Record-driven Simulated Flight Training Mission Recommendation

In consumer-grade flight simulation, there is no convenient personalized solution for users' flight training that can reasonably generate new mission scenes based on their training performance. This paper introduces a novel task focused on generating flight mission scenes for users guided by their training records. We present a comprehensive framework capable of autonomously producing mission scenes. Integrating a Mission Type Recommendation and a Scene Planning module allows for the analysis of users' proficiency and the output of scene parameters tailored to different flight simulation platforms using large language models (LLMs). Our user study to validate the effectiveness of the method on difficulty rationality, scene authenticity and parameter effectiveness.