Model-based approaches to emergency evacuation in buildings: a systematic literature review in the context of architectural practice and design

This review systematically analyses contemporary model-based approaches to emergency evacuation in buildings, drawing on 53 peer-reviewed studies published between 2015 and 2025. It emphasises the relevance of these approaches to architectural design workflows and performance-based safety engineering. The unique perspective adopted for this review focuses on the ways past research connects to, or is relevant for, architectural practice and design. Five modelling families are identified—Agent-Based Modelling (ABM), Social Force Models (SFM), Cellular Automata (CA), Multi-Agent Systems (MAS), and Society of Fire Protection Engineers (SFPE) guidelines and steering-type engineering models—and are examined using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and population, exposure, comparator, and outcome (PECO) frameworks. These models are mapped across hazard contexts, design dimensions, and integration strategies. The findings show that ABM and CA are the most versatile, supporting analyses of spatial layouts, exit schemes, and evacuation dynamics, while MAS introduces adaptive decision-making through machine learning (ML) and reinforcement learning. Gaps remain in modelling human-centred detailing, vulnerable populations, and hazards beyond fire, such as seismic or multi-hazard events. Integration with Building Information Modelling (BIM) and hazard simulators demonstrates a trajectory toward digital twins that embed evacuation safety within design processes, making these methods accessible to architectural practitioners. By consolidating these directions into a comparative framework, this review contributes to advancing computer-aided design and simulation in architecture, while strengthening the evidence base for fire safety and emergency engineering.