Natural Human Behaviour in Train Evacuation Simulations: Evidence from a Classroom-Based Experiment

Abstract

Passenger evacuation during train emergencies is often evaluated using procedural simulation models that prioritise regulatory compliance and idealised passenger movement. Although effective for technical assessment, these models frequently oversimplify human responses and fail to represent realistic passenger behaviour observed in actual emergency situations. As a result, evacuation time predictions may be overly optimistic. This study investigates the influence of incorporating natural passenger behaviours into evacuation simulations compared with conventional procedural approaches. An experimental, classroom-based evacuation simulation was conducted involving thirteen adult participants. Twelve evacuation scenarios were examined, consisting of seven procedural scenarios and five scenarios incorporating natural behaviours. The behavioural scenarios included hesitation, distraction, assisting elderly passengers, carrying luggage, and waiting for companions. Evacuation times were recorded for each trial, participant actions were analysed using video observations, and post-simulation questionnaires were used to capture decision-making and perception. The results indicate that scenarios incorporating natural behaviour produced greater variability and generally longer evacuation times than procedural simulations. The highest behavioural complexity resulted in a maximum evacuation time of 31.4 seconds, with an average increase of 8.58% across behavioural scenarios. Questionnaire findings confirmed that social interaction, situational awareness, and perceived responsibility influenced evacuation behaviour, with social dependency more frequently causing delays than facilitating movement. This study concludes that incorporating realistic passenger behaviour is essential for improving evacuation simulation accuracy and supporting more reliable railway safety evaluation and design.