Adaptive Neural Control for Flexible Joint Manipulators with Uncertainties: A Fully Actuated System Approach

In this paper, an adaptive command-filtered control scheme is proposed for flexible joint manipulators with dynamic uncertainties and external disturbances, leveraging the fully actuated system (FAS) framework. Distinct from traditional firstorder state-space method, the FAS approach directly addresses the second-order dynamics of the system without resorting to firstorder transformation, thereby simplifying controller design and reducing structural complexity. To circumvent the 'explosion of complexity' inherent in recursive backstepping procedures, a second-order command filter technique is integrated, transforming intricate differentiation operations into algebraic computations. A radial basis function neural network (RBFNN) is employed to approximate lumped uncertainties, enhancing the system's adaptability to unknown dynamics and external disturbances. Rigorous stability analysis based on Lyapunov theory guarantees that all closed-loop signals remain uniformly ultimately bounded. The efficiency of the proposed scheme is conclusively validated via simulations on a two-link robotic system, demonstrating superior trajectory tracking performance and robustness against perturbations.