Bridging the Gap between Social and Transportation Networks: An Integrated, Dynamic Evacuation Decision Making Model Open Access
Downloadable ContentDownload PDF
Human behavior during an emergency evacuation is complex and difficult to predict accurately. The complexity of human behavior in these situations lies in the social interactions that have been shown to have a significant impact on decision making. Furthermore, when a hazardous situation arises and evacuation is necessary, there is a requirement for a well prepared system that ensures the safety and security of individuals in impacted regions. From a transportation engineering perspective, the main goal is to move as many individuals as possible to shelter or more secure location. Sociologists and psychologists focus on understanding the wide range of human behaviors under these conditions. Based on the lack of cohesion between these disciplines, the main objective of this research is to present a framework that combines socio-psychological and transportation engineering considerations in order to determine how individual characteristics influence both social network formation and evacuation decision making under extreme conditions. By understanding and including the cascading effects of an idea throughout the social networks in a given area, a more complete evacuation framework is achieved. The need for a re-evaluation of current evacuation strategies and the associated demand modeling techniques has been identified by both research practitioners and the United States Government. From a general perspective, evacuation plans are outdated, new technologies are underutilized and expertise is lacking or misused. Compounding matters is a lack of cohesion both within government agencies and research practitioners; the former leading to disjointed planning efforts and the latter to mathematical models that do not capture the evacuation process in its entirety. Theoretical formulation for estimating evacuation demand needs to be revisited, with particular attention being given to the dynamic aspects of the decision making process. Specifically, models that capture the decision process in its entirety should account for mode and destination choice as well as the influence of social factors on all decisions. Research in this dissertation represents the first steps in addressing these deficiencies within the evacuation decision making framework.Research on the microscopic level aims at expanding upon current social network modeling practices in order to identify both marital and social links within the population. In both cases network evolution models are applied to the dataset whereas networks are borne out of seed networks consisting of a subset of the population. After a social network has been effectively estimated for the given region, microscopic and macroscopic data are combined within the framework of a decision making model in order to transform the cascading effects of social influence into binary decision variables; allowing for estimation of potential demand loading on the transportation network during an emergency evacuation. Here, mathematical formulation considers both an individual’s specific characteristics and the influence from linked individuals within the region. This stage of the model features considerations for not only the initial stay/leave decision that an individual must make, but also the modal and destination decisions that are inherent to an evacuation scenario. In order to calibrate the models developed in this research, survey data from over 250 individuals in the Washington, DC region was collected and extensively analyzed. Using this data, a genetic algorithm calibration procedure was applied in order to calibrate evacuation demand timing and mode and destination decisions. Simulation results demonstrated that allowing for agents to communicate with one another produced accurate demand loading curves as well as mode and destination matrices. Furthermore, the impact of individual demographic variables, which are typically used to simulate evacuation, were successfully captured by model results.