Scholarlink

<p>To address the security control issue of intelligent terminals undergoing changes due to task variations, a model—free sliding mode (SM) resilient controller is developed for uncertain nonlinear cyber—physical systems (CPSs) under the influence of Denial—of—Service (DoS) attacks and False Data Injection (FDI) attacks. Firstly, the unified mathematical model of cyber—attacks is formulated. An initial—value error conversion function is incorporated to transform the error with an arbitrary initial value into a variable having an initial value of zero. Subsequently, a reinforcement learning (RL) algorithm built on a single hidden—layer neural network is constructed. This algorithm is capable of real—time estimation of the CPS model and perturbation, which exhibits strong time—varying characteristics. Leveraging the prescribed performance control method (PPCM) and the predefined time convergence (PTC) theory, a novel model—free sliding mode resilient controller is designed for the CPS under cyber—attacks. The predefined time convergence of the control system is rigorously proven through theoretical analysis. Simulation results indicate that this control strategy can guarantee that the CPS with an arbitrary initial state converges to the desired trajectory within the specified time. Moreover, the trajectory tracking error satisfies the prescribed performance requirements. The steady—state error is less than 0.0008, and the intensity of cyber—attacks has no impact on the control effectiveness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Model-free sliding mode resilient control of cyber-physical system based on reinforcement learning

  • 期刊论文

  • 发表日期：

    2025年6月26日

查看全文
  • Chun-Wu Yin

摘要

To address the security control issue of intelligent terminals undergoing changes due to task variations, a model—free sliding mode (SM) resilient controller is developed for uncertain nonlinear cyber—physical systems (CPSs) under the influence of Denial—of—Service (DoS) attacks and False Data Injection (FDI) attacks. Firstly, the unified mathematical model of cyber—attacks is formulated. An initial—value error conversion function is incorporated to transform the error with an arbitrary initial value into a variable having an initial value of zero. Subsequently, a reinforcement learning (RL) algorithm built on a single hidden—layer neural network is constructed. This algorithm is capable of real—time estimation of the CPS model and perturbation, which exhibits strong time—varying characteristics. Leveraging the prescribed performance control method (PPCM) and the predefined time convergence (PTC) theory, a novel model—free sliding mode resilient controller is designed for the CPS under cyber—attacks. The predefined time convergence of the control system is rigorously proven through theoretical analysis. Simulation results indicate that this control strategy can guarantee that the CPS with an arbitrary initial state converges to the desired trajectory within the specified time. Moreover, the trajectory tracking error satisfies the prescribed performance requirements. The steady—state error is less than 0.0008, and the intensity of cyber—attacks has no impact on the control effectiveness.