Combining Agent-Based Models with Virtual Reality to collect micro-level behavioural data in a smoke evacuation situation

Abstract

The main research objective of this thesis was to fill the gap in micro-level behavioural data, that is missing to improve agent-based models. Furthermore, it aimed to enhance an evacuation model with a dynamic smoke environment. The dynamic smoke model was implemented to a static evacuation model, and the behaviour of the agents was adapted. New variables were introduced to the agents, so that they could perceive risk and cope with these risks. The micro-level behaviour data was collected by performing experiments in Virtual Reality. The experiments focussed on three elements: general behaviour when encountering smoke, smoke density and the location of the smoke. In the experiments, a mixed 5x2 design was used. Each participant went through five VR simulations, which were built in Unity3D, with either a low density or a high density. By conducting experiments with the evacuation model, the results of the VR experiments were compared to the model. Results were similar on average, however, differences in the environments made it difficult to compare. Future research should be focussed on a more realistic smoke model, a more realistic VR environment that matches the evacuation model better and a larger sample size in the experiments.