Unmanned aerial vehicles (UAVs) have become essential in disaster management, aiding in search and rescue, damage assessment, supply delivery, and communication restoration. However, traditional UAV control methods struggle with real-time decision-making and adaptability in dynamic disaster scenarios. Deep reinforcement learning (DRL) offers a promising solution by enabling UAVs to learn optimal strategies in uncertain and complex environments autonomously. This paper explores the advantages and limitations of deploying UAVs in disaster scenarios, outlines the core principles and applications of DRL in autonomous systems, and investigates research efforts that combine these technologies to address specific disaster response tasks. Finally, it synthesizes the gathered information to identify current issues and potential future directions for research.