Collecting Data in an Immersive Video Environment to Set up an Agent-Based Model of Pedestrians’ Compliance with COVID-Related Interventions

Abstract

Setting up any agent-based model (ABM) requires not only theory to define the agents' behavior, but also suitable methods for calibration, validation, and scenario analysis, which are highly dependent on the available data. When modelling aspects related to the COVID-19 pandemic during the pandemic itself, finding existing data and behavioral rules was rarely possible as conditions were fundamentally different from before and collecting data put people at risk. Here, we present a method to set up and calibrate an ABM using an immersive video environment (IVE). First, we collect data in this reproducible and safe setting. Based on derived behavior, we set up an ABM of pedestrians responding to one-way street signs, installed to stimulate physical distancing. Using bootstrapped regression, we integrate the IVE data into the ABM. Model experiments show that the street signs help to reduce pedestrian densities below critical distance-keeping thresholds, though only when the number of pedestrians is not too high. Our work contributes to the understanding of pedestrian movement dynamics during pandemics. In addition, the proposed data collection and calibration method using the IVE may be applied to other simulation models in which effects of interventions in the physical environment are modelled.