Resilient AI Architectures for Disaster-Responsive Transportation Systems: Dynamic Routing and Coordination During Natural and Human-Made Crises

Abstract

The increasing frequency and intensity of natural and human-made disasters pose severe challenges to global transportation systems, often resulting in mobility paralysis, delayed emergency response, and significant socioeconomic disruption. Traditional Intelligent Transportation Systems (ITS) lack the adaptability and resilience required to operate effectively under rapidly changing crisis conditions. This study proposes a resilient Artificial Intelligence (AI) architecture integrating deep reinforcement learning (DRL) and multi-agent coordination to ensure continuous mobility during floods, earthquakes, and mass evacuations. Using a simulation-based experimental design in SUMO with Python integration, the framework was tested across multiple disaster scenarios and compared with conventional ITS models. The results revealed a 45–55% reduction in route recovery time, over 89% sustained network throughput, and a resilience index exceeding 0.85, demonstrating the system’s ability to autonomously adapt and stabilize traffic flows under severe disruptions. Statistical validation confirmed significant improvements in coordination efficiency and adaptability. The findings highlight the transformative potential of AI-driven architectures for disaster-responsive transportation, enabling decentralized decision-making and faster system recovery. The proposed framework contributes to the broader goals of smart-city resilience and sustainable urban mobility, offering a viable pathway for real-world implementation in next-generation disaster management systems.