Event-triggered trajectory-tracking guidance for reusable launch vehicle based on neural adaptive dynamic programming

To improve the trajectory-tracking guidance performance for reusable launch vehicle under various uncertainties and distortions, an event-triggered (ET) guidance method based on neural adaptive dynamic programming (ADP) is proposed. Firstly, the reference trajectory and corresponding steady-state control are generated optimally offline based on Gauss pseudo-spectral method. Secondly, the approximate optimal feedback controller based on single-critic ADP is designed. The great adaptation capacity inheriting from reinforcement learning technique ensures the tracking errors to converge to zero, and yet no offline dataset or pre-training is required. Event-triggered mechanism is introduced to reduce online training computation and save data transmission resource. Event-triggered condition is designed and the asymptotic stability of the event-triggered guidance system is proved. Comprehensive simulations are conducted and results validate the effectiveness of the feedback controller based on ADP and the significantly improved efficiency of ET mechanism. Besides, the improved performance of the proposed guidance method over traditional method of linear quadratic regulator has also been verified through simulations.