Neural network-based synchronisation of free-floating space manipulator's joint motion and mother spacecraft's attitude for active debris removal

A free-floating space manipulator attached to a spacecraft introduces challenges in simultaneously controlling the motion of the space manipulator and its mother spacecraft's attitude. This study aims to develop a neural network-based control approach to synchronously control the space manipulator motion and spacecraft attitude, improving the control performance in trajectory tracking, error reduction and eliminating uncertainties that arise from external disturbances, high-frequency noise, oscillations and imprecise knowledge of changes in the control system. Image-based Visual Servoing (IBVS) is used to provide inputs in terms of image features of the debris to the conventional controllers such as sliding mode control (SMC) and proportional-integral-derivative (PID). SMC is used to control the motion of the space manipulator. The unscented Kalman filter (UKF) provides the estimate of the spacecraft's attitude as an input to the PID controller to control the attitude. PID controller provides a feed-forward compensation to the SMC to counter spacecraft reactions to manipulator motion, while maintaining the attitude of the spacecraft. The neural network is introduced in the control strategy to enhance the performance of conventional controllers by dynamically optimising their gains and coefficients. This adaptability improves trajectory tracking accuracy, response to changes in the system and autonomy. The stability of this control approach is proven using the Lyapunov stability theorem, demonstrating a global asymptotic stability. The neural-network-based synchronous control approach is tested and validated by numerical simulations and comparative analysis in the MATLAB-Simulink environment. The results demonstrate an enhanced control performance in terms of accurate trajectory tracking, faster 100% convergence to zero error and more robustness to uncertainties. Outcomes highlight the potential of neural network-based control approaches in real-world applications that manage the free-floating space manipulators during uncooperative debris capture.