This article presents a co-design method of dynamic periodic event-triggering mechanism (DPETM) and output-based control law for networked control systems (NCSs) subject to inevitable imperfections such as limited network resources, exogenous disturbances and successive packet dropouts. The developed DPETM has the potential to make full use of scarce network resources compared to the periodic sampled-data control on the one hand, and is natural for practical digital platforms instead of a discretized version of continuous event-triggered control (CETC) one the other hand. In this paper, the whole closed-loop system is modeled a unified hybrid system by incorporating bilateral DPETM, successive packet dropouts effects and external disturbances, and an improved hybrid Lyapunov function is constructed for stability analysis. Based on the hybrid framework, some sufficient conditions are presented to guarantee$\mathcal {L}_{2}$-stability of the system while tolerating a maximum allowable number of successive packet dropouts (MANSD). Furthermore, new linear matrix inequality (LMI) conditions for co-design of control law and triggering strategy are derived by introducing some constraints with the free variables and a modified Young’s inequality. In addition, this paper considers two scenarios relating to whether an acknowledgment scheme is employed or not. Finally, the effectiveness of the proposed method is verified by two illustrated examples and comparative studies. Note to Practitioners—With the increasing requirement for reducing unnecessary waste of communication resources in modern industries, it is necessary to consider an alternative paradigm of the time-triggered control. In many digital control applications such as smart power grid, vehicle and remote surgery, the control tasks are often executed periodically, which makes it difficult for the theoretical results of CETC to be directly applied to practical digital platforms. The DPETM proposed in this paper, which only verifies the triggering conditions periodically at each sampling instant, is more suitable for practical digital platforms such as time-sliced embedded architectures compared to CETC. This paper investigates the stability problem for NCSs with DPETM subject to packet dropouts. Different from most of the existing literature on randomness of packet loss, this paper focuses on impact of the continuity of packet dropouts on system performance. Based on the hybrid framework and the Lyapunov approach, a co- design methodology of triggering condition and control law is proposed to ensure that the system satisfies the desired performance and can admit a MANSD.

This paper investigates the problem of data-driven and dynamic periodic event-triggered control (DPETC) for unknown cyber-physical systems (CPSs) against Denial-of-Service (DoS) attacks and successive packet losses. An unified hybrid system framework is proposed to describe the whole CPS, in which DoS attacks are characterized in terms of duration and frequency. Building on the hybrid system framework, we present model-based co-design conditions for a dynamic periodic event-triggering mechanism (DPETM) and controller to ensure$\mathcal {L}_{2}$-stability of the CPSs while tolerating both DoS attacks and a maximal allowable number of successive packet dropouts (MANSDs). By integrating the model-based conditions with the data-based parametric representation and employing a linear fractional transformation (LFT) approach, corresponding data-based sufficient conditions are derived. Finally, numerical simulations of a batch reactor system demonstrate the effectiveness of the proposed co-design approach.

In this paper, a novel model of Gabor Filtering based Deep Network (GFDN) for hyperspectral image classification is proposed. First, spatial features are extracted via Gabor filtering from the three principal components. Gabor filter can capture physical structures of hyperspectral images, such as specific orientation information. Then, the Gabor features and spectral features are simply staked to form combined features. Finally, high-level features are learnt by a stacked sparse auto-encoder deep network. Since the limited training samples negatively affect the classification performance in deep learning, here, an effective way is designed to simulate more training samples. By using both the real and virtual samples, the parameters of deep network can be better learnt and updated, leading to more robust and accurate classification results. Experiments on the real hyperspectral data set reveal the superior performance of the proposed method over some well-known classification methods.