Simulating Indoor Evacuation of Pedestrians: The Sensitivity of Predictions to Directional-Choice Calibration Parameters

Abstract

The increasing occurrence of safety-related incidents like fire and terror attacks in crowded public facilities and mass gatherings has heightened the importance of planning for efficient evacuations through optimizing evacuation routes and architectural designs. This calls for the development of simulation and analytical tools that can replicate occupants’ responses and thereby their most likely movement patterns. Such models must be accurate to prevent inappropriate design and planning. One major factor connected to prediction accuracy is the sensitivity of modeling outputs to the value of their various parameters. We report on implementation of a calibrated model of directional choices in a microscopic simulation model of pedestrians’ evacuation. We show how estimates of the aggregate measures of prediction are sensitive to the parameters of this tactical level (i.e., directional choice) model. Results demonstrate that the prediction of the total evacuation time and average individual evacuation times are closely correlated with one another in terms of their variation, and are both very sensitive to the specification of each directional choice parameter. Simulated evacuation time could vary up to nearly 30% depending on parameter values. The observed sensitivity highlighted the significance of importing well-calibrated parameters into such simulation models and practicing consistent degrees of accuracy for all levels of decision modeling. We also inferred that the two aggregate measures (i.e., total evacuation time and average individual evacuation times) can be used interchangeably as the basis for evacuation optimization or sensitivity analysis practices.