System-on-a-chip (SoC)-based hardware acceleration for foreground and background identification

The rapid growth of embedded vision applications and accessibility in recent years has instigated a philosophical shift in algorithm and implementation design for artificial intelligence. With the popularization of high-definition video, the amount of data available to be processed has also increased substantially, posing massive computational and communication demands. Hardware acceleration through specialization has received renewed interest in recent years; such acceleration has generally been implemented using two chips, with the image signal processing (ISP) part being performed by a DSP, a GPU or an FPGA and the video content analytics (VCA) part being executed by a processor. GPUs consume a substantial amount of power; thus, it is challenging to deploy them in embedded environments. However, the new generation of SoC-FPGAs that are fabricated with both the microprocessor and FPGA on a single chip consumes less power and can be built into small systems, thereby offering an attractive platform for embedded applications. This study presents the hardware acceleration of a real-time adaptive background and foreground identification algorithm in a SoC, including the capture, processing and display stages. The algorithm can be performed in either 2D or 3D space. The proposed platform uses photometric invariant color, depth data and local binary patterns (LBPs) to distinguish background from foreground. The system uses minimal cell resources, an elastically pipelined architecture is used to absorb variations in processing time, and each pipeline stage is optimized to use the available FPGA primitives. Additionally, the communication-centric architecture used in this work simplifies the implementation of embedded vision algorithms.     

A class of fast fixed-time synchronization control for the delayed neural network

This paper deals with the synchronization control of a class of delayed neural networks using a fast fixed-time control theory. By employing Lyapunov stability theory, a novel sufficient criterion is derived such that two neural networks can be synchronized within a fixed-time. Compared with some existing results, the proposed controller can render two neural networks faster synchronized. A numerical example is given to demonstrate the effectiveness of the criterion.     

A particle inerter system for structural seismic response mitigation

This study introduces a novel particle inerter system (PIS) designed for vibration mitigation of structures. The new system comprises an inerter subsystem, a spring, and a tuned particle element, where the spring is used for tuning the particle element and the inerter subsystem is set for energy absorption and dissipation. The structural performance and the vibration mitigation effect of the PIS are assessed in terms of displacement and acceleration responses. An optimal design method is developed for a PIS under a performance-oriented design framework. Following the criterion of lightweight control, the added mass of the PIS is minimized under the constraints of target displacement and acceleration responses. A parametric analysis is performed and the robustness of the PIS for seismic response mitigation is verified. Design cases are carried out for the illustration of the proposed design method. The results show that the structural displacement and acceleration responses can be reduced significantly with the help of a PIS. Compared with the particle tuned mass damper with the same parameters, both the energy absorption and dissipation effects of the PIS are increased and the relative displacement response of the container in the PIS is reduced by the inerter subsystem. Under the same performance target, the required physical mass of the container and particles in the PIS is minimized and is significantly smaller than that of the conventional particle tuned mass damper.     

Adaptive and online network intrusion detection system using clustering and Extreme Learning Machines

Despite the large volume of research conducted in the field of intrusion detection, finding a perfect solution of intrusion detection systems for critical applications is still a major challenge. This is mainly due to the continuous emergence of security threats which can bypass the outdated intrusion detection systems. The main objective of this paper is to propose an adaptive design of intrusion detection systems on the basis of Extreme Learning Machines. The proposed system offers the capability of detecting known and novel attacks and being updated according to new trends of data patterns provided by security experts in a cost-effective manner.     

A new approach based on discrete-time high-order neural networks with delays and impulses

This paper is concerned with the stability of discrete-time high-order neural networks (HONNs) with delays and impulses. Without applying the Lyapunov function, some sufficient conditions, which ensure the exponential stability and asymptotic stability of considered networks involving delays and impulses, are derived based on the fixed point theory. Finally, several numerical examples are given to demonstrate the effectiveness of the obtained results.     

A novel event-triggered mechanism for networked cascade control system with stochastic nonlinearities and actuator failures

This paper deals with the event-triggered control for networked cascade control systems. Unlike conventional event-triggered schemes that predetermine a fixed threshold to reduce the data-releasing rate, this paper proposes a novel event-triggered mechanism (ETM) in an adaptive way. Under this ETM, it has the following merits: (1) the data-releasing rate remains at a lower level so as to save limited network bandwidth; (2) the reliability of control systems can be improved since the threshold of ETM is increased gradually with the elapse of time till the next event is generated. An integrated model of networked cascade control systems with consideration of stochastic nonlinearity, actuator failures and ETM is established. Sufficient conditions are obtained to ensure the mean-square stability and stabilization of networked cascade control systems. Finally, two examples are exploited to show the effectiveness of the proposed method.     

