. Figure 7 shows that Participants in the leader intervention condition were evacuated for shorter distances (30 m vs 43.3 m) and time (35.9s vs 64s) than participants in the range intervention condition. Because leaders were able to directly enter the venue to guide participants, making security information spread more e ciently 24 .Under the scope of intervention, participants are slow to accept information, and participants need to take the initiative to seek environmental information, resulting in uncertainty in the decisions made, low evacuation e ciency and increased mortality.
Regardless of the intervention behavior, there were implications for participants' escape route selection (Table 1 Evacuation routes). Leaders who are familiar with the environment will give participants environmental information and lead participants to evacuate from the shortest route 24 . Thus, participants guided by leader intervention behavior all chose to evacuate from the shortest route. However, when participants are guided by range intervention, participants need to actively perceive safety information such as identi cation, before the direction and speed of movement change dynamically 45 .

Interaction effect of spatial knowledge and completeness of intervention behavior
The interaction of the degree of spatial knowledge completeness with the intervention behavior had a signi cant effect on the evacuation time and distance of the participants. The ndings show that for participants with complete spatial knowledge, leader intervention behavior can pacify participants and reduce evacuation time and distance. and alternative route guidance when familiar routes are blocked 27 .
Scope intervention can again determine the correctness of evacuation routes by identifying and broadcasting, enhancing participants' familiarity with routes and reducing evacuation errors 24 .
For participants with incomplete spatial knowledge, the effect of leader intervention behaviors was more effective on evacuation than scope intervention. Leaders can directly lead participants to evacuate from the shortest route, reducing the time for participants to nd escape routes and improve evacuation e ciency. Scope intervention allowed participants to nd stable and reliable guidance information, such as emergency export instructions 24,41 .However, since participants do not know environmental information and require longer searching for identi cation, participants may take longer to nd the right route, so participants relied more on leader instructions for evacuation 13 , so scope interventions are less e cient than the leader intervention in evacuation.

Effect of dangerous emergency degree on re emergency evacuation
This study demonstrates that participants in highly dangerous emergencies are more likely to escape.
The evacuation time and distance of participants in low hazardous emergencies were greater than the time and distance of population evacuation in highly hazardous emergencies (59s vs 44.5s 38.9m vs 36.1m).It was previously proposed that crowds panic in highly dangerous emergencies and that pedestrian mobility will be signi cantly reduced 45 .However, this study proved that under high emergencies, population evacuation is more e cient and has shorter evacuation time and distance, which is inconsistent with previous studies.
In a low-degree dangerous emergency, the crowd can evacuate more calmly because it is not greatly affected, and the highly dangerous emergency will strain the crowd and lead to accelerated action, so the evacuation time is greater than the time of people to walk in a highly dangerous emergency 45 (59s vs 44.5s).
Evacuation distance decreases in high hazard emergencies compared to low hazard emergencies (38.9m VS 36.1m).Because of highly dangerous emergencies, participants form cluster states that favor collective escape 7 .In low-risk emergencies, participants tended to escape with increased randomness of route selection 44 , resulting in reduced population evacuation e ciency.

The three interactions
In conclusion, the interactive effects of spatial knowledge integrity, intervention behavior, and different emergencies have a signi cant impact on population evacuation. In low-risk emergencies, for participants with incomplete spatial knowledge, leader intervention provides better guided population evacuation (31.7s vs 74.2s 30m vs 45.8m) relative to scope intervention. Participants with incomplete spatial knowledge needed more time to nd escape routes due to a lack of knowledge of environmental information. Although labeling was important for participants to nd escape routes, leaders were able to guide evacuation more e ciently 10 .
For participants with complete spatial knowledge, scope intervention provides better guided evacuation behavior (44.6s vs 61.9s 30m vs 41.2m) relative to leader intervention. Participants with complete spatial knowledge tend to choose "safe" routes that they know, rather than seeking additional environmental cues, trying alternative routes that they are unfamiliar with 19 .So although leader intervention can lead participants to evacuate from the best path, people are still more likely to con rm their correct decisions by identifying familiar routes.
In highly dangerous emergencies, leader intervention performs better evacuation guidance than scope intervention (32.5s VS 56.8s 30m VS 42.3m), whether spatial knowledge is complete. In highly dangerous emergencies, panic reached a maximum, decreased perceptual function and increased loss of orientation, resulting in di cult evacuation decision-making and preventing the population to escape as soon as possible 27 .And the smoke can interfere with the population identi cation of evacuation signs 27 . But trained leaders can remain emotionally calm in highly dangerous emergencies and make correct decisions rationally to guide effective evacuations.

