Adaptive Wavelet Neural Network Research Articles

Wave-CapNet: A Wavelet Neuron-based Wi-Fi Sensing Model for Human Identification

Gait is regarded as a unique feature for identifying people, and gait recognition is the basis of various customized services of the IoT. Unlike traditional techniques for identifying people, the Wi-Fi-based technique is unconstrained by illumination conditions and such that it eliminates the need for dense, specialized sensors and wearable devices. Although deep learning-based sensing models are conducive to the development of Wi-Fi-based identification, the latter technique relies on a large amount of data and requires a long training time, where this limits the scope of its use for identifying people. In this study, we propose a Wi-Fi sensing model called Wave-CapNet for human identification. We use data processing to eliminate errors in the raw data so that the model can extract the characteristics in channel state information (CSI). We also design a dedicated adaptive wavelet neural network to extract representative features from Wi-Fi signals with only a few epochs of training and a small number of parameters. Experiments show that it can identify human gait with an average accuracy of 99%. Moreover, it can achieve an average accuracy of 95% by using only 10% of the data and fewer than five epochs and outperforms state-of-the-art (SOTA) methods.

SEA-GWNN: Simple and Effective Adaptive Graph Wavelet Neural Network

The utilization of wavelet-based techniques in graph neural networks (GNNs) has gained considerable attention, particularly in the context of node classification. Although existing wavelet-based approaches have shown promise, they are constrained by their reliance on pre-defined wavelet filters, rendering them incapable of effectively adapting to signals that reside on graphs based on tasks at hand. Recent research endeavors address this issue through the introduction of a wavelet lifting transform. However, this technique necessitates the use of bipartite graphs, causing a transformation of the original graph structure into a bipartite configuration. This alteration of graph topology results in the generation of undesirable wavelet filters, thereby undermining the effectiveness of the method. In response to these challenges, we propose a novel simple and effective adaptive graph wavelet neural network (SEA-GWNN) class that employs the lifting scheme on arbitrary graph structures while upholding the original graph topology by leveraging multi-hop computation trees. A noteworthy aspect of the approach is the focus on local substructures represented as acyclic trees, wherein the lifting strategy is applied in a localized manner. This locally defined lifting scheme effectively combines high-pass and low-pass frequency information to enhance node representations. Furthermore, to reduce computing costs, we propose to decouple the higher- order lifting operators and induce them from the lower-order structures. Finally, we benchmark our model on several real- world datasets spanning four distinct categories, including citation networks, webpages, the film industry, and large-scale graphs and the experimental results showcase the efficacy of the proposed SEA-GWNN.

Neural network-based variable stiffness impedance control for internal/external forces tracking of dual-arm manipulators under uncertainties

The desired interaction between manipulators, objects, and environments has resulted in the internal/external force control for dual-arm manipulators being in increasing demand. Consequently, this study focused on the internal/external force tracking for dual-arm manipulator systems under external disturbances, geometries, and stiffness uncertainties which continuously lead to unsatisfactory internal force tracking. The proposed scheme is based on a two-level adaptive impedance control scheme, where the stiffness coefficient is adjusted to adapt to uncalibrated objects. An object-level hybrid impedance controller was used to regulate the external disturbance to produce a compliant response. A manipulator-level neural network-based variable stiffness impedance controller (NNVSIC) was proposed to regulate the internal force under various uncertainties. Additionally, an adaptive wavelet neural network was designed to compensate for the geometric estimation errors of the object. The variable stiffness coefficient could automatically adapt to an unknown object during the cooperation process. One advantage of the proposed method is that no prior knowledge was required. The same controller parameters could be adapted to various objects. The asymptotic stability of the proposed NNVSIC was proven via Lyapunov stability analysis. A series of experiments were conducted using two self-developed nine-degrees-of-freedom redundant manipulators. Furthermore, hard and soft objects of various geometries and stiffnesses were used to verify the effectiveness of the algorithm. The experimental results demonstrated the efficiency and superiority of the proposed controller through performance comparison with various algorithms.

