Physics-based model and data dual-driven approaches for predictive evacuation
Yuxin Zhang, Zhiguo Yan, Hehua Zhu, Pingbo Tang
Abstract
Physics-based models or data-driven methodologies can help acquire the evacuation process for optimizing evacuation and rescue plans. However, neither of these methodologies can predict the process rapidly and precisely. Physics-based models rely on physical rules of human behavior but are computationally expensive. Data-driven approaches need a significant amount of diverse data but lack an understanding of underlying human behavior in spinning emergencies. This short communication aims to initiate systematic discussions about a physics-based model and data dual-driven approach, combining both approaches’ strengths. This combined approach puts forward iterative updating loops, using physics-based models to identify evacuation stages, capture operational mechanisms, and act as rational boundaries while using data-driven methods to process each stage’s results rapidly. The authors synthesize a roadmap highlighting bottlenecks and potential research directions for achieving such a combined approach, calling for attention and collaboration in predictive evacuation for disaster response.