Agent-Based Simulation of the Infection Risk in Variable Indoor Geometries

In this paper, we introduce an agent-based pedestrian simulation with aerosol modeling, which we use for analyzing the risk of infection with airborne diseases, with special attention to indoor scenarios and the corresponding geometry. For our analysis, we simulate a realistic supermarket scenario, and analyze the influence of geometric factors for the risk of infection regarding aerosol concentration. Using such a defined set of geometry allows for a targeted analysis of risk factors. Specifically, we examine if angular structures bear higher viral loads than flat structures, which is confirmed by our experiments. An artificial neural network (ANN) specifically trained on simulation data is able to identify adjacent geometric structures based on aerosol concentration with up to 94% accuracy.