Pinning impulsive stabilization for BAM reaction-diffusion neural networks with mixed delays

The exponential stabilization of BAM reaction-diffusion neural networks with mixed delays is discussed in this article. At first, a general pinning impulsive controller is introduced, in which the control functions are nonlinear and the pinning neurons are determined by reordering the state error. Next, based on the designed control protocol and the Lyapunov–Krasovskii functional approach, some novel and useful criteria, which depend on the diffusion coefficients and controlling parameters, are established to guarantee the global exponential stabilization of the considered neural networks. Finally, the effectiveness of the proposed control strategy is shown by two numerical examples.     

A clustering algorithm based TS fuzzy model for tracking dynamical system data

This paper presents an optimal fuzzy partition based Takagi Sugeno Fuzzy Model (TSFM) in which a novel clustering algorithm, known as Modified Fuzzy C-Regression Model (MFCRM), has been proposed. The objective function of MFCRM algorithm has been developed by considering of geometrical structure of input data and linear functional relation between input–output data. The MFCRM partitions the data space to create fuzzy subspaces (rules). A new validation criterion has been developed for detecting the right number of rules (subspaces) in a given data set. The obtained fuzzy partition is used to build the fuzzy structure and identify the premise parameters. Once, right number of rules and premise parameters have been identified, then consequent parameters have been identified by orthogonal least square (OLS) approach. The cluster validation index has been tested on synthetic data set. The effectiveness of MFCRM based TSFM has been validated on benchmark examples, such as Boiler Turbine system, Mackey–Glass time series data and Box–Jenkins model. The model performance is also validated through high-dimensional data such as Auto-MPG data and Boston Housing data.     

Observer-based fault detection for uncertain nonlinear systems

This paper addresses L₂ observer-based fault detection issues for a class of nonlinear systems in the presence of parametric and dynamic uncertainties, respectively. To this end, three different types of uncertain affine nonlinear system models studied in this paper are described first. Then, the integrated design schemes of L₂ observer-based fault detection systems are derived with the aid of Hamilton–Jacobi inequalities (HJIs), respectively. Numerical examples are also provided in the end to demonstrate the effectiveness of the proposed results.     

Terminal iterative learning control for discrete-time nonlinear systems based on neural networks

The terminal iterative learning control is designed for nonlinear systems based on neural networks. A terminal output tracking error model is obtained by using a system input and output algebraic function as well as the differential mean value theorem. The radial basis function neural network is utilized to construct the input for the system. The weights are updated by optimizing an objective function and an auxiliary error is introduced to compensate the approximation error from the neural network. Both time-invariant input case and time-varying input case are discussed in the note. Strict convergence analysis of proposed algorithm is proved by the Lyapunov like method. Simulations based on train station control problem and batch reactor are provided to demonstrate the effectiveness of the proposed algorithms.     

Distributed cooperative learning for a group of uncertain systems via output feedback and neural networks

This paper studies the problem of adaptive neural network (NN) output-feedback control for a group of uncertain nonlinear multi-agent systems (MASs) from the viewpoint of cooperative learning. It is assumed that all MASs have identical unknown nonlinear dynamic models but carry out different periodic control tasks, i.e., each agent system has its own periodic reference trajectory. By establishing a network topology among systems, we propose a new consensus-based distributed cooperative learning (DCL) law for the unknown weights of radial basis function (RBF) neural networks appearing in output-feedback control laws. The main advantage of such a learning scheme is that all estimated weights converge to a small neighborhood of the optimal value over the union of all system estimated state orbits. Thus, the learned NN weights have better generalization ability than those obtained by traditional NN learning laws. Our control approach also guarantees the convergence of tracking errors and the stability of closed-loop system. Under the assumption that the network topology is undirected and connected, we give a strict proof by verifying the cooperative persisting excitation condition of RBF regression vectors. This condition is defined in our recent work and plays a key role in analyzing the convergence of adaptive parameters. Finally, two simulation examples are provided to verify the effectiveness and advantages of the control scheme proposed in this paper.     

Bayesian network based extreme learning machine for subjectivity detection

Subjectivity detection is a task of natural language processing that aims to remove ‘factual’ or ‘neutral’ content, i.e., objective text that does not contain any opinion, from online product reviews. Such a pre-processing step is crucial to increase the accuracy of sentiment analysis systems, as these are usually optimized for the binary classification task of distinguishing between positive and negative content. In this paper, we extend the extreme learning machine (ELM) paradigm to a novel framework that exploits the features of both Bayesian networks and fuzzy recurrent neural networks to perform subjectivity detection. In particular, Bayesian networks are used to build a network of connections among the hidden neurons of the conventional ELM configuration in order to capture dependencies in high-dimensional data. Next, a fuzzy recurrent neural network inherits the overall structure generated by the Bayesian networks to model temporal features in the predictor. Experimental results confirmed the ability of the proposed framework to deal with standard subjectivity detection problems and also proved its capacity to address portability across languages in translation tasks.     

