In this paper, we review the current situation and development trend of multiple unmanned aerial vehicle (multi-UAV) collaboration in adversarial environments based on deep reinforcement learning(DRL). Firstly, the application of multi-UAV in military and civilian fields was introduced. Secondly, the basic principles of DRL are outlined, including multi-agent DRL methods and DRL applications, and their application achievements in fields such as gaming and robot control are discussed. Subsequently, this paper analyzed the key technologies of multi-UAV collaboration in adversarial environments based on DRL, including optimization of collaborative strategy, collaborative area search, and collaborative communication. The case study demonstrates the effectiveness of DRL in improving the decision-making ability, adaptability and robustness of multi-UAV systems. At the same time, the challenges faced by current research are pointed out, such as computational resource constraints, complex task processing and illegal application risks. Finally, the article looks forward to the future development trend. It emphasizes the importance of algorithm optimization, model innovation, adaptive and robust improvement, and cross-domain applications. This paper aims to provide a comprehensive summary and a reference for future research directions for relevant researchers.