ADGWN: adaptive dual-channel graph wavelet neural network for topology identification of low-voltage distribution grid

Maintaining accurate topology of the low-voltage distribution grid (LVDG) are critical to the operations and maintenance of power distribution systems. However, this goal is hard to achieve due to the fast-changing LVDG topology. To this end, we focus on the abnormal customer-transformer relationships identification in the LVDG and propose an identification method based on an Adaptive Dual-channel Graph Wavelet Neural Network (ADGWN) consisting of two identical GWNs connected with the attention mechanism. In the proposed ADGWN, two GWNs learn customer embedding simultaneously from the LVDG topology graph and the feature graph that is constructed from customer electricity consumption data with the k-Nearest Neighbor algorithm. The topology identification results of these two GNNs are then adaptively fused to form the ultimate identification result with the attention mechanism by dynamically balancing the aforementioned two types of information. To validate the performance of our proposed method, we further build a real benchmarking dataset from customer electricity consumption data collected from a certain substation in Shanghai, China. Experimental results show that the proposed ADGWN achieves 100.0% LVDG topology identification accuracy and significantly outperforms the state-of-the-art. Our proposed method can help operators of power distribution systems maintain the accurate topology in a timely and economic manner.

Practical torque sensorless super-twisting control of manipulators based on a novel integral non-singular fast terminal sliding mode with fixed-time convergence

ABSTRACT In this paper, a robust controller has been proposed for first-order uncertain non-linear systems by introducing a novel hybrid sliding surface. Using this method, contrary to previous proposed sliding surfaces, the state trajectory begins moving on the correct sliding mode and the reaching phase can be reduced to the minimum time. Thus, the transient response speed of the system and consequently the efficiency and robustness of the system increase. Moreover, the convergence of the system state in bounded time, independent of the initial conditions of the system, is guaranteed without singularity problems. Then, for torque sensorless control of permanent-magnet DC motors, as one of the most prevalent joint actuators in robotics, an adaptive super-twisting controller, based on the proposed hybrid sliding surface, has been presented. Using this method, continuous finite-time control of the DC motor’s torque without a torque sensor is realized. The DC motor load-side torque is estimated based on an adaptive wavelet neural network and it is compensated in the control loop. In addition to the stability proof and fixed-time convergence of control schemes, a set of simulation and experimental studies have been provided to evaluate the efficiency of control algorithms and investigate their feasibility.

A novel adaptive noise reduction method for field natural gas pipeline defect detection signals

To suppress noise interference and improve defect identification accuracy, we propose a transfer learning-based adaptive wavelet neural network (AWNN) method for pipeline defect detection signal denoising. Simulated noisy signals and field detection signals are used to determine the network's initial structure and fine-tune its parameters. The Meyer wavelet basis function is chosen as an excitation function based on the defect signal waveform. The pretrained structure is determined by a search algorithm. The network parameters are adaptively fine-tuned according to the sample entropy of the noise. The new genetic beetle antennae search strategy (GBA) optimization algorithm is proposed to improve the AWNN to avoid local optima. Compared with traditional noise reduction methods, the AWNN has excellent ability to recovery the pure denoised signal with a higher SNR and smaller RMSE. The AWNN can realize adaptive noise reduction of different field signals, providing reliable input for subsequent defect diagnosis.

Application of the adaptive fuzzy wavelet neural network for two‐axis trajectory control *

The accuracy of contour tracking is an important issue to be addressed for the X-Y stage in a machining tool. Applying a better control scheme can help to increase the tracking accuracy with the effects of systems' uncertainties, nonlinearities and disturbances. In this paper, to prevent these negative effects and achieve high accuracy tracking control of the two-axis motion stage, the adaptive fuzzy wavelet neural network (AFWNN) PI sliding mode controller is proposed. The adaptive algorithm is combined with the basis function of the wavelet neural network and fuzzy system to approximate the plant model functions. Then, the controller with PI sliding mode structure could be built based on the estimated system function. In this method, the time domain positioning property of wavelet and the reasoning of the fuzzy system could provide the well-estimated plant model. Moreover, the adaptive algorithm and sliding mode control scheme help to reduce the effects of parameter variation and uncertainties. The stability is analyzed via Lyapunov stability theorem to prove the asymptotic stability of the system. In the simulation and experimental results, the contours are planned through the nonuniform rational B-spline (NURBS), which is considered a high accuracy interpolator. The tested curves consist of circle, bow, heart, star and trident. The average and standard deviations of the tracking errors are compared to other methods to illustrate the performance of the proposed method. The results demonstrate that the method can reduce errors more effectively and have better localization for trajectory tracking.