Adaptive neural network decentralized stabilization for nonlinear large scale interconnected systems with expanding construction

A backstepping-based adaptive neural network decentralized stabilization approach is presented for the expanding construction of a class of nonlinear large scale interconnected systems in this paper. The expanding construction of large scale interconnected systems is to add some new subsystems into the original system during the operation of the original system. For stabilization of the expanding system, it is more realistic to keep the decentralized control laws of the original subsystems unchanged. And the decentralized control laws of the new subsystems must be designed to stabilize both itself and the resultant large scale system. In this paper, neural networks are used to approximate the unknown nonlinear functions in the new subsystems and the unknown nonlinear interconnection functions. The decentralized control laws and the parameter adaptive laws of the new subsystems are designed by using backstepping technique for the expanding construction of the large-scale interconnected system. Based on Lyapunov stability theory, the uniform and ultimate boundedness of all signals in the closed-loop system is proved. Two illustrative examples show the feasibility of the presented approach.     

A LPV modeling of turbocharged spark-ignition automotive engine oriented to fault detection and isolation purposes

This paper illustrates the derivation of a linear parameter varying (LPV) model approximation of a turbocharged Spark-Ignition (SI) automotive engine and its usage in designing a model-based fault detection and isolation (FDI) scheme. The LPV approximation is derived from a detailed nonlinear mathematical model of the engine on the basis of the well known Jacobian approach. The resulting LPV representation is then exploited for synthesizing a bank of LPV-FDI $H_{\infty }/H_{?}$ Luenberger observers. Each observer is in charge of detecting a particular class of fault and is designed for having low sensitivity to all other exogenous inputs so as to allow an effective fault isolation. The adopted FDI scheme is gain-scheduled and exploits a set of engine variables, assumed to be measurable on-line, as a scheduling parameters. The goodness of the LPV approximation of the engine model and the effectiveness of the LPV-FDI architecture are demonstrated by several numerical simulations.     

Stop and Go adaptive strategy for synchronization of delayed memristive recurrent neural networks with unknown synaptic weights

Although the drive-response synchronization problem of memristive recurrent neural networks (MRNNs) has been widely investigated, all the existing results are based on the assumption that the parameters of the drive system are known in prior, which are difficult to implement in real-life applications. In the present paper, a Stop and Go adaptive strategy is proposed to investigate the synchronization control of chaotic delayed MRNNs with unknown memristive synaptic weights. Firstly, by defining a series of measurable logical switching signals, a switched response system is constructed. Subsequently, by utilizing the logical switching signals, several suitable parameter update laws are proposed, then some different adaptive controllers are devised to guarantee the synchronization of unknown MRNNs. Since the parameter update laws are weighted by the logical switching signals, they will work or stop automatically with the switch of the unknown weights of drive system. Finally, two numerical examples with their computer simulations are provided to illustrate the effectiveness of the proposed adaptive synchronization schemes.     

Mode-dependent stochastic synchronization criteria for Markovian hybrid neural networks with random coupling strengths

This paper is concerned with the stochastic synchronization problem for a class of Markovian hybrid neural networks with random coupling strengths and mode-dependent mixed time-delays in the mean square. First, a novel inequality is established which is a double integral form of the Wirtinger-based integral inequality. Next, by employing a novel augmented Lyapunov–Krasovskii functional (LKF) with several mode-dependent matrices, applying the theory of Kronecker product of matrices, Barbalat’s Lemma and the auxiliary function-based integral inequalities, several novel delay-dependent conditions are established to achieve the globally stochastic synchronization for the mode-dependent Markovian hybrid coupled neural networks. Finally, a numerical example with simulation is provided to illustrate the effectiveness of the presented criteria.     

A new singular system approach to output feedback sliding mode control for fractional order nonlinear systems

This paper studies the problem of output feedback sliding mode control (OFSMC) for fractional order nonlinear systems. A necessary and sufficient condition for the existence of a sliding surface is obtained by a new singular system approach and a linear matrix equality (LMI), which reduces the conservativeness of the system. Then an OFSMC law is designed based on a fractional order Lyapunov method, which ensures that the resulting fractional closed-loop system is asymptotically stable and the states of the fractional closed-loop system converge to the sliding surface in finite time. A fractional electrical circuit is discussed to illustrate the effectiveness of the proposed approach.     

Adaptive neural design frame for uncertain stochastic nonlinear non-lower triangular pure-feedback systems with input constraint

This paper dedicates to dealing with the adaptive neural design problem for uncertain stochastic nonlinear systems with non-lower triangular pure-feedback form and input constraint. On the basis of the mean-value theorem, the pure-feedback structure is first transformed into the desired affine structure, and then the well-known backstepping technology is applied to construct the actual input signal of the controller. Although the considered system has a fairly complex structure, a new adaptive neural tracking controller design frame is established via the flexible application of radial basis function (RBF) neural networks’ (NNs’) structural characteristics. The proposed design frame guarantees the control objective of this paper can be achieved. Finally, a simulation example is given to further illustrate the availability of the proposed control scheme.