IET Cyber-Physical Systems: Theory and Applications最新文献

Sampled-data synchronization problem for singular Markovian jump systems (SMJSs) subject to aperiodic sampled-data control is investigated. Firstly, via constructing mode-dependent one-sided loop-based Lyapunov functional (LBLF) and two-sided LBLF, two different stochastically admissible conditions are suggested for error SMJSs with aperiodic sampled-data. It is guaranteed that the slave system is stochastically synchronized to the master system on the basis of the proposed stochastically admissible conditions. Secondly, two corresponding mode-dependent aperiodic sampled-data controller design approaches are provided for error SMJSs based on two different conditions, separately. Finally, the validity of these approaches is demonstrated by a direct current (DC) motor model. It also demonstrated that the two-sided LBLF method possesses a larger upper bound of sampled-data period than the one-sided LBLF method.

This study focuses on the distributed adaptive tracking control of heterogeneous multi-agent systems with unknown leader dynamics in a directed graph. In contrast to the reported leader-following consensus studies, the prior knowledge of the leader is supposed to be cognised to some or all of followers, the situation that the leader's dynamics is totally unrecognised but can be learnt for each individual follower is considered. A data-driven learning algorithm using the systems data is developed to reconstruct the unknown systems matrix. Then, an adaptive distributed dynamic compensator is exploited to provide the leader's state estimation in a directed graph. Afterwards, a dynamic output feedback control law for each agent is projected. Theoretical analysis shows that the proposed algorithms not only ensure that all followers can identify the unknown system matrix, but also guarantee that the distributed output leader-following consensus control with heterogeneous dynamics is achieved without any global information. Finally, a numerical example is provided to testify the proposed algorithms.

Video has become an essential medium to monitoring, identification and knowledge sharing. For industrial applications, especially Industrial Internet of Things (IIoT), videos encoded with specific protocols are transferred to smart gateways. In a typical IIoT scenario, the protocol of the video is firstly recognised, which prepares for subsequent video tasks. Due to the constrained resources in such scenarios, the video quality can be deteriorated during encoding and compression processes, which is challenging for IIoT. Recently, there have been extensive works focussing on the protocol identification (PI) and video quality enhancement (VQE) tasks on IIoT edge devices using deep neural networks (DNNs). Since DNNs often require high computational resources, complex networks can hardly be deployed on edge devices. An IIoT system which can efficiently identify the stream protocol and enhance the video quality is proposed in this study. The light-weighted network designs and inference optimisation techniques have been proposed for PI and VQE to realise efficient deployments. Our proposed system employed on an IIoT edge device can achieve an accuracy of higher than 97.52% with fast inference speed for PI. For the VQE task, our system has demonstrated superior performance (15.230 FPS, 0.773 FPS/W) in comparison with the state-of-the-art methods.

The resource allocation problem in a distributed multi-agent system is considered in this study. First, the authors develop a predefined-time distributed algorithm and analyse its convergence analysis using the Lyapunov stability theory, in which the local constraint is ensured by a differential projection operator. Thus, a predefined time is obtained by a time-varying time-based generator. Second, to reduce the communication consumption between agents, the authors develop a static as well as a dynamic-based event-triggered control scheme, where the information broadcast only occurs at some discrete time instants. Moreover, the three proposed algorithms converge precisely to the global optimal solution. Besides, the Zeno behaviour is excluded in the above static and dynamic event-triggered mechanisms. Finally, the authors test the proposed algorithms' efficiency based on the provided numerical examples.

The novel concept of a Cyber-Human Social System (CHSS) and a diverse and pluralistic ‘mixed-life society’ is proposed, wherein cyber and human societies commit to each other. This concept enhances the Cyber-Physical System (CPS), which is associated with the current Society 5.0, a social vision realised through the fusion of cyber (virtual) and physical (real) spaces following information society (Society 4.0 and Industry 4.0). Moreover, the CHSS enhances the Human-CPS, the Human-in-the-Loop CPS (HiLCPS), and the Cyber-Human System by intervening in individual behaviour pro-socially and supporting consensus building. As a form of architecture that embodies the CHSS concept, the Cyber-Human Social Co-Operating System (Social Co-OS) that combines cyber and human societies is shown. In this architecture, the cyber and human systems cooperate through the fast loop (operation and administration) and slow loop (consensus and politics). Furthermore, the technical content and current implementation of the basic functions of the Social Co-OS are described. These functions consist of individual behavioural diagnostics, interventions in the fast loop, group decision diagnostics and consensus building in the slow loop. Subsequently, this system will contribute to mutual aid communities and platform cooperatives.