Adaptive Multi-Scale Wavelet Neural Network for Time Series Classification

Wavelet transform is a well-known multi-resolution tool to analyze the time series in the time-frequency domain. Wavelet basis is diverse but predefined by manual without taking the data into the consideration. Hence, it is a great challenge to select an appropriate wavelet basis to separate the low and high frequency components for the task on the hand. Inspired by the lifting scheme in the second-generation wavelet, the updater and predictor are learned directly from the time series to separate the low and high frequency components of the time series. An adaptive multi-scale wavelet neural network (AMSW-NN) is proposed for time series classification in this paper. First, candidate frequency decompositions are obtained by a multi-scale convolutional neural network in conjunction with a depthwise convolutional neural network. Then, a selector is used to choose the optimal frequency decomposition from the candidates. At last, the optimal frequency decomposition is fed to a classification network to predict the label. A comprehensive experiment is performed on the UCR archive. The results demonstrate that, compared with the classical wavelet transform, AMSW-NN could improve the performance based on different classification networks.

Research on UUV Recovery Active Disturbance Rejection Control Based on LMNN Compensation

As a kind of intelligent marine equipment, its successful recovery is the basis of Unmanned Underwater Vehicle (UUV)’s normal working and completing the mission successfully. However, the mathematical model of UUV is severely coupled and highly nonlinear, and may be subject to complex interference in the recovery process. Besides, the model could not be determined completely. Active disturbance rejection control (ADRC) method does not need the beforehand information of the unknown disturbance and also can ensure the stability. But conventional ADRC method with fixed parameters could not adjust to UUV complicated motion control. This paper introduces the adaptive wavelet neural network optimized by Levenberg-Marquardt algorithm, and puts forward a novel ADRC control method improved by wavelet neural network algorithm and Levenberg-Marquardt algorithm, for partial system identification and uncertain model compensation. Moreover, with pitching moment change considered in the process of UUV recovery, two dimensionless hydrodynamic coefficients are defined based on near-wall effect. The simulation experiments have been tested to verify the effectiveness of the proposed control. The results indicate that the ADRC with Levenberg-Marquardt neural network could control UUV recovery process in the variable disturbance environment more stable and reduce the ADRC computational burden.

A Mixed Optimization Method Based on Adaptive Kalman Filter and Wavelet Neural Network for INS/GPS During GPS Outages

To improve the navigation performance of the navigation system combining inertial navigation system (INS) and global positioning system (GPS) under complicated environments, especially GPS outages, a navigation method - wavelet neural network based on random forest regression (RFR-WNN) to assist adaptive Kalman filter (AKF) - is proposed. AKF is employed to correct INS errors, the Kalman filter is improved by introducing adaptive factor, to suppress the influence of the complex environment and random errors on the filtering accuracy; RFR-WNN is used to construct a high-precision prediction model when GPS works well, and to provide the required observations for AKF update when GPS outages. To solve the problem that the single neural network structure is easy to cause the overfitting, unstable and low prediction accuracy due to the lack of comprehensive training samples, RFR is introduced to optimize the single WNN, which can improve the generalization ability and prediction accuracy. In order to verify the effectiveness and advancement of the proposed method, vehicle navigation experiments were carried out, the results indicate that the proposed method has better navigation accuracy and performance than compared methods during GPS outages, and this advantage is more obvious in the case that fewer samples are collected.

Robust Optimal Control of High-Speed Permanent-Magnet Synchronous Motor Drives via Self-Constructing Fuzzy Wavelet Neural Network

This article presents the design of an adaptive backstepping robust optimal control (ABROC) approach for achieving performance with high dynamic of high-speed micro permanent-magnet synchronous motor (HS-MPMSM) drive. First, a backstepping controller is designed for stabilizing the HS-MPMSM drive. To enhance the performance of the control system against external disturbances and parameter variations, an adaptive backstepping robust controller (ABRC) is developed. The ABRC combines a backstepping controller, an adaptive self-constructing fuzzy wavelet neural network (SCFWNN) identifier, and a robust controller. The proposed identifier is developed to approximate the nonlinear functions online. Furthermore, the robust controller is designed to recover the SCFWNN approximation errors. As the online adaptive control laws are derived via Lyapunov theory, thus, the ABRC stability is assured. To attain the optimal control performance, an infinite horizon optimal controller using a critic neural-network (NN) is developed and combined with ABRC to construct the ABROC approach. The critic NN is developed to approximate the optimal value function of the Hamilton–Jacobi–Bellman equation, which is used to develop the optimal controller. The experimental results are presented to verify the effectiveness of the proposed approach. The results validate that the ABROC approach is robust against parameter uncertainties and external disturbances.

