Reinforcement Learning Control With Self-Regulated Prescribed Performance for Input-Saturated Systems Under External Disturbances

This paper presents an optimal control scheme for a class of nonlinear systems subject to input saturation and external disturbances, achieved through the integration of reinforcement learning (RL) and a self-regulated prescribed performance control (SRPPC) algorithm. The proposed architecture employs an Identifier-Critic-Actor RL structure, implemented via interval type-2 fuzzy logic systems (IT2FLSs), to accurately approximate unknown nonlinear dynamics while optimizing control performance. To circumvent the singularity issues inherent in conventional PPC, the SRPPC strategy is developed to dynamically initialize the prescribed performance bounds (PPBs) and autonomously relax them during severe disturbances or actuator saturation, thereby maintaining the tracking error within a feasible envelope. By synergizing RL-based optimization with the SRPPC safety mechanism, the resulting RL-SRPPC controller not only minimizes operational costs but also guarantees transient safety and strict adherence to performance specifications. Numerical simulations on a one-link manipulator demonstrate the robustness and superiority of the proposed scheme compared to existing methodologies.