Conclusions And Limitations
Page 12/18 The results show that: rst, the distance and time of participants with complete spatial knowledge are shorter in emergency evacuation. Second, in a low-risk emergency, leader intervention is better for participants with incomplete spatial knowledge to guide evacuation; participants with complete spatial knowledge are better for scope intervention. Third, in highly dangerous emergencies, leader intervention evacuation better than scope intervention whether spatial knowledge is complete. Fourth, the interaction between the completeness of spatial knowledge and the intervention mode was signi cant, with a positive effect on the evacuation time and distance of the participants.
A key advantage of our study is using VR technology for virtual reality simulation. Previous studies rely on computer simulations and often fail to realistically re ect real population evacuation behavior. The technique is applied in the experiment to realistic the evacuation behavior, and the research results are more realistic.
Another advantage of our study is the application of different emergency scenarios to experiments. Fire changed dynamically, but previous studies conducted only one simulation of an emergency scenario, and the observed population evacuation behavior was not comprehensive enough. Therefore, the study of different emergency scenarios can be more objective and accurate analyzed and enhance the practical signi cance of experiments. This paper shows that population evacuation e ciency is higher in high emergencies than in low emergencies, overturning the results of previous studies.
The results of this study have three practical implications, rstly, shopping malls have complete spatial knowledge. In emergencies, they strongly tend to follow the familiar route for evacuation, so their evacuation route choices are highly predictable. This means that the mall emergency department can investigate the composition of the personnel, predict the route of emergency evacuation according to the spatial knowledge of the personnel, and formulate emergency plans accordingly. Secondly, participants with incomplete spatial knowledge tend to move slower and stochastic road nding decision-making, suggesting that they are a major challenge for emergency evacuation. These personnel lack knowledge of the environment and rely heavily on environmental cues to make their own way-nding decisions.
Therefore, during the emergency evacuation, guidance information should be displayed through signs and staff instructions to help people with safe evacuation. Thirdly, in highly dangerous emergencies, population decision uncertainty increases due to the complex environment and the high-risk factor. The emergency department of the mall should send more leaders to intervene and directly enter the mall to lead the evacuation. In low-risk emergencies, the population tends to escape randomly, targeting participants with complete spatial knowledge, the identi cation can better ensure that participants walk the correct route. Leaders' participants with incomplete spatial knowledge, better guided evacuation due to a lack of environmental information.
Admittedly, this study also has three limitations that can be addressed in future studies. First, the limitations of intervention behavior, there are many types such as static identi cation, dynamic identi cation, trained leaders, mobile devices, mobile robots, and wireless sensor networks 46 . This paper considers only two major intervention behaviors, the trained leaders, and assistive devices. In addition, this paper only applies the intervention behaviors separately, and does not consider the impact of using at the same time two or more intervention behaviors on the re evacuation simultaneously. Secondly, reallife re is a process from small to large dynamic change. This paper only low-degree re and heavy re scenarios are considered, and the scene of re dynamic change process has not been considered.
Whether the results obtained in this paper are applicable to emergency evacuation under dynamic change re scenarios remains to be studied. Thirdly, this paper only considers the degree of spatial knowledge, but there are people with complete spatial knowledge in real life. It remains to be veri ed whether the results of this study apply to those with complete spatial knowledge. How people can improve the completeness of their spatial knowledge through spatial exploration, and the impact of the improved spatial knowledge completeness on evacuation behavior is another question worth studying. In uence of spatial knowledge integrity on evacuation