An attitude control scheme for ejection seats based on BP neural network

In order to improve the life-saving performance of the fourth-generation ejection seat under unfavorable conditions at low altitudes, we developed a dynamic model for free flight of the ejection seat and a dual-channel attitude control model for the ejection seat with low-to-medium ejection speeds. An attitude control scheme based on the above models using the backpropagation (BP) neural network PID controller was further proposed to control the pitching and rolling attitudes of the ejection seat. The proposed scheme uses a genetic algorithm to design a parameter optimization model for the ejection seat attitude control, which generates a set of optimal PID control parameters corresponding to different ejection states. Based on this dataset, a BP neural network model was developed and trained to directly predict the PID parameters for the attitude control of the ejection seat with low-to-medium ejection speeds. Finally, the PID controller outputs two sets of rudder deflection angles of the gas rudder thrust-vector nozzle. The lift and drag forces generated by the deflection of the gas rudder act on the ejection seat’s dynamic model, thereby adjusting the attitude of the ejection seat during the free-flight phase. The control scheme, validated by MATLAB/Simulink with different initial ejecting speeds, pitch angles, and roll angles, successfully adjusted the pitching and rolling attitudes of the ejection seat to normal values during the free flight within the rocket package’s working time, demonstrating the promising application prospects of the proposed attitude control scheme for ejection seats with low-to-medium ejection speeds.