In this research, reinforcement learning techniques are combined with traditional direct shooting methods to generate optimal proximal spacecraft maneuvers. Open- and closed-loop controllers, parameterized by neural networks, are developed for terminally constrained, fuel-optimal relative motion trajectories using three different thrust models. Neurocontroller performance robustness to parametric uncertainty and bounded initial conditions is assessed. This research demonstrates that neurocontrollers offer a flexible and robust alternative approach to the solution of complex controls problems in the space domain and present a promising path forward to more capable, autonomous spacecraft.
展开▼
