When Routes Fail: Redistribution After Network Edge Removal in a Virtual Reality Evacuation Experiment

Sudden infrastructure failures can disrupt evacuation networks and force rapid route reconfiguration under time pressure. While evacuation research often focuses on travel times or path efficiency, less attention has been given to discrete route choice behavior following abrupt topological disruption. This study investigates how individuals redistribute across alternative paths after the removal of a critical network edge in an immersive virtual reality (VR) flood scenario based on an urban district in Frankfurt, Germany. Participants (N = 128) initially followed a designated evacuation route. Midway, a bridge collapse removed a key network edge, forcing rerouting at a downstream decision node with three available outgoing branches. We model the street system as an undirected graph with locally directional choice sets and analyze resulting route choices. We quantify redistribution using empirical branch selection probabilities, a chi-square test against uniform allocation, and Shannon entropy as a concentration measure. Results reveal stable and non-uniform redistribution patterns across branches, with branch preferences aligning with simple geometric properties of the available alternatives, including directional continuity, path length, and detour. Redistribution was largely invariant to demographic and experience-related attributes, suggesting a dominant role of geometry in this setting. Even in the absence of crowd interaction, consistent branch distributions emerge. These findings highlight the importance of topological structure in shaping evacuation behavior under disruption and demonstrate how immersive experiments can inform network-level resilience analysis.