Compensation of Geometric Parameter Errors for Terrestrial Laser Scanner by Integrating Intensity Correction

The accuracy of geometric parameters (mainly referred to the incidence angle and measuring distance) in a terrestrial laser scanner (TLS) is not only influenced by the TLS intrinsic systematic instrumental error but also the extrinsic received intensity data. However, the current error compensation methods for geometric parameters mainly focus on the calibration of TLS intrinsic systematic instrumental error and rarely consider the extrinsic intensity data correction. For this reason, this article presents a new method integrating the TLS intrinsic systematic instrumental error calibration and extrinsic intensity data correction to compensate the TLS geometric parameter error. The error compensation procedure is implemented as follows. First, the error compensation mathematical model integrated with TLS intrinsic systematic instrumental error calibration parameters and extrinsic intensity data correction coefficient is established. Second, the hybrid harmonic analysis (HA) and the adaptive wavelet neural network (AWNN) algorithm are proposed to calculate the TLS incidence angle error compensation values. Subsequently, the cubic spline interpolation (CSI) is applied to compute the measuring distance error compensate values. Finally, the TLS (model FARO Focus S150) and the hemispherical angle calibration instrument were used to evaluate the proposed compensation method. The experimental results demonstrate that the geometric parameters are significantly influenced by the intensity data received from TLS, and the proposed method can effectively improve the overall accuracy of the TLS incidence angle and measuring distance.

RETRACTED: Forecasting biofuel production using adaptive integrated optimization network model

Abstract Recently, biofuel production has been increasing all over the world over the past few decades owing to the relative changes in energy price and policy intervention. Several research studies are focussed on producing optimal biofuel production using numerous approaches. However, those approaches have few drawbacks such as computational complexity, less accuracy, high error value rate, etc. So to overcome such shortcomings, this paper aims in developing a novel Adaptive Integrated Optimization Network (AION approach) to attain optimal biofuel production with high accuracy and minimum error value rate. Moreover, the proposed AION approach comprises four significant phases namely Pre-processing of data, Re-construction of component, Prediction of an individual and Ensemble forecasting. The initial process utilizes an empirical mode decomposition (EMD) approach to divide a raw complicated data. The reconstruction phase employs the Fine to coarse (FTC) reconstructing approach that rebuilds the IMF into two different t-testing components namely the low-frequency component and the high-frequency component. Moreover, Gray Wolf Optimization (GWO) based Adaptive Wavelet Neural Network (AWNN) and Long Short Term Memory-Recurrent Neural Network (LSTM-RNN) approaches are applied to predict the low and the high-frequency components. Then the forecasted results are summed by using an ADD operator so as to obtain absolute forecasting results. The experimental analysis reveals that the proposed AION approach provides an optimal biofuel production with enhanced accuracy, minimum error value rate, and complexity issues.

Wavelet Adaptive Control for Robotic Manipulator With Input Deadzone

In this paper, the adaptive wavelet neural network (WNN) tracking control problem is investigated for robotic manipulator with input deadzone. The WNN is used to approximate the unknown nonlinear function and the derivative of virtual control in the system, which avoids the problem “explosion of complexity” in the traditional backsteppinng control methods, the requirement of input instruction signal is reduced. The robust term is designed to compensate the approximation errors of WNN and external disturbance. Since wavelet functions are used in WNN, its learning capability is precede to the traditional neural network for system identification. It is proved that the proposed controller can guarantee that all signals of the closed-loop system are uniformly ultimately bounded. Simulation results demonstrate the effectiveness of the proposed approach.

