This article investigates adaptive fuzzy bipartite consensus control for multi-agent systems (MASs) with guaranteed performance. By simplifying the bipartite consensus control problem as a local tracking problem, the prescribed performance of the bipartite consensus error is ensured. In addition, distributed adaptive fuzzy bipartite consensus control scheme is developed for each agent using a series of smooth functions and the fuzzy logic system (FLS). It is essential to have only one adaptive method in each controller. With the proposed control scheme, it is proved that all the signals in the closed-loop system are bounded and the bipartite consensus error converges to a user-prescribed interval asymptotically as time goes to infinity. Finally, the effectiveness of the proposed scheme is validated by a networked pendulum system.
This paper discusses the drag-free attitude control problem of two low-earth orbit drag-free satellites with two test masses, which are used for the detection of space gravitational waves. The control schemes of the science mode are designed, and a robust controller capable of resisting disturbances are proposed. First of all, the whole system dynamics model of the satellite with two test masses is established. Decoupling the complex system dynamics model based on characteristics of dynamic coupling and control time-frequency, three loops called the spacecraft attitude control loop, the drag-free control loop and the suspension control loop are produced. Then, taking consideration of the control requirements in frequency domain and the spectral models of various external disturbances and sensor noises, constrains for sensitivity function and complementary sensitivity function of each control loop are derived utilizing the mixed-sensitivity method of $H_{\infty}$ robust control theory. With selecting appropriate weight functions, an $H_{\infty}$ robust controller is designed. Finally, simulation results indicate that not only the inter-satellites pointing error but also the pose errors and residual acceleration of test masses satisfy the science performance specification, which verifies the effectiveness of the designed controller.
Spare parts life circle value is sensitive to various factors and has the traits of obvious uncertainty, so the comprehensive analysis of the cost-effectiveness becomes more complex. When procuring and selecting spare parts, it is necessary not only to consider efficiency, but also to consider costs; It is necessary not only to consider whether it is “affordable” but also to consider whether it is “affordable”, and strive to achieve the principle of harmonizing the advanced nature of spare parts with the sustainability. In this paper, A type of solution to the comprehensive assessment on spare parts life circle cost-effectiveness which is gray analysis method has been applied. On this basis, the corresponding weight according to relative importance of the primary factor of efficiency function to spare parts life circle cost has been determined. Finally, the assessment method is presented, and a practical example is analyzed. There are two points as for conclusions. (1) the analysis method to cost-effectiveness is an excellent mode on the science basis between the optimal usage of spare parts guarantee budget and the usage standard of its budgets, and it can satisfy the comprehensive balance relationship between a guarantee budget-usage of mission request for spare parts of an aviation department (2) the method of gray system valuation can handle the indetermination information within the analysis of the expenses-effect problem nicely, the analyzing result which has been synthesized can provides a type of path performances for the guarantee effects of spare parts such as good or bad, so as to have a better military and economy value.
This paper presents a novel approach for detecting the presence of valve stiction in control loops using a convolutional neural network (CNN) model. The proposed approach uses a combination of wavelet reconstruction and filtering techniques to preprocess the process variable versus output variable (PV(OP)) plot data before feeding it into the CNN model. The model was trained using simulated data and evaluated using data from the ISDB dataset with the F1 score as the primary evaluation metric. With our proposed approach, the model achieved the F1 score of 0.90, with a precision of 0.87 and recall of 0.93, on the test set. We also investigated the effects of different parameters on the model's performance, such as the number of epochs and output filters of convolutional layers, as well as the use of different model inputs. The results showed that the threshold for wavelet reconstruction had a significant impact on the model's performance and using both PV(OP) plots from filtered and wavelet reconstruction data as input provided better performance. Our findings demonstrate the potential of using CNN models with preprocessed PV(OP) plot data to improve the detection of valve stiction in control loops, and our approach may have broader applications in the field of process control.
The problem of synchronization in networks of neural mass model populations with discrete couplings is consid-ered. The considered network is hybrid one, therefore Mikheev approach is applied to transform it to the network with time-varying delayed couplings. Thus the problem of hybrid network synchronization is reduced to the studying of synchronization in networks with delayed couplings, which was previously solved by analytical means. It is showed that the Laplace matrix spectrum and maximum sampling interval are defining for networks dynamics. The dynamics of 5 neural mass model populations with discrete couplings was simulated for 3 different situations. The first case deal with the asymptotic synchronization, when both maximum eigenvalue of Laplacian and maximum sampling interval are small enough. The second case is about $\varepsilon$ -synchronization, which is achieved for small enough maximum eigenvalue of Laplacian and big sampling intervals. And the last case is desynchronization of oscillations, which has been observed for big values of Laplacian eigenvalues and sampling intervals.
