Learning to Update Engine Models with Deep Reinforcement Learning

Model updating can reduce the reality gap between the digital space and physical space, which is essential for accurately building digital models of aviation equipment. However, with data requirements and the complexity of application environments continue to increase, traditional update techniques show great limitations. The main reason is that traditional update technologies are limited by computational efficiency, resulting in the inability to balance accuracy and real-time. In this work, we propose a model update method based on reinforcement learning to infer model parameters. The proposed method does not require any ground truth parameters. Experimental results on the twin-rotor turbofan engine model verify the superiority of the proposed method compared with other state-of-the-art methods.