Reinforcement Learning Based Attitude Tracking Control of Spacecraft with Actuator Saturation and Inertial Uncertainty

Abstract

This paper investigates an attitude tracking control problem of spacecraft subject to actuator saturation and inertial uncertainties. Aiming at alleviating the negative effects of actuator saturation and parameter uncertainties, an adaptive prescribed performance attitude control scheme is proposed via exploring the reinforcement learning policy. By constructing a critic network and an action network, the approximations of the optimal objective function and the optimal control gains of the spacecraft attitude tracking error system are realized. A group of numerical examples is organized to validate the effectiveness and robustness of the proposed attitude control scheme with respect to actuator saturation and inertia uncertainties.