SlideShare uma empresa Scribd logo
1 de 4
Baixar para ler offline
A Survival Guide during Fire Outbreak inside a Cinema


                                                    Hsin-Huei Cheng

                                            Instructor : Demetri Terzopoulos

                                        University of California at Los Angeles



                                                         Abstract

        A crowd behavior simulation is performed in a situation where fire outbreak occurs inside a cinema. Each
        individual agent inside the simulation will decide their own escape pathways based on their local
        perception of the environments (i.e., agents and obstacles). It is found that agents who do not interact with
        their local environment (i.e. people that are stayed clam) will escape more efficiently than those who
        interact with other agents. This model allows for the dynamic simulation of different crowd behaviors in
        emergent situations, which will be useful for survival guide design.



        1. Introduction

“Oh My God! There’s a fire!”. Imagine when you are                This paper is organized as follow. The design of a
inside a cinema watching a famous movie, but                     virtual environment model is discussed in Section 2.
suddenly the fire alarm rings, what would you do?                In addition, the details of the autonomous model and
Would you cry, run or wait to get rescued? Everyone              the different variations will be presented in Section 3.
behaves differently in facing an emergency situation.            The implementation methods will be discussed in
People’s reactions ultimately influence the overall              Section 4, followed by results and discussions in
human survival rate in the incident. In the occasion of          Section 5. Finally, the conclusions and the future work
fire outbreak in an enclosed area, people were found             will be provided in Section 6.
killed by stepping on each other, instead of being killed
by the fire. Many human and environmental factors are            2. Virtual Environmental Model
coupled together that will affect the final outcomes.
The most critical factor is the personality of                    In real life, people interact with local their
individuals. Other factors like the environment, the             environments in deciding the escaping pathway.
physical health situation, the time of making decision,          Similarly, in the simulation, each agent interacts with
and crowd’s behavior etc. also play important roles.             their local environment in order to decide the best
                                                                 escaping route. Therefore, the design of the virtual
 In this project, I simulate the real time dynamics of           environment is crucial for the correct implementation
people movement inside a cinema in response to a fire            of the simulation.
alarm. Each person is an autonomous agent. They can
interact with each others, as well as the static obstacles         To begin with, the virtual cinema environment will
inside the cinema. The agents make decision for their            be briefly introduced. The plain view of the cinema
local pathways, and this information is collected in a           layout is shown in Figure 1. As shown in the figure,
central decision maker at each time step to implement            the grids in pink color represent the exits. Each agent
the agents’ motion. In particular, each agent was                selects the nearest exit at each time step, which is
randomly assigned with different properties in order to          determined by the Euclidean distance between the exit
simulate different crowd behaviors.                              and the agent. The seats are illustrated in blue girds.
                                                                 These seats are stationary obstacles, where agents can
only go around the obstacles instead of walking across     3.1 Autonomous Model
them.
                                                            The agents were represented by circular-shaped
 The virtual environment model consists of a               objects in the simulation. Agents with different
stationary map, a mobile map, and a path map (Figure       properties are represented with different colors. The
2). The stationary map records the grid locations          properties of the agents are summarized as follow:
where the obstacles (i.e. seats) occupy. This map is
stationary with respect to time. For dynamic moving            -   Physical factors: different walking speeds
agents, the mobile map stores the information for                  represent different physical conditions
occupied cells in real time. Agents that are within            -   Mental factors: selfishness and helpfulness
specific cell are recorded in the “occupied agent list”.       -   Group factor: belongs to a group or alone
The dynamic map is updated at each time step. Both
stationary and mobile maps are necessary input             Each agent possesses three different states. The initial
information for the central decision maker to              state is indicated as “sitting” state. When the agents
implement the paths. The implementation will be            are alerted by the fire alarm, their states change from
stored in the path map.                                    the “sitting” state to the “running” state. After a
                                                           defined time period, if the agents are trapped inside
                                                           the cinema, then their states will turn into the “dead”
                                                           state, which implies that the agents are killed by the
                                                           fire. For those who have successfully reached the
                                                           exits, their states will turn into “alive” states. A
                                                           summary of state changes is illustrated in Figure 3.




Figure 1. The plain view of cinema.



                                                           Figure 3. State change of an agent



                                                           3.2 Autonomous Behavior Model

                                                            The general behavior of agents can be classified as
                                                           below (Figure 4).

