面向空间碎片清除的旋转系绳经验回放网络 预定时间控制

To address the fast stabilization control problem of space tether system under multiple constraints such as capture impact and abrupt mass change during debris removal missions，a predefined-time adaptive neural network control strategy based on an experience replay mechanism is proposed. First，a coupled dynamic model integrating tether swing and satellite attitude is established. A predefined-time performance function is introduced to transform the system state errors，and a controller is designed combined with the non-singular terminal sliding mode theory to ensure the system converges within a user-defined time while satisfying physical constraints. Furthermore，to solve the weight drift problem caused by the difficulty in satisfying the persistent excitation condition in traditional neural network adaptive control，a Radial Basis Function （RBF） neural network with an experience replay mechanism is designed. By using historical sample data to break temporal correlation，high-precision approximation of the system’s lumped uncertainties and external disturbances is realized. Simulation verification is carried out in two typical scenarios：static equilibrium stabilization after debris capture and dynamic equilibrium stabilization after ejection. The results show that the proposed algorithm can effectively suppress system oscillations，drive the states to converge to the equilibrium point within the predefined time， and achieve significantly better disturbance estimation errors than traditional adaptive methods.