Search

Computer models are used to simulate pedestrian behaviour for safety at mass events. Previous research has indicated differences between physical crowds of co-present individuals, and psychological crowds who mobilise collective behaviour through a shared social identity. This thesis aimed to examine the assumptions models use about crowds, conduct two studies of crowd movement to ascertain the behavioural signatures of psychological crowds, and implement these into a theoretically-driven model of crowd behaviour. A systematic review of crowd modelling literature is presented which explores the assumptions about crowd behaviour being used in current models. This review demonstrates that models portray the crowd as either an identical mass with no inter-personal connections, unique individuals with no connections to others, or as small groups within a crowd. Thus, no models have incorporated the role of self-categorisation theory needed to simulate collective behaviour. The empirical research in this thesis aimed to determine the behavioural effects of self-categorisation on pedestrian movement. Findings from a first study illustrate that, in comparison to a physical crowd, perception of shared social identities in the psychological crowd motivated participants to maintain close proximity with ingroup members through regulation of their speed and distance walked. A second study showed that collective self-organisation seemed to be increased by the presence of an outgroup, causing ingroup members to tighten formation to avoid splitting up. Finally, a computer model is presented which implements the quantified behavioural effects of self-categorisation found in the behavioural studies. A self-categorisation parameter is introduced to simulate ingroup members self-organising to remain together. This is compared to a physical crowd simulation with group identities absent. The results demonstrate that the self-categorisation parameter provides more accurate simulation of psychological crowd behaviour. Thus, it is argued that models should implement self-categorisation into simulations of psychological crowds to increase safety at mass events.