Adaptive dynamic programming-based optimal control for chained tethered satellite formation reconfiguration

This paper addresses the problem of optimal reconfiguration control for a chained tethered satellite formation with full-state constraints, model uncertainties, limited actuator capabilities, and external disturbances. An optimal control approach based on an observer and adaptive dynamic programming technique is proposed. First, a neural network-based disturbance observer is used to estimate the model uncertainties and external disturbances. In the backstepping framework, a specially designed cost function is introduced to handle the full-state constraints and limited actuator capacities. Using the adaptive dynamic programming technique, an actor neural network is employed to approximate the optimal control law, while a critic neural network is used to approximate the cost function. The adaptive weight update laws for the neural networks are derived by minimizing the approximation error, leading to the determination of the optimal control law. Finally, numerical simulations of the reconfiguration control for a four-satellite chained tethered satellite formation are conducted to validate the effectiveness and feasibility of the proposed control strategy. The results show that the proposed controller reduces the fuel consumption by more than 70% with better control performance compared to the conventional method.