Adaptive attitude takeover control of noncooperative spacecraft via Gaussian process online learning

The increasing number of space objects has posed new demands for spacecraft attitude takeover control methods capable of handling significant uncertainties. Cellular spacecraft have recently attracted considerable attention as a simple, cost-effective, and reusable solution for the capture of spacecraft. To address attitude takeover control of noncooperative targets, particularly those capable of generating unknown torques, a control approach integrating Gaussian process (GP)–based online learning and adaptive control is proposed. Sparse online Gaussian process (SOGP) is employed to realize online learning for system uncertainties, and the learned model is incorporated into an adaptive control framework for controller design. The control error of the proposed method is rigorously proven to be bounded in a probabilistic sense. For multiple cellular spacecraft system, a distributed control allocation problem with energy balance constraints is formulated and solved using the SDC-ADMM algorithm. Numerical simulation results demonstrate that the proposed control strategy achieves high-precision attitude control of the combined spacecraft, yields superior tracking performance compared with conventional control methods, and verifies the effectiveness of the proposed distributed control allocation scheme.