Phased evacuation: An optimisation model which takes into account the capacity drop phenomenon in pedestrian flows

Experiments have been organised for studying pedestrian behaviour in oversaturated conditions: the data showed clear empirical evidence of the capacity drop phenomenon. When high density conditions occur upstream a bottleneck, often the maximum bottleneck capacity (i.e. the maximum number of pedestrians that can flow through the bottleneck in a time interval) drops. This phenomenon could have relevant implications for fire safety: if, during a building evacuation, it is possible to control densities upstream the sections where flows from different rooms/floors merge, the flows through the bottlenecks can be equal to the maximum capacity and not to the dropped one and the building evacuation time could be shortened significantly. Alarm times and egress routes are important means for controlling densities and therefore for avoiding capacity drops during the evacuation process, especially in high-rise buildings. The paper presents a methodology for assessing to which extent the building evacuation time can be reduced acting on egress routes and alarm times. Specifically, the proposed methodology embeds an optimisation algorithm that explores the research space defined by all the possible route-schedule plans (where a route-schedule plan is defined by a set of alarm times and a set of egress routes) looking for the plan that minimises the building evacuation time. Given a route-schedule plan, the related building evacuation time is established by simulating the dynamic of the egress with a proposed new movement model. The proposed movement model is able to reproduce the capacity drop phenomenon emerged in the experimental work and therefore to assess the building evacuation time taking into account the current capacities of bottlenecks in the building. The route-schedule planning refers to static situations where all egress paths are available and the environment has good visibility and to high-rise buildings with a well-trained population. In this context, the route-schedule planning is chosen off-line and the building occupants could be trained on it.