Terrorist attacks have become both more frequent and deadlier in recent decades. Knife attacks by terrorists in particular are both high impact and relatively frequent. Understanding how individuals move when confronted by an attacker can save lives by informing the planning of transportation hubs, entertainment venues and other spaces where large numbers of people congregate. Using pedestrian dynamics to predict responses to such attacks has the potential to further improve the survival of those involved by providing insights to emergency service responders when a terrorist situation has occurred in a crowded venue. However, given the lack of appropriate data, it is currently impossible to accurately predict pedestrian movement responses to terrorist attacks. This paper describes a practical study that developed a methodology and implemented a set of experiments examining responses to knife-based attacks by an unexpected and hostile individual. The experiments used financial incentives and an ‘aggressor’ to recreate a knife-based terrorist attack in an ethically viable setting. The participants of this study (n = 80) were tracked using ultra wide-band sensors to provide temporal and spatial positional data in relation to the attacker and to each other. Participants subsequently completed a questionnaire to report their psychological response during the experiment, thus allowing psychological perception and movement responses to be compared.

The results show that participants were stressed and reported fear during the experiment, indicating that the experimental approach is a potentially valid proxy for a real-life attack. The analysis highlights that participants’ decisions to stay or run away were strongly influenced by the actions of other nearby participants, while participants’ continuous movement response (movement speed) was highly dependent on relative position to the aggressor. Participant demographics were a factor with female participants less likely to move and slower moving, while older participants were also slower moving. From these results we have been able to construct a model of how pedestrians may react when faced with a sudden attack. We discuss our findings and their potential to contribute to emergency planning and response, finally we comment on opportunities for further study.