Neural-Approximation-Based Fixed-Time Sliding Mode Control with Varying Exponent for Robotic Manipulators Under Actuator Saturation
Abstract
High-precision tracking for robotic manipulators is complicated by coupled nonlinear dynamics, uncertain parameters, external disturbances, and the bounded torque that every joint actuator can deliver. To meet these challenges simultaneously, this work develops a novel adaptive fixed-time control framework resting on three components. First, a continuous nonsingular sliding manifold with state-dependent exponents is constructed, which removes the switching action from the sliding variable and stays nonsingular across the whole state space. Second, a radial basis function neural network is embedded in the control law to learn the lumped uncertainty online, removing the need for an accurate dynamic model or a preset disturbance bound. Since the network compensates the lumped uncertainty, the switching gain that remains in the control law can be kept small, which reduces chattering. Third, an auxiliary dynamic system compensates the actuator-saturation effect. A Lyapunov-based stability proof establishes that the tracking error enters a small residual set around the origin within a settling time bounded independently of where the manipulator starts. Comparative simulation studies confirm the effectiveness and superiority of the proposed strategy.