Inclusive crowd evacuation modeling under heterogeneous mobility constraints

Emergency evacuations in built environments pose significant challenges for individuals with disabilities, yet traditional simulation models often fail to account for heterogeneous mobility needs. While considerable advances have been made in pedestrian dynamic modeling, a critical gap persists in the realistic incorporation of disability-specific movement limitations and environmental barriers. This paper presents an inclusive evacuation simulation framework based on an extended social force model, explicitly integrating wheelchair users and visually impaired individuals. The model modifies agent parameters such as desired speed, relaxation time, body size, and barrier navigation capability to reflect empirical observations. Key enhancements include a probabilistic falling mechanism under high crowd pressure and dynamic interaction with environmental obstacles. A single-room evacuation scenario involving 50 agents, including 20% disabled individuals, was simulated using this framework. Results demonstrated that the presence of disabled individuals increased total evacuation time by approximately 50% compared to an all-able-bodied crowd, led to a 40% reduction in peak evacuation throughput after crowd falls, and caused arching, clogging, and faster-is-slower effects to intensify. Two fall incidents occurred within the first 4 s, resulting in partial door blockage and additional delays. Heatmaps revealed localized congestion zones induced by mobility impairments, and kinetic energy analysis illustrated significant dissipation due to frictional interactions at the exit. The findings underscore the necessity of inclusive modeling to identify critical vulnerabilities in evacuation plans and highlight the importance of design interventions such as wider doorways, alternative accessible exits, and controlled evacuation flow for heterogeneous crowds. This work offers a robust foundation for performance-based inclusive design and supports future extensions into multi-level structures, dynamic assistance modeling, and optimization-based evacuation planning.