Resilience Optimization of Transportation Network via Deep Reinforcement Learning: An Emergency Resource Allocation Framework

Urban transportation networks often face disruptions from emergencies such as natural disasters or service failures. Enhancing their resilience is essential for maintaining system reliability. This paper proposes an emergency resource allocation framework based on deep reinforcement learning to optimize the recovery of bus stops in disrupted networks. Node importance is quantified using betweenness centrality, and a Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is applied to continuously allocate resources. Techniques such as validity masking and power normalization are incorporated to ensure stable learning and efficient policy generation. Comparative experiments with SAC, PPO, GA, and PSO demonstrate that the proposed method outperforms alternatives in terms of recovery speed, resilience metrics, and robustness. The results validate the effectiveness of integrating structural analysis with reinforcement learning for building resilient urban transportation systems.