Multilayer-perceptron-based on-orbit learning attitude takeover control for non-cooperative spacecraft

This paper investigates the attitude takeover control for non-cooperative spacecraft that remain capable of actively generating unknown torques. A neural network-based control framework is proposed, in which a multilayer perceptron (MLP) network is employed to perform on-orbit learning and real-time prediction of the unknown torques generated by the target spacecraft. The predicted torques are subsequently incorporated into the control law as compensation terms, thereby mitigating the significant disturbances caused by the target spacecraft’s output torques in attitude takeover control system. Numerical simulation results demonstrate that the proposed method is capable of driving the non-cooperative target spacecraft with torque generation capability to a stable desired attitude, achieving attitude takeover control of the non-cooperative spacecraft.