                                                               A. helpful agents meet helpful agents: the slower
                                                                  agent has the priority to move to the next grid
                                                               B. selfish agents meet helpful agents: selfish
Figure 2. Upper left: stationary map; upper right:                ones have the priority to move to the next grid
mobile map; lower: path map                                    C. selfish agents meet selfish agents: one of them
                                                                  will be given the priority to move
                                                               D. group members meet other individuals: group
                                                                  members have the higher priority if the group
3.       Autonomous Agent Behavior                                is already in the process of moving. Group
                                                                  members cannot be separated.
  The dynamic agents are the major character in the
simulation. Here, I will first introduce the properties
of the agents and then describe the behavior models
based on these properties.
is the next moving position of each agent at next time
                                                            step.

                                                             Since all the agents movements have to be
                                                            coordinated at the same time, a path decision maker is
                                                            required to coordinate the movement. This is achieved
   (a)             (b)          (c)            (d)          by the function “confirmPath”. For each grid on the
                                                            path map, “confirmPath” checks the occupied status.
Figure 4. (a) H1 has lower moving speed so it has the       There are some complicated rules for the
priority to move. (b) S1 has the priority to move. (c)      implementation, which are outlined as follow:
Either S1 or S2 has the priority. (d) G1 has the priority
to move if one of the group members (e.g. G0) first             A. A grid can only be occupied by one agent p
passed by the grid                                                 and “confirmPath” approves p’s next moving
                                                                   position if
                                                                - p does not belong to any group
3.3 Sensing Area                                                - p belongs to a moving group and the previous
                                                                   member just walked through the grid
In reality, people can only sense their local                   - p belongs to a group and is the leader of the
environment to make a decision. In the virtual                     group
environment, each agent is provided with the ability to
sense its local neighborhood, in order to decide the            B. in the situation where the grid is going to be
best moving direction. For instance, if the number of              occupied by more than one agent
neighbors of an agent is less than that of its right hand       - compare each candidate’s properties and
side, the agent will choose the direction where there              follow the rules described in Section 4.2 to
are fewer neighbors (i.e. higher chance to escape                  find the “winner” who can occupy the
faster). With this concept of neighborhood sensing,                grid at the next time step. The agents who are
each agent updates a sensing list which records the                “lost” in the comparison will remain at the
occupied cells within a specified area of interest at              same position.
each time step. An example of the sensing area is
shown in Figure 5.
                                                            Following the method, agents can progressively reach
                                                            the nearest goal (i.e., the exit).




Figure 5. The 2 tier sensing area of agent p. The light
blue area represents the 1st tier, whereas the dark blue
region represents the 2nd tier.
                                                            Figure 6. The pseudo codes of the method.

4. Method Description

An outline of the methodology used for computing a          5. Results and Discussion
path for an agent is provided in Figure 6. The function
“predictNext” is to find the optimal intermediate           In this section, I will illustrate some agent behaviors
positions (i.e. sub-goals) toward the final goal. The       based on the behavior model mentioned in Section
function refers to the stationary map, as well as the 2-    4.2. These demonstrations are displayed on the
tier sensing list to find the next direction that has a     websites at
higher escape probability. The output of this function      http://web.me.com/stella0107/stella/CS275_AL.html.
Different crowd behaviors were simulated based on           Acknowledgements
different agents’ properties and some interesting
phenomena were observed. For instance, Demo 3 and           I would like to thank Prof. Terzopoulos’s lectures that
4 demonstrate a situation where agents located on the       let me understand the beauty of Artificial Life. I
left hand side of the cinema do not interact with other     would also like to thank the instructor Prof.
agents, therefore they escape to the exit based on the      Terzopoulos, as well as all the people who were
shortest pathway; while agents located on the right         involved in the development of the simulator for
side are programmed to interact with other agents,          CS174/274C. Special thanks to Professor Faloutsos
therefore a longer decision making process is required      and the teaching assistant Gabriele Nataneli.
for individual agent to execute their pathway. This is
reflected from the simulated behavior model, where
some agents become hesitate to decide which way to          References
go. They interact with the dynamic environment in
real time and cannot find a good solution immediately.      [1] “Composite Agents”, Hengchin Yeh, Sean Curtis,
When people meet an emergency, they often become                Sachin Patil, Jur van den Berg,Dinesh Manocha,
too afraid or anxious to make quick and reasonable              Ming Lin, SCA2008.
decision.
                                                            [2] "Environmental Modeling for Autonomous
In this project, I did not perform some of the                  Virtual Pedestrians," W. Shao, D. Terzopoulos,
interesting features that would make the simulation             SAE 2005 Transactions Journal of Passenger
more realistic, such as collision avoidance and global          Cars: Electronic and Electrical Systems, 114(7),
path planning. In addition, one problem I want to               February, 2006, 735–742. Compilation of the
address is the grid based data structure. The grid size I       most outstanding SAE technical papers of 2005.
used for the stationary, the mobile, and the path maps
are all the same to make the computation easier and         [3] "Artificial fishes: Physics, locomotion, perception,
faster. However, this leaded to a situation where some          behavior," X. Tu, D. Terzopoulos, Proc. ACM
agents will get stuck on the boundary of the grid               SIGGRAPH'94 Conference, Orlando, FL, July,
without further movement. Owing to this limitation,             1994, in Computer Graphics Proceedings, Annual
the statistics of the crowd behaviors were not carried          Conference Series, 1994, 43–50.
out.
                                                            [4] "Artificial life for computer graphics," D.
6. Conclusion and Future Work                                   Terzopoulos, Communications of the ACM,
                                                                42(8), August, 1999, 32–42.
Crowd behavior simulation is a very interesting topic,
especially when the simulation is coupled with the          [5] “Steering     Behaviors   For     Autonomous
complex personalities of individual agents. In this             Characters”, Craig W. Reynolds, Sony Computer
project, I have demonstrated a crowd behavior model             Entertainment America.
based on individual agents’ properties, where some
interesting observations were found. Changing the
behavior models will lead to different outcomes that
may be outside our imagination. The results presented
in this report are instructive to educate people how to
escape efficiently from an emergency situation, such
as fire outbreak inside a cinema.

