Crowd dynamics in emergency situations

Abstract

Safety is a very big issue when designing new buildings, structures and spaces. In cases of emergency, many people lose their lives due to difficulties in reaching suitable exits in time or due to the chaos caused by other people with the same intentions. Following the maxim that 'prevention is better than cure', engineers try their utmost to build such spaces to provide a safer environment. Although various guidelines exist for the development of buildings, they cannot be considered as a panacea since different situations may require some alterations. Simulations can provide clearer insight for such requirements and are currently a widely used tool to model crowded hazardous situations so that casualties are avoided or minimised. The aim of this project is to investigate and develop possible solutions to model crowds in emergency situations. By modelling a particular environment, a hazard, a crowd and its behaviour, an approximate outcome of how a crowd will react to an emergency egress scenario will be simulated. The intention of this simulation is to provide cues on how the geometric configuration of the building affects the flow and egress rates of a crowd. Previous research has shown that the modelling of crowds in emergency situations is not an easy feat to achieve since, in reality, people think and react in many different ways; predicting chaotic behaviour is difficult. A framework was developed comprising of a two-dimensional building editor, a plug-in behavioural model subsystem and the encompassing simulator. The editor is presented using an intuitive user interface, allowing a user to create traversable, exit and hazardous areas. Fire is used as an example of a hazard and is simulated using a cellular automaton model. A crowd behaviour model based upon an agent-based, distributed behavioural force model was attempted and provided with the aforementioned framework. The behaviour model is evaluated based on the current localised perception of an individual and in tum provide the necessary instructions to the underlying motor control system. The model exhibits individuals searching for the shortest exits and trying to reach their intended destination with minimal collisions whilst avoiding hazardous areas. The proposed model exhibits an emergent, self organising system and manages to simulate a crowd with the intended behaviour. This was a result based upon a series of tests which were composed in order to study the dynamics of the simulated crowd with regards to different scenarios. A comparison to another popular crowd behaviour model (Helbing's Social Forces model) is performed in order to analyse the inconsistencies with our proposed model and elicit possible enhancements. Due to the chaotic nature of the problem at hand, the crowd models provided are by no means the direct reflection of reality, but the rudimentary behaviours used are suitable for analysing emergency egress scenarios and can possibly aid by providing a better understanding to cater for similar future situations.