Abstract: Urban computing, grounded in the collaborative modeling and integrated analysis of multi-source heterogeneous data, has become a critical technical foundation for fine-grained urban governance and intelligent decision-making in smart cities. With the rapid growth of urban time-series data, spatial data, and knowledge graphs, several critical challenges have emerged, including unified modeling of multi-scale spatio-temporal features, effective alignment of heterogeneous modalities, and the collaborative representation of numerical and semantic information. This paper systematically reviews representative research progress in urban time-series data fusion, urban spatial data fusion, and urban knowledge graph fusion. Existing methods are categorized into some typical paradigms: data-level fusion, feature-level fusion, model-level fusion, and knowledge-driven fusion. A comparative analysis is conducted to evaluate these approaches in terms of spatio-temporal modeling, multimodal collaboration, and semantic expression capabilities. Finally, key challenges and future research directions for deep collaborative fusion of urban multi-source data are discussed.