As today's power grid is evolving into a densely interconnected cyber-physical system (CPS), a high fidelity and multifaceted testbed environment is needed to perform cybersecurity experiments in a realistic grid environment. Traditional standalone CPS testbeds lack the ability to emulate complex cyber-physical interdependencies between multiple smart grid domains in a real-time environment. Therefore, there are ongoing research and development (R&D) efforts to develop an interconnected CPS testbed by sharing geographically dispersed testbed resources to perform distributed simulation while analysing simulation fidelity. This paper presents a networked federation testbed for cybersecurity evaluation of today's and emerging smart grid environments. Specifically, it presents two novel testbed architectures, including cyber federation and cyber-physical federation, identifies R&D applications, and also describes testbed building blocks with experimental case studies. It also presents a novel co-simulation interface algorithm to facilitate distributed simulation within cyber-physical federation. The resources available at the PowerCyber CPS security testbed at Iowa State University (ISU) and the US Army Research Laboratory are utilised to develop this platform for performing multiple experimental case studies pertaining to wide-area protection and control applications in power system. Finally, experimental results are presented to analyse the simulation fidelity and real-time performance of the testbed federation.

Learning-enabled components (LECs) such as deep neural networks are used increasingly in cyber-physical systems (CPS) since they can handle the uncertainty and variability of the environment and increase the level of autonomy. LECs, however, may compromise system safety since their predictions may have large errors, for example, when the data available at runtime are different than the data used for training. This study considers the problem of efficient and robust out-of-distribution detection for learning-enabled CPS. Out-of-distribution detection using a single input example is typically not robust and may result in a large number of false alarms. The proposed approach utilises neural network architectures that are used to compute efficiently the nonconformity of new inputs relative to the training data. Specifically, variational autoencoder and deep support vector data description networks are used to learn models for the real-time detection of out-of-distribution high-dimensional inputs. Robustness can be improved by incorporating saliency maps that identify parts of the input contributing most to the LEC predictions. We demonstrate the approach using simulation case studies of an advanced emergency braking system and a self-driving end-to-end controller, as well as a real-world data set for autonomous driving. The experimental results show a small detection delay with a very small number of false alarms while the execution time is comparable to the execution time of the original LECs.

It is meaningful to study the real-time state monitoring and identification of integrated energy system and grasp its state in time for stable operation. A state identification method based on multi-class data equalisation and extreme gradient boost (XGBoost) is proposed for integrated energy systems. First, Latin hypercube sampling is used to simulate the load at different moments. Different system states are set up and combined with the simulative load at different moments to determine the system operation state at different moments. Then, the energy flow model is used to calculate the system power flow under different states, and the feature indexes are obtained to form the original data set. Aiming at the unbalanced data, the oversampling technology is used to preprocess data to achieve the balance of data sets. The pre-processed data is utilised to train the XGBoost, and the optimal hyperparameters of the model are obtained based on the K-fold cross-validation and grid search. Finally, the pre-processed data set is used to verify the proposed method. The calculation results show the accuracy of the identification model reaches 87.79%. Compared with traditional methods, the model can accurately identify the operating state of the electricity–heat energy system at any time section.

High proportion of wind power in the power grid leads to the problem of power system frequency instability, which requires the wind farm itself to have the ability of frequency adjustment; therefore, it is particularly important to conduct modelling of wind farm frequency regulation (WFFR) response characteristics. During the modelling process, it is generally necessary to establish a model for each working condition separately, which will bring huge workload. In addition, the accuracy of the model decreases when the frequency response is non-linear. Therefore, this paper investigates the modelling of WFFR response characteristics in different working conditions. A data preprocessing method based on WFFR strategy and modelling methods is introduced. Then, data-based transfer function models of WFFR response characteristics for different working conditions are constructed. After that, the gaps between different models are measured using a gap metric technique to analyse dynamic similarity between models. Finally, in order to make up for the defect of transfer function models, a non-linear autoregressive with exogenous input neural networks (NARXNN) model of WFFR response characteristics is constructed utilising lumped data of all working conditions; then, the trained model is tested by the data of each working condition to verify the accuracy and universality.