Future work includes the optimization of the data
structures, as well as the behavior models. The project
is a small-scale simulation model with a few agents at
the moment. The ultimate goal is to design a large-
scale simulation that can realistically simulate the real
life situation.

Mais conteúdo relacionado

Semelhante a Cinema Fire Escape Guide

Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...
Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...
Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...Kalle
 
Unit II & III_uncovered topics.doc notes
Unit II & III_uncovered topics.doc notesUnit II & III_uncovered topics.doc notes
Unit II & III_uncovered topics.doc notessmithashetty24
 
Agent architectures
Agent architecturesAgent architectures
Agent architecturesguesta6bfe2
 
Biomimicry And Fuzzy Modeling
Biomimicry And Fuzzy ModelingBiomimicry And Fuzzy Modeling
Biomimicry And Fuzzy ModelingJake Langford
 

Semelhante a Cinema Fire Escape Guide (7)

AI Lesson 02
AI Lesson 02AI Lesson 02
AI Lesson 02
 
CS4700-Agents_v3.pptx
CS4700-Agents_v3.pptxCS4700-Agents_v3.pptx
CS4700-Agents_v3.pptx
 
Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...
Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...
Mc Lendon Using Eye Tracking To Investigate Important Cues For Representative...
 
Unit II & III_uncovered topics.doc notes
Unit II & III_uncovered topics.doc notesUnit II & III_uncovered topics.doc notes
Unit II & III_uncovered topics.doc notes
 
Agent architectures
Agent architecturesAgent architectures
Agent architectures
 
Agent architectures
Agent architecturesAgent architectures
Agent architectures
 
Biomimicry And Fuzzy Modeling
Biomimicry And Fuzzy ModelingBiomimicry And Fuzzy Modeling
Biomimicry And Fuzzy Modeling
 

Último

Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Miguel Araújo
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersThousandEyes
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)Gabriella Davis
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityPrincipled Technologies
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024BookNet Canada
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Alan Dix
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Paola De la Torre
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking MenDelhi Call girls
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024Scott Keck-Warren
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonAnna Loughnan Colquhoun
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdfhans926745
 

Último (20)

Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
Neo4j - How KGs are shaping the future of Generative AI at AWS Summit London ...
 
A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)A Domino Admins Adventures (Engage 2024)
A Domino Admins Adventures (Engage 2024)
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
#StandardsGoals for 2024: What’s new for BISAC - Tech Forum 2024
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...Swan(sea) Song – personal research during my six years at Swansea ... and bey...
Swan(sea) Song – personal research during my six years at Swansea ... and bey...
 
Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101Salesforce Community Group Quito, Salesforce 101
Salesforce Community Group Quito, Salesforce 101
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
 
The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024SQL Database Design For Developers at php[tek] 2024
SQL Database Design For Developers at php[tek] 2024
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 

