Realistic evacuation simulation through micro and macro scale agent-based modelling including demographics, agent patience and evacuation route capacities

Abstract

Disasters affect millions of people annually, causing large social impacts, and detrimental economic impacts. Emergency professionals recurrently tackle these impacts, therefore they require assessment methods to understand potential consequences and enable the delivery of resilient resolutions. One method of achieving this is through numerical modelling, specifically agent-based modelling. However, current models simulating human behaviours and movement are bespoke in nature and non-transferable. It has also been found that current modelling tools have either focused on the microscale (e.g. individual confined spaces) or macroscale (e.g. city scale), without considering how the two scales may be interlinked. Further to this, the inclusion of human behaviour has been over-simplified and generic, lacking the inclusion of unique populations with varied characteristics. The aim of this research is to develop a modelling framework, utilising agent-based modelling, to form a more robust representation of human behaviour within an enhanced evacuation model environment. This will allow emergency planners to be better prepared, reduce the interruption after an event (thereby reducing social and economic impacts) and potentially reduce the mitigation required beforehand. The individual agents within the framework capture a range of robust human behaviour indicators (e.g. walking speed, obedience, and patience), allowing the accurate replication of an emergence scenario response. Initially, the research focused on creating a macroscale evacuation model for a test city, to assess whether the inclusion of varied population characteristics and groups of people affected evacuation time. The varied characteristics included a range of ages, gender, and mobility in the form of walking speed. It was then possible to compare this with the parameters of existing evacuation models. This research has found that by enhancing the representation of human behaviour within a model environment more accurate predictions of evacuation time can be produced. To produce more robust human behaviour, models must include a range of population characteristics (such as age, gender and mobility), the grouping of agents and walking speed ratio. When all the variables are included in the model, there is an average increase of 70% in the time to evacuate Newcastle city centre. Even with less variables, i.e. only considering population characteristics, there has been an average increase of 45% in the time to evacuate Newcastle city centre compared with existing models. To further examine human behaviour and the more intricate and detailed behaviours such as patience, a microscale model was created to consider the capacity of the pathways and to introduce congestion. The two microscale models were created of a pavement and a crossroads, ii to replicate people passing and waiting behind slower people, whilst still including the varied population characteristics. When capacity is captured at the microscale, there is an average 61% increase in the time to exit the pavement and when on a crossroads there is an average 87% increase in the time to exit compared to 1.34m/s (3mph) models. Overall, this research has found that there is a need to provide more robust representation of human behaviour characteristics within evacuation models. This must be carried out not only at the macroscale in terms of enhancing population demographics but also at the microscale by capturing intricate behaviours such as taking over and giving way. Without an ability to exhibit these characteristics evacuation simulations cannot effectively capture human behaviour and therefore produce robust simulation times. The inclusion of more representative human behaviour in simulations and the continual need to improve provides the opportunity to reduce the likelihood of increased fatalities and injuries caused by those unable to evacuate in time due to current underestimations. The improvement of computational simulation of evacuations alongside existing simulation techniques allows emergency professionals to plan and prepare better for a range of events to protect global communities.