关键词: 有人-无人协同空战, 空战运行范式, 决策技术架构, 深度强化学习, 人机交互

Abstract: Amid the global wave of military transformation, manned–unmanned cooperative air combat has become a new focus of major-power military competition owing to its disruptive warfighting potential. However, current research still faces two primary challenges. First, a systematic analysis of the operational paradigm for manned–unmanned cooperative air combat under typical aerial confrontation patterns is lacking, making it difficult to effectively guide key technology development. Second, relevant technical research lacks an integrated architecture and remains largely fragmented, thereby limiting overall combat effectiveness. To address the first challenge, this study analyzes the role positioning and coordination logic of manned–unmanned cooperative air combat under typical aerial confrontation patterns, and extracts the information interaction relationships and decision-making technology requirements therein. To address the second challenge, this study constructs a three?tier intelligent decision?making technology architecture for manned–unmanned cooperative air combat, consisting of a demand?pull layer, a technology research layer, and a simulation?validation layer, and further explores feasible implementation approaches for five key technical links within the architecture. On this basis, experimental simulations are conducted using specific manned–unmanned cooperative air combat micro?task scenarios. The technical effects are demonstrated at different levels and stages, visually presenting the application method and overall effectiveness of the proposed architecture. Through these efforts, this study aims to provide a clear research route and a systematic methodological framework for both academia and engineering, and to promote the large?scale application of manned–unmanned cooperative air combat in diverse and complex missions.

Key words: manned–unmanned cooperative air combat, air combat operational paradigm, decision-making technology architecture, deep reinforcement learning, human–machine interaction