Deep Learning for Cross-Domain Data Fusion in Urban Computing: Taxonomy, Advances, and Outlook

As cities continue to burgeon, Urban Computing emerges as a pivotaldiscipline for sustainable development by harnessing the power of cross-domaindata fusion from diverse sources (e.g., geographical, traffic, social media,and environmental data) and modalities (e.g., spatio-temporal, visual, andtextual modalities). Recently, we are witnessing a rising trend that utilizesvarious deep-learning methods to facilitate cross-domain data fusion in smartcities. To this end, we propose the first survey that systematically reviewsthe latest advancements in deep learning-based data fusion methods tailored forurban computing. Specifically, we first delve into data perspective tocomprehend the role of each modality and data source. Secondly, we classify themethodology into four primary categories: feature-based, alignment-based,contrast-based, and generation-based fusion methods. Thirdly, we furthercategorize multi-modal urban applications into seven types: urban planning,transportation, economy, public safety, society, environment, and energy.Compared with previous surveys, we focus more on the synergy of deep learningmethods with urban computing applications. Furthermore, we shed light on theinterplay between Large Language Models (LLMs) and urban computing, postulatingfuture research directions that could revolutionize the field. We firmlybelieve that the taxonomy, progress, and prospects delineated in our surveystand poised to significantly enrich the research community. The summary of thecomprehensive and up-to-date paper list can be found athttps://github.com/yoshall/Awesome-Multimodal-Urban-Computing.