Dynamic Road-Network Hybrid Optimization for Vehicle-Based Emergency Evacuation (Case Study: Qom)

Abstract

Urban evacuation is a typical complex-system problem in which heterogeneous agents, capacity-limited shelters, and time-varying road conditions interact on a large network. This study proposes a hybrid optimization framework for vehicle-based emergency evacuation that combines an exact baseline solved by DOcplex Python with metaheuristic algorithms, namely, GWO, GA, PSO, and SA, to preserve solution quality under both static and dynamic conditions. The model instantiates a multivehicle, multidepot routing setting with two evacuee classes (injured/noninjured), multiple vehicle types (bus-ambulance, bus, minibus), and dual shelter categories (hospital/residential), embedded on a GIS-derived road graph. On a real network from District 8, Qom (Iran), the exact layer delivers the best performance in normal traffic (e.g., mean evacuation time ≈ 79.4 min vs. 85.1–93.8 min for metaheuristics), under crisis conditions involving road closures and counterflow; however, the GWO- and GA-based approaches produce the most effective routes, with evacuation times of approximately 93.4–93.6 min, without requiring full reoptimization. This two-layer design treats evacuation as control on a changing network, balancing optimality (when feasible) with adaptability (when conditions shift), and demonstrates measurable gains in clearance time and operational resilience. The results highlight how hybrid computation and network-aware dynamics can bridge the gap between theoretically optimal routing and real-time decision support in disasters.