Emergency evacuation in crowded environments remains a major safety challenge due to complex crowd dynamics, congestion, and uncertainty during emergencies. Recent advances in autonomous robotic systems have introduced new opportunities for adaptive evacuation guidance and intelligent crowd management; however, research in this area remains fragmented across crowd modeling, robotics, human–robot interaction, and artificial intelligence. This paper presents a comprehensive survey of robot-assisted emergency evacuation systems, covering crowd behavior analysis, crowd modeling approaches, robotic guidance strategies, learning-based methods, simulation platforms, digital twins, and evaluation frameworks. A taxonomy of robot-assisted evacuation strategies is proposed, and the strengths, limitations, and applicability of existing approaches are critically analyzed. Furthermore, key research challenges and future directions are identified, including trust-aware guidance, safe reinforcement learning, scalable multi-robot coordination, simulation-to-reality transfer, and benchmark standardization. By providing a unified interdisciplinary perspective, this survey aims to serve as a reference framework for the development and evaluation of next-generation intelligent evacuation systems. This work was conducted at Arab International University (AIU), Damascus, Syria.Official website: https://www.aiu.edu.sy