Cinema Fire Escape Guide

  • 1. A Survival Guide during Fire Outbreak inside a Cinema Hsin-Huei Cheng Instructor : Demetri Terzopoulos University of California at Los Angeles Abstract A crowd behavior simulation is performed in a situation where fire outbreak occurs inside a cinema. Each individual agent inside the simulation will decide their own escape pathways based on their local perception of the environments (i.e., agents and obstacles). It is found that agents who do not interact with their local environment (i.e. people that are stayed clam) will escape more efficiently than those who interact with other agents. This model allows for the dynamic simulation of different crowd behaviors in emergent situations, which will be useful for survival guide design. 1. Introduction “Oh My God! There’s a fire!”. Imagine when you are This paper is organized as follow. The design of a inside a cinema watching a famous movie, but virtual environment model is discussed in Section 2. suddenly the fire alarm rings, what would you do? In addition, the details of the autonomous model and Would you cry, run or wait to get rescued? Everyone the different variations will be presented in Section 3. behaves differently in facing an emergency situation. The implementation methods will be discussed in People’s reactions ultimately influence the overall Section 4, followed by results and discussions in human survival rate in the incident. In the occasion of Section 5. Finally, the conclusions and the future work fire outbreak in an enclosed area, people were found will be provided in Section 6. killed by stepping on each other, instead of being killed by the fire. Many human and environmental factors are 2. Virtual Environmental Model coupled together that will affect the final outcomes. The most critical factor is the personality of In real life, people interact with local their individuals. Other factors like the environment, the environments in deciding the escaping pathway. physical health situation, the time of making decision, Similarly, in the simulation, each agent interacts with and crowd’s behavior etc. also play important roles. their local environment in order to decide the best escaping route. Therefore, the design of the virtual In this project, I simulate the real time dynamics of environment is crucial for the correct implementation people movement inside a cinema in response to a fire of the simulation. alarm. Each person is an autonomous agent. They can interact with each others, as well as the static obstacles To begin with, the virtual cinema environment will inside the cinema. The agents make decision for their be briefly introduced. The plain view of the cinema local pathways, and this information is collected in a layout is shown in Figure 1. As shown in the figure, central decision maker at each time step to implement the grids in pink color represent the exits. Each agent the agents’ motion. In particular, each agent was selects the nearest exit at each time step, which is randomly assigned with different properties in order to determined by the Euclidean distance between the exit simulate different crowd behaviors. and the agent. The seats are illustrated in blue girds. These seats are stationary obstacles, where agents can
  • 2. only go around the obstacles instead of walking across 3.1 Autonomous Model them. The agents were represented by circular-shaped The virtual environment model consists of a objects in the simulation. Agents with different stationary map, a mobile map, and a path map (Figure properties are represented with different colors. The 2). The stationary map records the grid locations properties of the agents are summarized as follow: where the obstacles (i.e. seats) occupy. This map is stationary with respect to time. For dynamic moving - Physical factors: different walking speeds agents, the mobile map stores the information for represent different physical conditions occupied cells in real time. Agents that are within - Mental factors: selfishness and helpfulness specific cell are recorded in the “occupied agent list”. - Group factor: belongs to a group or alone The dynamic map is updated at each time step. Both stationary and mobile maps are necessary input Each agent possesses three different states. The initial information for the central decision maker to state is indicated as “sitting” state. When the agents implement the paths. The implementation will be are alerted by the fire alarm, their states change from stored in the path map. the “sitting” state to the “running” state. After a defined time period, if the agents are trapped inside the cinema, then their states will turn into the “dead” state, which implies that the agents are killed by the fire. For those who have successfully reached the exits, their states will turn into “alive” states. A summary of state changes is illustrated in Figure 3. Figure 1. The plain view of cinema. Figure 3. State change of an agent 3.2 Autonomous Behavior Model The general behavior of agents can be classified as below (Figure 4). A. helpful agents meet helpful agents: the slower agent has the priority to move to the next grid B. selfish agents meet helpful agents: selfish Figure 2. Upper left: stationary map; upper right: ones have the priority to move to the next grid mobile map; lower: path map C. selfish agents meet selfish agents: one of them will be given the priority to move D. group members meet other individuals: group members have the higher priority if the group 3. Autonomous Agent Behavior is already in the process of moving. Group members cannot be separated. The dynamic agents are the major character in the simulation. Here, I will first introduce the properties of the agents and then describe the behavior models based on these properties.
