Modelling of behaviours in response to terrorist activity

Abstract

In this paper we present an approach to the modelling of human interaction in complex environments and its application to a security related scenario; the evacuation of a railway station subsequent to the detonation of an improvised explosive device. The intent of the experiments reported in this paper is to investigate the application of our existing software capabilities to the proof-of-concept scenario described above. Our simulation framework, Simulacron, allows the development of multiple interacting modules which address matters such as motivation, scheduling and movement, controlled both by internal goals and external influences. The ability to integrate this interaction modelling with our existing, rich palette of capabilities makes this approach to simulation building of particular interest. In this paper, we pay particular attention to the motivation module, which allows the coupling of internal characteristics with external conditions to control behaviour. Sydney's Central Station is represented in this scenario with a directed graph of some 200 locations. Through this, five thousand commuters move over the course of two simulated hours at a temporal granularity of 5 seconds. Typical run time was on the order of one hour on a consumer desktop using a single thread. Commuter behaviours, which have been selected based on actual, observed behaviours in similar circumstances, are represented by a number of states, driven by transition rules using a set of hypothetical characteristics. We report the results of several variations to the base scenario where we modify commuter and assailant characteristics. Additionally, we report on a set of simulations in which simple crowd management strategies are implemented to investigate their potential impact on the evacuation process. We conclude that the results broadly agree with expected behaviours and justify further development. This would include the incorporation of psychologically-grounded characteristics and the extension of the work to a full validation scenario. We foreshadow a range of future initiatives including a study to determine the interaction between evasion resulting from perceived assailant threat and fatality rates. The work to date has provided us with a simulation environment into which we can now introduce more experimentally derived parameter values, covering not only internal states but also allowing the introduction of passenger schedules and the incorporation of incoming and outgoing trains. The incorporation of additional features, such as: more realistic evasion, security force interdiction, multiple assailants including decoys, caring for the injured, families, biological imperatives, such as hunger, and disorientation during evacuations, is planned for the future.