Attitude control for dumbbell-shaped spacecraft subject to parameter uncertainty via fully-actuated system approach and deep reinforcement learning

This paper studies the attitude control problem for a dumbbell-shaped flexible spacecraft with parameter uncertainty. To this end, based on the dynamic formulation and transformation of dumbbell-shaped spacecraft, a composite control framework is synthesised by integrating feedback and observer. A parameterised feedback controller is constructed via fully-actuated system approach and a disturbance observer based on deep reinforcement learning is proposed to deal with parameter uncertainty. The proposed method offers an intelligent and effective solution from a practical viewpoint. Finally, numerical simulations validate the superiority of the method with comparison.