  • 3. is the next moving position of each agent at next time step. Since all the agents movements have to be coordinated at the same time, a path decision maker is required to coordinate the movement. This is achieved (a) (b) (c) (d) by the function “confirmPath”. For each grid on the path map, “confirmPath” checks the occupied status. Figure 4. (a) H1 has lower moving speed so it has the There are some complicated rules for the priority to move. (b) S1 has the priority to move. (c) implementation, which are outlined as follow: Either S1 or S2 has the priority. (d) G1 has the priority to move if one of the group members (e.g. G0) first A. A grid can only be occupied by one agent p passed by the grid and “confirmPath” approves p’s next moving position if - p does not belong to any group 3.3 Sensing Area - p belongs to a moving group and the previous member just walked through the grid In reality, people can only sense their local - p belongs to a group and is the leader of the environment to make a decision. In the virtual group environment, each agent is provided with the ability to sense its local neighborhood, in order to decide the B. in the situation where the grid is going to be best moving direction. For instance, if the number of occupied by more than one agent neighbors of an agent is less than that of its right hand - compare each candidate’s properties and side, the agent will choose the direction where there follow the rules described in Section 4.2 to are fewer neighbors (i.e. higher chance to escape find the “winner” who can occupy the faster). With this concept of neighborhood sensing, grid at the next time step. The agents who are each agent updates a sensing list which records the “lost” in the comparison will remain at the occupied cells within a specified area of interest at same position. each time step. An example of the sensing area is shown in Figure 5. Following the method, agents can progressively reach the nearest goal (i.e., the exit). Figure 5. The 2 tier sensing area of agent p. The light blue area represents the 1st tier, whereas the dark blue region represents the 2nd tier. Figure 6. The pseudo codes of the method. 4. Method Description An outline of the methodology used for computing a 5. Results and Discussion path for an agent is provided in Figure 6. The function “predictNext” is to find the optimal intermediate In this section, I will illustrate some agent behaviors positions (i.e. sub-goals) toward the final goal. The based on the behavior model mentioned in Section function refers to the stationary map, as well as the 2- 4.2. These demonstrations are displayed on the tier sensing list to find the next direction that has a websites at higher escape probability. The output of this function http://web.me.com/stella0107/stella/CS275_AL.html.
  • 4. Different crowd behaviors were simulated based on Acknowledgements different agents’ properties and some interesting phenomena were observed. For instance, Demo 3 and I would like to thank Prof. Terzopoulos’s lectures that 4 demonstrate a situation where agents located on the let me understand the beauty of Artificial Life. I left hand side of the cinema do not interact with other would also like to thank the instructor Prof. agents, therefore they escape to the exit based on the Terzopoulos, as well as all the people who were shortest pathway; while agents located on the right involved in the development of the simulator for side are programmed to interact with other agents, CS174/274C. Special thanks to Professor Faloutsos therefore a longer decision making process is required and the teaching assistant Gabriele Nataneli. for individual agent to execute their pathway. This is reflected from the simulated behavior model, where some agents become hesitate to decide which way to References go. They interact with the dynamic environment in real time and cannot find a good solution immediately. [1] “Composite Agents”, Hengchin Yeh, Sean Curtis, When people meet an emergency, they often become Sachin Patil, Jur van den Berg,Dinesh Manocha, too afraid or anxious to make quick and reasonable Ming Lin, SCA2008. decision. [2] "Environmental Modeling for Autonomous In this project, I did not perform some of the Virtual Pedestrians," W. Shao, D. Terzopoulos, interesting features that would make the simulation SAE 2005 Transactions Journal of Passenger more realistic, such as collision avoidance and global Cars: Electronic and Electrical Systems, 114(7), path planning. In addition, one problem I want to February, 2006, 735–742. Compilation of the address is the grid based data structure. The grid size I most outstanding SAE technical papers of 2005. used for the stationary, the mobile, and the path maps are all the same to make the computation easier and [3] "Artificial fishes: Physics, locomotion, perception, faster. However, this leaded to a situation where some behavior," X. Tu, D. Terzopoulos, Proc. ACM agents will get stuck on the boundary of the grid SIGGRAPH'94 Conference, Orlando, FL, July, without further movement. Owing to this limitation, 1994, in Computer Graphics Proceedings, Annual the statistics of the crowd behaviors were not carried Conference Series, 1994, 43–50. out. [4] "Artificial life for computer graphics," D. 6. Conclusion and Future Work Terzopoulos, Communications of the ACM, 42(8), August, 1999, 32–42. Crowd behavior simulation is a very interesting topic, especially when the simulation is coupled with the [5] “Steering Behaviors For Autonomous complex personalities of individual agents. In this Characters”, Craig W. Reynolds, Sony Computer project, I have demonstrated a crowd behavior model Entertainment America. based on individual agents’ properties, where some interesting observations were found. Changing the behavior models will lead to different outcomes that may be outside our imagination. The results presented in this report are instructive to educate people how to escape efficiently from an emergency situation, such as fire outbreak inside a cinema. Future work includes the optimization of the data structures, as well as the behavior models. The project is a small-scale simulation model with a few agents at the moment. The ultimate goal is to design a large- scale simulation that can realistically simulate the real life situation.