Reinforcement Learning-Based Event-Triggered Predefined-Time Optimal Fuzzy Control for Nonlinear Constrained Systems

This article focuses on the switching event-triggered predefined-time fuzzy optimal control for full-state constrained nonlinear systems based on reinforcement learning (RL) algorithms. To fulfill the asymmetric full-state time-varying constraints (AFTCs), an emerging universal transformed function and error transformation were delicately adopted because they assist in eliminating the feasibility requirement induced by the related constrained issues. In addition, an RL-based optimal dynamic surface control solution was performed by leveraging the characteristics of predefined-time stability. The highlights of this study are that a modified nonsingular predefined-time filter featuring the hyperbolic tangent item and an adaptive parameter was constructed to overcome the curse of dimensionality and that the AFTC property was integrated into the optimal framework. Moreover, communication costs and control expenses were considerably minimized using an amended switching event-triggered mechanism. Under the predefined-time stability criterion, the reported control tactic ensures that the tracking error can approach the vicinity of the origin within a user-specified time while excluding the violation of the AFTC. Herein, two illustrative analyzes with comparisons are provided to confirm the efficacy and benefits of the reported control algorithm.