Abstract: The increasing frequency of extreme disasters driven by global climate change has underscored cascading-failure risks caused by complex interdependencies among urban infrastructure systems, making them a pivotal challenge for disaster-resilient urban planning. Existing studies remain constrained by three limitations. The identification of infrastructure interdependencies is limited by data availability and methodological tools, producing a narrow set of recognized categories. The refined identification of critical facilities often overlooks social systems and residents' behavior, resulting in incomplete simulation frameworks. Resilience-improvement strategies still focus mainly on single-system responses and emergency countermeasures, while collaborative multi-system planning remains underdeveloped. Against the background of rapid progress in artificial intelligence, this paper proposes research directions and technical pathways: using multisource heterogeneous data and large language models to improve the granularity and precision of interdependency identification; developing facility-society-population coupled cyber-physical simulation models to characterize urban spatial vulnerability; and formulating intelligent, multi-system collaborative recovery strategies to accelerate the restoration of urban service capacity. The study provides theoretical guidance and feasible technical approaches for key issues in disaster-resilient planning.