Aiming at the problem of feature point interference and noise in visual servoing (VS) control, an image feature tracking control method based on adaptive dynamic programming (ADP) is proposed. The output signal of the image sensor is used as the feedback signal of the VS manipulators system for closed-loop control. While the adaptive state observer is used to estimate the image disturbances in the VS model of the manipulator, and the cost function is improved according to the observed disturbance values. On the basis of ADP technology, a critic neural network (NN) was constructed to derive the Hamilton-Jacobi-Bellman (HJB) equation. According to Lyapunov theory, it is proved that the visual servoing manipulators system is asymptotically stable. Finally, numerical simulations are given to verify the effectiveness of the algorithm.
In this paper, combining the advantages of regularized echo state network and improved sparrow search optimization, a novel prediction control method is proposed. Compared with the traditional recurrent neural network, the prediction model based the regularized echo state network has the advantages of simple training and less computational burden. Through using the reverse learning strategy and the Levy flight strategy, the sparrow optimization algorithm is improved for optimizing the reservoir parameters of the regularized echo state network. Finally, for the predictive control of the buck converter, the simulation experiments demonstrate the effectiveness of the proposed predictive control method.
In this paper, a new attacked data compensation (ADC) algorithm under denial of service (DoS) attacks and event-triggered based data driven predictive iterative learning control (DDPILC) method for a class of nonlinear networked control systems (NCSs) is proposed. A new attacked data compensation algorithm by using historical and forecast data information is proposed to against DoS attacks when the system suffers from DoS attacks. Meanwhile, In order to save network communication resources, an event triggerd mechanism is proposed in this paper. The convergence of the tracking control error is given by a rigorous theoretical analysis. Finally, the effectiveness of the proposed method is further verified by simulation.
Hysteresis system identification is a research topic in nonlinear system identification of long history. This paper proposes a novel recurrent neural network architecture to carry out hysteresis system identification. We first modify the original Prandtl-Ishlinskii (PI) model with one threshold parameter to a PI model with two threshold parameters, and then develop recurrent neural network based on the hybrid of the modified PI model and the Preisach model. We then evaluate our proposed network on systems simulated by different models, and explores the potential of model based neural networks for hysteresis system identification. As shown by numerical simulations, the proposed recurrent neural network can achieve better performances in describing different kinds of hysteresis systems.
After training, the parameters and weights of the standard echo state network are fixed, which will result in poor prediction accuracy for different types of input data. Therefore, a weak-consciousness echo state network (W-ESN) for time series prediction is proposed in this article. By introducing the concept of “label” in the network, the network can select the optimal network weights and parameters for different tasks based on the defined labels. Moreover, the network structure is improved by adding multiple reservoirs in series. Finally, the effectiveness of the W-ESN is proved by two examples.
In this paper, we study a class of fractional-order memristive neural networks (FMNNs) with time-delays and interactions. The main objective is to achieve finite-time synchronization (FTS) through the implementation of a feedback controller, utilizing techniques such as Filippov solutions, differential inclusion theory, and Lyapunov stability theorem. In fact, the interactions between two networks are diverse, and this paper proposes a new memristive neural networks model between two networks with interactive connections. In addition, this article proposes a series of inequalities based on the boundedness of the interaction functions to solve the interaction terms. Finally, the effectiveness of the proposed approach is demonstrated through a simulation example.
General Matrix-Matrix Multiplication (GEMM) is a commonly used kernel in machine learning, scientific computing and many other applications. Designing a customized GEMM accelerator can bring obvious performance and power consumption benefits. In this paper, we first perform a detailed workload characterization for different sizes of GEMM kernels. Then, we make a comprehensive design space exploration to find the Pareto optimal architecture configurations. Lastly, we compare two versions of multiple GEMM accelerator systems with main-stream BLAS libraries (e.g., OpenBLAS, MKL and cuBLAS). The proposed GEMM acceleration hardware system shows higher energy efficiency than existing software implementations.
Recommender systems (RSs) for products and services stay central to the operations of commercial banks in China. Efficiency, effectiveness and resilence of relevant methods are therefore the main focus. In this paper, a novel Behaviour Trace-based Unsupervised Deep Embedding (BTUDE) architecture has been developed for online recommendation of multiple products in the mobile bank application of a large commercial bank in China (LCB). BTUDE is implemented with a Variational AutoEncoder (VAE) model in which a ResNet1D network and its reverse implementation, modified for 1D time series processing, play the role of encoder and decoder, respectively. In this architecture, Customer behaviour data grouped in many time series are first processed with a novel normalization method, and then fed into the VAE model. As a quantitative form of the customer profile system, BTUDE represented as the latent space embedding generated by VAE is built up independent of scenarios for any specific product recommendation. In recommendation events for each specific product, BTUDE similarity between potential customers and those already possessed the product can be used as the feature to further detect target customers via a two-component Gaussian Mixture Model (GMM). Different types of tests have been carried out. Online A/B tests have been conducted for a money market fund product. Performance for wealth management and life insurance products are also evaluated. The promising results show that BTUDE is effectively adapted to online environment for recommendation shifting among multiple products in a fast pace.
