Abstract: With the rapid development of Low Earth Orbit (LEO) mega-constellation networks, satellite communication systems have become massive in scale with highly dynamic topologies, making intelligent networking a research focal point. This article surveys intelligent networking technologies for LEO mega-constellations, analyzes their networking characteristics and challenges, proposes an overall architecture for intelligent networking, and systematically reviews five key directions: service awareness and demand prediction, dynamic topology and mobility management, satellite–ground access strategies, inter-satellite routing strategies, and intelligent management of multidimensional resources. First, we review methods for service awareness and demand prediction, covering traditional single-satellite approaches, AI-based single-satellite approaches, and multi-satellite collaborative prediction. Second, we discuss dynamic topology and mobility management, tracing the evolution from traditional static management to dynamic schemes oriented to multidimensional resources. Third, for satellite–ground access strategies, we describe beam-scheduling methods for single satellites as well as multi-satellite cooperative beam scheduling. Fourth, we examine inter-satellite routing strategies in depth, surveying both centralized and distributed approaches. Finally, we discuss intelligent management of multidimensional resources from the perspectives of static and dynamic resource management. On this basis, we analyze the remaining challenges in intelligent networking for LEO mega-constellations, critically assess the effectiveness and limitations of current techniques, and outline future research directions.

Key words: LEO mega-constellations, intelligent networking technologies, traffic awareness and prediction, dynamic topology and mobility management, satellite-ground access, inter-satellite routing, management of multidimensional resources.