Prescribed-performance predefined-time tracking control of input-saturated robotic manipulators via actor-critic reinforcement learning

Accurate trajectory tracking of robot manipulators in the presence of unknown nonlinear dynamics, parametric uncertainties, and time-varying external disturbances is a challenging problem. This paper addresses the predefined-time tracking control problem for robot manipulators with unknown nonlinear dynamics, time-varying external disturbances, and input saturation constraints. A reinforcement learning-based optimal control framework is developed, in which an actor–critic neural network is employed to approximate the optimal policy and value function online, without requiring precise knowledge of system dynamics. To guarantee prescribed performance requirements, a predefined-time error transformation combined with a bounded constraint formulation is developed, facilitating systematic controller synthesis and ensuring that the tracking error remains confined within a compact prescribed set for all time. Based on this design, a predefined-time prescribed performance control scheme is derived, guaranteeing that the transformed error dynamics converge to the equilibrium within a user-specified settling time independent of initial conditions. Rigorous analysis based on predefined-time stability theory and Lyapunov methods establishes that all closed-loop signals remain uniformly bounded, while the tracking error converges to zero within a predefined time. Finally, simulation results validate the effectiveness and superiority of the proposed approach.