Learning-based control for deployment and retrieval of a spinning tethered satellite formation system

This paper investigates the nonlinear dynamics and control of the deployment and retrieval for a spinning tethered satellite formation system via artificial intelligent method. A dynamic model of the spinning tethered formation system is developed to describe the attitude motions of the system, involving the relative rotations of the tethers to the central main satellite. Considering the system with symmetric and asymmetric configurations, a learning-based control strategy with low time cost is proposed to achieve the stable deployment and retrieval of tethers. In the strategy, a nonlinear model predictive control law accounting for the control constraints and nonlinear dynamics is developed to achieve the control goal. Based on a deep learning method, a dataset including control input and state output obtained offline is trained to form deep neural networks. An online feedback control of the system can be achieved by conducting real-time mapping from the system state to the control input using the neural networks. Finally, numerical simulations for deployment and retrieval of the system with different configurations are presented to demonstrate the computational efficiency and to validate the effectiveness of the control strategy.