Hurricane-resilient city: Markov decision process-based problem formulation and machine learning-based solution

Recent years have witnessed rapid advances in assessing interdependent infrastructure performance under multi-hazard hurricane events and optimizing actions to enhance community resilience. This study focuses on improving the resilience of coastal cities through systematic short-term actions. Building on a review of state-of-the-art research, the problem is formulated as a Markov decision process (MDP)-based optimization, requiring efficient simulation of complex hurricane–city interactions and effective optimization of sequential actions. To address these challenges, a machine learning-based methodology is developed, combining supervised learning for fast surrogate modeling with reinforcement learning for optimal decision-making. Following a divide-and-conquer strategy, global action optimization is decomposed into local optimizations to reduce computational complexity. The hurricane life cycle is further divided into three phases: pre-hurricane preparedness, during-hurricane response, and post-hurricane recovery. Case studies using a virtual city demonstrate the potential of machine learning to enhance resilience and identify knowledge gaps for future research.