2011-2012 & 2012-2013 CIFE Seed Project
Civil and Environmental Engineering, Stanford University
Research Student: Mei Ling Chu
Safe egress is one of the key design issues identified by facility planners, managers and inspectors. Current computational tools for the simulation and design of emergency egress rely heavily on assumptions about human individual and social behaviors, which have been found to be oversimplified, inconsistent and even incorrect. This research aims to develop a framework for modeling human and social behaviors from the perspectives of human decision-making and social interaction and to incorporate such behaviors in a dynamic computational model suitable for emergency egress analysis. For the year-1 project, we have developed a theoretical framework of crowd behavior and a proof-of-concept, multi-agent based crowd simulation model. In the year-2 project, we plan to study the scalability and modularity issues of the simulation framework, to validate the models, and to incorporate engineering analyses, such as performance-based assessment of facilities, design of egress and emergency plans.
Methodology[Back to Top]
In order to design a simulation tool that can flexibly incorporate human and social behaviors, we will continue to conduct a systematic study on crowd behavior in emergency situations. Features and rules governing decision-making process will be extracted from social science literature and disaster studies. The information will then be used as the basis for the design and implementation of the prototype platform.
We conjecture that the process of human decision-making in emergency situations is affected by three basic factors: individual decision, the group’s pre-existing and emergent norm, and crowd characteristics. Figure 1 depicts a simplified organizational view of the features that define the occupants and their characteristics in an evacuation situation. Based on this three-level organization, features and behavioral rules are extracted from different human and social theories and are systematically classified into these three categories.
Figure 1: A simplified organization defining occupants’ behaviors and the environment
Simulating human behaviors in emergencies is a complex task that involves psychological, social, physical and environmental factors. By categorizing different behavior features at the individual, group, and crowd level, a general framework can be established. This staged representation would allow us to assume different levels of group and crowd effects on evacuating individuals and test the impact of these effects on the overall evacuation patterns. As shown in Figure 2, an occupant’s action is a result of individual preference, group norm, and crowd mood. The group and crowd behavior can affect individual decision, depending on the occupants and the situation.
Figure 2: An illustration of individual decision-making process[Back to Top]
This research extends the MASSEgress, a multi-agent based framework, to allow group influence and social behavior models to be defined and incorporated for simulation. Furthermore, extensive validation tests will be conducted to examine the features implemented and to test the behavioral theories based on real scenarios and collected data.
Figure 3: Overall architecture of the framework
Figure 3 schematically depicts the system architecture of the multi-agent simulation framework. The Global Database, Crowd Simulation Engine and Agent Behavior Model constitute the key modules of the framework. The Global Database maintains all the information about the physical environment and the agent population during a simulation. It obtains physical geometries from the Geometric Engine and sensing information from the Sensing Data Input Engine, as well as the agent population distribution and physical parameters from the Population Generator. The Agent Behavior Model contains the agent decision profiles and agent group information. The Global Database and the Agent Behavior Model interact with each other through the Crowd Simulation Engine, which generates visual output and event logs.
The simulation platform is designed to simulate egress patterns both in emergency situations and under normal circumstances. We plan to validate and test our framework with statistical and archived data of prior disaster events based on historical records and with real life data on people movements from industrial collaborators. Many of historical accidents in office buildings and facilities have been studied by social scientists and disaster management researchers. Many of such reports and videos are now publicly available, for example, the ESRC research project on past emergency events (Cocking and Drury 2008) and the NIST investigation of the World Trade Center disaster (Averill et al. 2005). Furthermore, researchers at Disney R&D have agreed to collaborate on this research, particularly in validation data sharing. Through this collaboration (which also has a strong partnership with GAMMA program at the University of North Carolina), we will make use of the state-of-the-art technology in Disney R&D to record real-life data about occupants’ trajectories and group interactions that are particularly useful to our validation tests about group behavior.
Figure 4 shows the proposed validation process. First, demographic data and group statistics are collected as model input. Defining different social behavior models and environmental conditions, simulations will be conducted to produce visual outputs and to facilitate analysis (such as statistical clustering patterns and evacuation routes). The results will then be compared with observed crowd patterns.
Figure 4: Proposed validation process[Back to Top]
The data collected so far has been mainly in the format of videos footage. While the theme of our study is emergency evacuation, it is extremely difficult to take videos during emergencies because the occurrence of emergency events is relatively unpredictable and often dangerous. Although there is CCTV footage which possibly capture the events, access to emergency evacuation data is usually restricted and is not available for public use. Occasionally, for large-scale emergency events, such as the stampede which occurred in the 2010 Love Parade, videos taken by witnesses can be found online. Even though these videos capture only local crowd patterns, they lend many insights on the human responses in real emergencies. Our plan is to continue collect and archive emergency videos for analyses. If necessary, qualitative records such as surveys, interview scripts and guidelines would be collected as well.
In collaboration with our industrial partner, Disney Imagineering, and University Public Safety Department, we have started to systematically and strategically collect crowd movement data, especially during mass gatherings at public areas. We have conducted field experiments during the opening in one of the highest traffic intersections at Disneyland Anaheim, CA, and at an evacuation drill in one of the Disney parks. We will continue to work collaboratively with Disney R&D to collect data from outdoor park areas and in certain indoor facilities. In collaboration with the on-campus public safety office and event management office, we plan to gather new data for sports arenas.[Back to Top]