Ultrasonography is the preferred method for detecting lung lesions, which has been widely recognized and adopted. In order to reduce the work stress and risk of infection for health care workers, lung ultrasound (LUS) scanning robots can be used to perform this task instead of doctors. For the LUS scanning robot, the first critical step is the detection of the scanned areas of the patient's lungs. The algorithms using computer vision and deep learning have made significant progress in the field of target detection, which are being applied more and more frequently to robotic autonomous decision-making. However, some of the performance of the traditional convolutional neural networks (CNN) still needs to be improved due to the more stringent requirements for real-time, accuracy and hardware cost of LUS scanning robots. The effectiveness of CNN combined with computer vision for detecting scanned sites in LUS scanning robots is explored and a lightweight Yolo network based on the attention mechanism is proposed. We use Yolo V4- Tiny model as the backbone network. Then the depth feature information of the feature layer output from the backbone network are extracted by adding space and channel attention mechanism, and the prediction results are output by using Yolo head model. Meanwhile, we also optimize the loss function of the network to further improve the performance of the network. Compared to several classical algorithms, the proposed method improves the target detection performance of the LUS scanning robot, which achieves 98.83% average precision, 96.87% precision, and 91.00% F1 score. We apply the proposed algorithm to the designed LUS scanning robot system and conduct clinical experiments. The results have shown that the proposed method has a 3D localization accuracy of 7.53 ± 0.37mm for the lung scanned sites, which has the potential to be applied to the target detection of the LUS scanning robot.
A distributed resilient finite-time control approach is proposed for formation control of multiple nonholonomic mobile robots with unknown actuator attacks and under directed communication topology. First, the leader's information is estimated in a finite time by introducing a distributed finite-time observer. Then, a resilient finite-time formation tracking control protocol is designed with the help of the observers, the adaptive control approach, and the input-output feedback linearization technique. Rigorous analysis is provided to demonstrate that all followers can maintain a desired geometric formation and track a leader robot within a finite time. Finally, the efficiency of the derived resilient formation control law is illustrated through a numerical example.
By using an event-based intelligent critic mechanism, this paper studies the optimal control problem for disturbed nonlinear continuous-time multi-input systems. First, the optimal control strategies, the worst disturbance strategy, and the Hamilton-Jacobi-Isaacs (HJI) equation are derived. Then, by introducing the event-triggered mechanism, the event-based optimal controls, the event-based worst disturbance, and the event-based HJI equation are given. Also, a novel triggering condition is proposed. Moreover, an intelligent critic mechanism is developed to approximately solve the event-based HJI equation, thereby obtaining the event-based near-optimal controls and the event-based approximate worst disturbance. Especially, during neural critic learning, we establish a novel weight updating rule. Finally, an example is given to verify the effectiveness of the established mechanism.
This paper investigates the sampled-data control problem for linear systems via a looped functional method. By integrating on both sides of system model from $t_{k}$ to $t_{k+1}$ , a new zero equation is obtained. According to system model, a simple looped functional is constructed, where the full information on the states from $x(t_{k})$ to $x(t)$ and from $x(t)$ to $x(t_{k+1})$ is involved. Based on the looped functional, by utilizing the improved free-matrix-based integral inequality, adding new zero term, and the convex analysis method, new stability and stabilization conditions are derived, which have less conservatism compared with existed ones. Finally, a well-known example is given to illustrate the effectiveness of the proposed method.
In view of the problems in the existing ice prediction model for transmission tower lines, such as the random selection of input and output vectors, the low prediction accuracy, and the low prediction efficiency, an IGWO-LSSVM transmission line ice prediction model based on grey correlation weight is proposed. Firstly, the data reduction model is obtained by extracting the effective information in the historical micro-meteorological data based on the gray correlation analysis, determining the correlation weight between the micro-meteorological information and the icing formation, and selecting the data information with high correlation degree with the line icing as the feature vector. Secondly, the prediction model is obtained by calculating the penalty factor and kernel function width of Least Squares Support Vector Machines (LSSVM) by the Improved Gray Wolf Optimizer (IGWO). Finally, the simulation analysis of a transmission line icing in Jilin province shows that the ice prediction model proposed in this paper has high prediction accuracy and small error.
Pine nematode disease, also named Pine wilt disease (PWD), has become a major disaster for the global forestry, devastating a large number of pine forests in the world and leading to great economic losses. In this paper, the small-sample forest image data is obtained through forest fire surveillance cameras and pre-processed by some augmentation methods. Then an automatic screening method for PWD based on the TransUNet network is proposed. This paper uses a small amount of image data captured by forest fire surveillance cameras instead of unmanned aerial vehicle (UAV) images. We perform pre-processing operations on the datasets, such as haze removal, clipping, and image filling. Multiple loss functions were selected, and a new loss function based on precision and recall named PR loss function, is developed in this paper. The results demonstrate that the proposed model has relatively high accuracy, with precision, recall, and F1 score of about 0.870, 0.756, and 0.807. Compared with many spectrum analysis methods and deep learning models, our proposed method can obtain better performance under small-sample and low-resolution restrictions.
