A composite control method for horizontal trajectory tracking of stratospheric airships based on reinforcement learning

High-precision trajectory tracking of stratospheric airships under complex wind disturbances and model uncertainties presents a challenging problem. To address the limitations of conventional control methods in resisting disturbances under uncertain conditions, this paper proposes an adaptive composite control method that integrates backstepping sliding mode control (BSMC) with reinforcement learning (RL). First, a BSMC controller is designed based on Lyapunov stability theory to provide stability guarantees and achieve baseline tracking performance. Meanwhile, an RL-based intelligent compensator is introduced to actively suppress complex disturbances through online learning. Innovatively, an adaptive weight allocation mechanism using a sigmoid function of the tracking error is designed, enabling smooth switching and cooperative action between the BSMC and RL controllers: when the error is small, only the BSMC is activated to ensure smooth operation; when the error increases, or strong disturbances occur, the RL compensator is triggered to accelerate convergence. Simulation results demonstrate that, compared to backstepping control and BSMC, the proposed composite controller significantly improves tracking accuracy under unknown disturbances and exhibits stronger robustness and environmental adaptability.