Reinforcement learning-based finite time control for the asymmetric underactuated tethered spacecraft with disturbances.

This article addresses an attitude stabilization control problem for the asymmetric underactuated tethered spacecraft subject to external disturbances, and a reinforcement learning(RL)-based finite time control scheme is proposed to enhance the control performance and energy efficiency of the closed-loop system. Firstly, the error dynamics of the underactuated tethered system in the presence of external disturbances is built based on the Lagrange's modeling technique. Then, a RL-based control algorithm is implemented by a radial basis function (RBF) neural network (NN), in which the actor–critic networks are developed to obtain the optimal performance index function and the optimal controller. According to the Lyapunov theorem, semi-global finite-time stability of all the closed-loop signals is achieved through rigorous mathematical analysis, and tracking errors can be ensured to an arbitrarily small neighborhood of the origin in a finite time. Finally, comparative simulation results with hierarchical sliding mode controller are presented to demonstrate the viability of the proposed strategy. [Display omitted] • The proposed algorithm does not rely on the harsh symmetry condition of mass matrix. • Actor-critic NNs are used to obtain the performance index function and controller. • A novel RL-based scheme is designed to ensure semi-global finite-time convergence. [ABSTRACT FROM AUTHOR]