TY - GEN
T1 - Multi-agent Modeling of Human Traffic Dynamics for Rapid Response to Public Emergency in Spatial Networks
AU - Shi, Xiaoru
AU - Lee, Hankang
AU - Yang, Hui
N1 - Publisher Copyright:
© 2024 IEEE.
PY - 2024
Y1 - 2024
N2 - Public emergencies pose catastrophic casualties and financial losses in densely populated areas, rendering communities such as cities, towns, and universities particularly susceptible due to their intricate environments and high pedestrian traffic. While simulation analysis offers a flexible and cost-effective approach to evaluating evacuation procedures, conventional evacuation models are often limited to specific scenarios and communities, overlooking the diverse range of emergencies and evacuee behaviors. Thus, there is an urgent need for an evacuation model capable of capturing complex structures of communities and modeling evacuee responses to various emergencies. This paper presents a novel approach to simulating responsive evacuation behaviors for multiple emergency situations in public communities through spatial network modeling and multi-agent modeling. Leveraging a community network framework adaptable to different community layouts based on map data, the proposed model employs a multi-agent approach to characterize responsive and decentralized evacuation decision-making. Experimental results show the model's efficacy in representing pedestrian flow and pedestrians' reactive behavior across various campuses based on real-world map data. Additionally, the case study highlights the potential of the proposed model to simulate pedestrian dynamics for a variety of heterogeneous emergencies. The proposed community evacuation model holds strong promise for evaluating evacuation policies and providing insights into resilient plans during public emergencies, thereby enhancing community safety.
AB - Public emergencies pose catastrophic casualties and financial losses in densely populated areas, rendering communities such as cities, towns, and universities particularly susceptible due to their intricate environments and high pedestrian traffic. While simulation analysis offers a flexible and cost-effective approach to evaluating evacuation procedures, conventional evacuation models are often limited to specific scenarios and communities, overlooking the diverse range of emergencies and evacuee behaviors. Thus, there is an urgent need for an evacuation model capable of capturing complex structures of communities and modeling evacuee responses to various emergencies. This paper presents a novel approach to simulating responsive evacuation behaviors for multiple emergency situations in public communities through spatial network modeling and multi-agent modeling. Leveraging a community network framework adaptable to different community layouts based on map data, the proposed model employs a multi-agent approach to characterize responsive and decentralized evacuation decision-making. Experimental results show the model's efficacy in representing pedestrian flow and pedestrians' reactive behavior across various campuses based on real-world map data. Additionally, the case study highlights the potential of the proposed model to simulate pedestrian dynamics for a variety of heterogeneous emergencies. The proposed community evacuation model holds strong promise for evaluating evacuation policies and providing insights into resilient plans during public emergencies, thereby enhancing community safety.
UR - http://www.scopus.com/inward/record.url?scp=85208228174&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85208228174&partnerID=8YFLogxK
U2 - 10.1109/CASE59546.2024.10711522
DO - 10.1109/CASE59546.2024.10711522
M3 - Conference contribution
AN - SCOPUS:85208228174
T3 - IEEE International Conference on Automation Science and Engineering
SP - 374
EP - 380
BT - 2024 IEEE 20th International Conference on Automation Science and Engineering, CASE 2024
PB - IEEE Computer Society
T2 - 20th IEEE International Conference on Automation Science and Engineering, CASE 2024
Y2 - 28 August 2024 through 1 September 2024
ER -