Spacecraft Proximity Operations under Motion and Input Constraints: A Learning-Based Robust Optimal Control Approach

This article proposes a learning-based robust control scheme for spacecraft proximity operations under motion constraints (i.e., approaching path and sensor field-of-view constraints), input saturation, and external disturbances. To enhance the robustness of the learning algorithm, a disturbance observer with finite-time convergence is first designed to provide accurate model information for online learning. By virtue of the barrier functions and hyperbolic tangent functions, a performance index is developed, which incorporates both motion and input constraints into the framework of adaptive dynamic programming. Then, following the actor-critic structure, an approximate optimal saturated control policy is obtained using two neural networks (NNs), wherein the weights of the NNs are updated online. It is shown that the derived controller can guarantee the boundedness of system states and network weight estimation errors, while ensuring the satisfaction of motion and input constraints despite the presence of external disturbances. Finally, numerical simulations are carried out for spacecraft proximity operations with a tumbling target to verify the effectiveness of our proposed method.