Adaptive Neural Sliding Mode Control for Singular Semi-Markovian Jump Systems Against Actuator Attacks
Adaptive Neural Sliding Mode Control for Singular Semi-Markovian Jump Systems Against Actuator Attacks
复制标题
奇异半马尔可夫跳跃系统的自适应神经滑模控制对抗执行器攻击 IEEE 系统
DOI：
10.1109/tsmc.2019.2898428
复制
发表时间：
2021-03-01
影响因子：
8.7
通讯作者：
Zou, Yuanyuan
中科院分区：
文献类型：
--
作者：
Cao, Zhiru;Niu, Yugang;Zou, Yuanyuan
针对执行器攻击的奇异半马尔可夫跳变系统（S - MJSs），研究了自适应滑模控制（SMC）问题。其中，转移率依赖于随机驻留时间且不是常数，并且系统状态不可获取。此外，通过通信网络传输的控制信号的脆弱性意味着执行器可能接收到被攻击的控制信号。为了降低执行器攻击的影响，采用神经网络技术来逼近对手注入的虚假信息。同时，引入滑模观测器来估计未测量的状态。提出了一种自适应滑模控制律，以确保估计状态和误差能够到达滑模面，并能保证奇异半马尔可夫跳变系统的随机容许性。最后，通过一个例子说明了本文中的方法。
The adaptive sliding mode control (SMC) problem is addressed for singular semi-Markovian jump systems (S-MJSs) against actuator attacks, in which the transition rates rely on the random sojourn time and are not constant, and the system states are unavailable. Moreover, the vulnerability of control signals transmitted via communication network means that the actuators may receive the attacked control signals. For the sake of reducing the effect of actuator attacks, the neural network technique is used to approximate the false information injected by adversaries. Meanwhile, a sliding mode observer is introduced to estimate the unmeasured states. An adaptive SMC law is proposed to guarantee that the estimation states and errors can reach to the sliding surfaces, and the stochastic admissibility of the singular S-MJSs can be ensured. In the end, an example is applied to illustrate the method in this paper.