Spacecraft Attitude Takeover Control by Multiple Microsatellites Using Differential Game
Baolin Wu, Keyu Chen, Danwei Wang, Yuxiang Sun
Abstract
This article proposes a differential game-based control scheme to address an attitude takeover control problem via microsatellites attached to the surface of target spacecraft. First, the attitude dynamics of combined spacecraft is reformulated as a general form so as to apply the reinforcement learning framework. Then, quadratic and Arctanh-type performance indices are designed in two cases of free input and input saturation, respectively. Accordingly, optimal control policy of each microsatellite is obtained and is dependent on value functions, which are solutions of a set of Hamilton–Jacobi–Bellman (HJB) equations. Single layer neural networks are employed to approximate value functions by policy iteration and the weights vectors are updated with the help of the concurrent learning algorithm so that the persistent excitation condition of control errors is loosened. Moreover, the necessity of interaction among microsatellites is eliminated by using a tracking differentiator technique to estimate angular accelerations which are not available directly through onboard devices. Stability of the closed-loop system is guaranteed by the Lyapunov method. Three cases of simulation are carried out to demonstrate the robustness and optimality of the proposed control scheme and to validate the effectiveness of the controller in the presence of actuators saturation.