Efficient evacuation in a multi-exit environment: an agent-based decision support model

A majority of research work carried out in crowd evacuation rely on simulation due to non-availability of real and realistic trial data. In this paper, an agent-based simulation study of an evacuating crowd is presented. The model is based on the microscopic behavioural rules formulated through small-scale empirical evidence in conjunction with crowd behavioural theories. In particular, the study focuses on the possibility of efficient evacuation from the environment with limited perceptions. Extending Moore's neighbourhood model, local congestion avoidance mechanism capable of detecting the relative displacement and orientation of the all the individuals in its neighbourhood is considered. Other strategies based on exit capacity and exit population are also modelled and tested. A probabilistic exit selection strategy is also designed that considers a sensitivity of an exit as a deciding factor. The simulation results show that the enhanced exit selection strategies make the proposed system more robust and increase the evacuation efficiency substantially.