Neural Network-Based Adaptive Control in Task-Space for Dual-Arm Space Manipulators

To address uncertainties in the system dynamic model, external disturbances, and dynamic coupling during on‑orbit operations of dual‑arm space manipulators, this paper proposes an adaptive control strategy that integrates a radial basis function (RBF) neural network with a sliding‑mode error–driven weight update law. The controller primarily employs an RBF neural network for active online learning and compensation of the system’s nonlinear dynamics and uncertainties, thereby enhancing control precision. Moreover, a sliding‑mode–driven adaptive law is proposed to update network weights in real time, enhancing the controller’s robustness and adaptability. The simulation results show that the proposed adaptive controller can effectively suppress the impact of uncertainties, allowing for fast and accurate tracking of the desired trajectory.