Flood Evacuation and Resource Allocation Modeling Using Agent-Based Approaches and Advanced Pathfinding Algorithms

Abstract

Efficient flood evacuation requires fine-grained modeling that accounts for local infrastructure and population vulnerability, factors often obscured in regional-scale simulations. This study presents a neighborhood-scale, data-driven evacuation model for Franklin Park, Illinois, integrating Agent-Based Modeling (ABM) and a dual-phase A-Star pathfinding algorithm to simulate both resident behavior and emergency response under flood conditions.

Geospatial datasets, including FEMA flood zones, road networks, and building footprints, were processed using QGIS to construct an accurate representation of the community. Agents represent population groups and responders, and evacuation paths are optimized in two stages: firetrucks navigate via Euclidean-heuristic routing on the road graph, then responders continue on foot using Manhattan heuristics through flooded zones.

Simulation results show that a single emergency vehicle would require approximately 137 hours to evacuate all residents in a worst-case flooding scenario, whereas deploying ten vehicles reduces this to 12 hours, demonstrating the model’s capacity to evaluate resource allocation strategies. The system’s tiered architecture enables future extensions such as dynamic flood growth and multi-vehicle dispatching.

This work underscores the value of localized evacuation modeling in enhancing equity and efficiency. It aligns with findings that highlight the need to incorporate spatial analysis and individual agent behavior into evacuation planning to mitigate disparities (Tonn and Guikema, 2018). Additionally, prior simulation research has demonstrated that agent-based models are effective for comparing staged and simultaneous evacuation strategies under varied urban constraints (Chen and Zhan, 2014).