A Longitudinal Control Method for UAV Carrier Landing Based on Deep Q-Network Method

This paper focused on the issues of strong ground effects and ship wake interference et al faced by high aspect ratio unmanned aerial vehicles (UAV) during carrier landing. A network was established using the Deep Q-Network (DQN) algorithm to intelligently optimize the longitudinal control parameters of UAV. This method could improve the accuracy of UAV carrier landing control and reduce the longitudinal error of carrier landing. Using the mature inner and outer loop series PID control law in industrial applications as the initial control method, this method applied the training results of the DQN algorithm for parameter optimization. In addition, the DQN algorithm was optimized during use. By establishing a simulation model and conducting longitudinal carrier landing process simulation experiments, it is shown that compared with the typical PID control law method with internal and external loops connected in series, the proposed UAV carrier landing longitudinal control system optimized based on DQN method has significantly improved carrier landing accuracy and stability of carrier landing longitudinal error under different environment disturbances.