In view of the difficulties in emergency evacuation caused by the increasing complexity of underground spaces in subway stations, and based on an extensive review of domestic and international literature, this paper first analyzes the evacuation characteristics and common features of different types of underground spaces, It examines evacuation scenarios, pedestrian behavior, and evacuation analysis methods in disaster contexts, and summarizes the evacuation behavior traits across station types. Secondly, the study delves into recent advancements in evacuation experiments and simulation technologies, covering physical and simulated experiments, including core algorithms, simulation processes, and inherent limitations of mainstream simulation platforms. It also highlights the impact of spatial features on simulation experiments and identifies current research hotspots. Finally, focusing on layout and path optimization for evacuation, the paper explores related methodologies and practical applications. On this basis, it points out that existing research is predominantly centered on theoretical modeling and quantitative analysis, with a lack of field validation that limits practical effectiveness. There is also a deficiency in interdisciplinary integration, especially concerning psychological and social dynamics of crowd behavior, which constrains the effectiveness of evacuation strategies in emergency practice. Therefore, the study suggests enhancing model realism through virtual reality, big data, and artificial intelligence technologies to improve simulation accuracy; and encourages interdisciplinary collaboration among psychology, sociology, urban planning, and safety engineering to comprehensively analyze crowd evacuation behaviors in different types of underground spaces in subway stations during emergencies, thereby providing a solid theoretical foundation and practical reference for future research.