Jumping Threshold-Based Dynamic Event-Triggered Self-Learning Control for IT-2 Fuzzy Semi-Markov Model Under Attack Detection and Mitigation

Abstract

This article investigates the optimal control synthesis for a class of semi-Markov jump systems subjected to external disturbances, by integrating advanced event triggered communication and resilient attack mitigation strategies. In an effort to accurately capture the inherent stochastic switching dynamics and high-order uncertainties, an interval type-2 fuzzy framework is established, wherein the optimal control objective is rigorously formulated through generalized algebraic Riccati equations. To alleviate network communication burdens, a jumping threshold-based dynamic event-triggered mechanism is architected, operating synergistically with a constrained Markov decision process-based framework for continuous attack detection and mitigation. Furthermore, an offline safe reinforcement learning algorithm is developed to systematically derive the optimal control gains. Grounded with Lyapunov stability theory, the asymptotic stability criterion for the closed-loop semi-Markov jump systems under the proposed jumping threshold-based dynamic event triggered scheme is rigorously established. Alongside the mathematical convergence of the optimal control gains are obtained. Finally, a comprehensive numerical simulation is provided to validate the robustness, communication efficiency, and practical effectiveness of the proposed control architecture.