A Deep Learning Predictor-Proportional Guidance Corrector Method for Rocket Deceleration Guidance

During the aerodynamic deceleration flight of reusable rockets, large uncertainties exist and the environment is very complicated, resulting in large deviations in terminal velocity and position. Therefore, it is often very difficult to satisfy the handover condition, especially for the terminal velocity, as they are very sensitive to uncertainties. To address this issue, a new predictor-corrector guidance method for rocket deceleration is developed in this paper. With the proposed method, the bias proportional guidance is employed to simultaneously control the terminal position and attitude angle, while the predictor-corrector guidance is to control the terminal velocity by correcting the guidance coefficient of BPN. The correction command for terminal velocity is derived based on the terminal velocity deviation and partial derivative. Moreover, a deep learning method is proposed for terminal velocity prediction to improve the prediction efficiency while ensuring accuracy. The effectiveness and advantages of the proposed method are demonstrated by numerical simulations.