A Prescriptive Simulation Framework with Realistic Behavioural Modelling for Emergency Evacuations

Abstract

Emergency and crisis simulations play a pivotal role in equipping authorities worldwide with the necessary tools to minimize the impact of catastrophic events. Various studies have explored the integration of intelligence into Multi-Agent Systems (MAS) for crisis simulation. This involves incorporating psychological behaviours from the social sciences and utilizing data-driven machine learning models with predictive capabilities. A recent advancement in behavioural modelling is the Conscious Movement Model (CMM), designed to modulate an agent's movement patterns dynamically as the situation unfolds. Complementing this, the model incorporates a Conscious Movement Memory-Attention (CMMA) mechanism, enabling learnability through training on pedestrian trajectories extracted from video data. The CMMA facilitates mapping a pedestrian's attention to their surroundings and understanding how their past decisions influence their subsequent actions. This study proposes an efficient framework that integrates the trained CMM into a simulation model specifically tailored for emergency evacuations, ensuring realistic outcomes. The resulting simulation framework automates strategy management and planning for diverse emergency evacuation scenarios. A single-objective method is presented for generating prescriptive analytics, offering effective strategy options based on predefined operational rules. To validate the framework's efficacy, a case study of a theatre evacuation is conducted. In essence, this research establishes a robust simulation framework for crisis management, with a particular emphasis on modelling pedestrians during emergency evacuations. The framework generates prescriptive analytics to aid authorities in executing rescue and evacuation operations effectively.