Neural model predictive control for spacecraft autonomous rendezvous and docking

This paper proposes a neural model predictive control (NMPC) strategy for autonomous rendezvous and docking (AR&D) with non-cooperative targets. The AR&D task is formulated as an optimal control problem, with the relative motion dynamics linearized using Hill’s equations to reduce online computational complexity. By leveraging neural networks, the proposed strategy significantly improves computational efficiency and facilitates real-time implementation. Simulation results demonstrate that the method effectively guides the chaser spacecraft to achieve precise position and velocity alignment with minimal fuel consumption. It also shows strong robustness and reliability in close-proximity docking scenarios, even under uncertainties arising from external disturbances and control deviations.