Abstract: Close-range air combat has the characteristics of multiple combat elements, rapid situational changes, and tense combat atmosphere, it’s decision-making method is a hot research topic in the field of artificial intelligence. At present, research on close range air combat algorithms is mostly conducted in simplified low precision scenarios or existing simulation systems, due to the complexity of practical problems and limitations in simulation effectiveness, the decision models for air combat are mostly simplified, which reduces the reference value of research results. In response to this issue, a visual air combat platform that meets research requirements was built based on Unity3D, and a set of aircraft maneuvering actions was designed. Based on the characteristics of the enemy friendly situation during close-range air combat, situation evaluation functions and reward functions were defined. On this basis, a one-on-one close-range air combat decision-making framework based on proximal policy optimization algorithm was constructed. The experimental results show that the decision model can drive the intelligent agent to make flexible maneuvering decisions based on the battlefield situation, and has strong autonomous decision-making ability, which verifies the effectiveness of the method.