Path-following control of unmanned surface vehicles with unknown dynamics and unmeasured velocities

The path-following control problem of an unmanned surface vehicle (USV) with unknown dynamics and unmeasured velocities is addressed in this paper. Main contributions are summarized as below: (1) considering that vehicle velocities are unmeasured directly, a finite-time velocity observer (FVO) is designed by utilizing available position information, which contributes to the path-following control and facilitates the implementation in practical engineering; (2) based on the traditional light-of-sight guidance, a heading-surge (HS) guidance scheme is proposed to guide USV surge velocity and heading angle, simultaneously; (3) in combination with wavelet neural network (WNN) and adaptive technique, unknown dynamics involving model uncertainties and environment interferences can be estimated accurately, thereby significantly enhancing system robustness. By the aid of the HS guidance based adaptive WNN controllers, all the error signals in the closed-loop system are uniformly ultimately bounded (UUB). Besides, simulation studies and comprehensive comparisons are conducted to demonstrate the satisfactory path-following performance and superiority of the proposed scheme.

Adaptive self-recurrent wavelet neural network and sliding mode controller/observer for a slider crank mechanism

In this paper, a novel control strategy based on an adaptive Self-Recurrent Wavelet Neural Network (SRWNN) and a sliding mode controller/observer for a slider crank mechanism is proposed. The aim is to reduce the tracking error of the linear displacement of this mechanism while following a specified trajectory. The controller design consists of two parts. The first one is a sliding mode control strategy and the second part is an SRWNN controller. This controller is trained offline first, and then the SRWNN weights are updated online by the adaptive control law. Apart from the hybrid control strategy proposed in this paper, a velocity observer is implemented to replace the use of velocity sensors. The outcomes obtained in the numerical experiment section prove that the smallest tracking error is obtained for the linear and angular displacements in comparison with other strategies found in literature due to the uncertainty and disturbance rejection properties of the sliding mode and the self-recurrent wavelet neural network controllers.

Adaptive self-recurrent wavelet neural network and sliding mode controller/observer for a slider crank mechanism

Adaptive self-recurrent wavelet neural network and sliding mode controller/observer for a slider crank mechanism

A hybrid approach to model and forecast the electricity consumption by NeuroWavelet and ARIMAX-GARCH models

Today, electrical energy plays a major role in production and consumption and is of special importance in economic decision-making process. Being aware of electrical energy demand for each period is necessary to correct planning. Therefore, the forecasting of electricity consumption is important among several economic sections. Besides the traditional models, in this paper, we offer a hybrid forecast approach that combines the adaptive wavelet neural network with the ARIMA-GARCH family models and uses the effective exogenous variables on electricity consumption. Based on this approach, two hybrid models are proposed. To assess the ability of the proposed models, we forecasted the daily electricity consumption by the hybrid and benchmark models for 60 days ahead in two separate seasons (summer and winter). The empirical results showed that the proposed models have more prediction accuracy compared with the other benchmark forecast models including neural network, adaptive wavelet neural network, and ARIMAX-GARCH family models.

Adaptive recurrent neural network motion control for observation class remotely operated vehicle manipulator system with modeling uncertainty

Precise motion control of remotely operated vehicles plays an important role in a great number of submarine missions. However, the high-performance operations are difficult to realize due to the uncertainty in system modeling with self-disturbance. On the basis of the multibody system dynamics, self-disturbances from the tether and manipulator have been systematically analyzed in order to transform them into observed forces. A novel S surface–based adaptive recurrent wavelet neural network control system has been proposed on the nonlinear control of underwater vehicles, with its recurrent wavelet neural network structure designed for the approximation of the uncertain dynamics. Moreover, a robust function has been proposed to improve system robustness and convergence. The comparison shows that the remotely operated vehicle operation performance including the three-dimensional path following and vehicle-manipulator coordinate control has been greatly improved.

Adaptive wavelet neural network for terrestrial laser scanner-based crack detection

Abstract Objective, accurate, and fast assessment of civil infrastructure conditions is critical to timely assess safety risks. Current practices rely on visual observations and manual interpretation of reports and sketches prepared by inspectors in the field, which are labor intensive, subject to personal judgment and experience, and prone to error. Terrestrial laser scanners (TLS) are promising for automatically identifying structural condition indicators, as they are capable of providing coverage for large areas with accuracy at long ranges. Major challenges in using this technology are in storing significant amount of data and extracting appropriate features enabling condition assessment. This paper proposes a novel adaptive wavelet neural network (WNN)-based approach to compress data into a combination of low- and high-resolution surfaces, and automatically detect concrete cracks and other forms of damage. The adaptive WNN is designed to sequentially self-organize and self-adapt in order to construct an optimized representation. The architecture of the WNN is based on a single-layer neural network consisting of Mexican hat wavelet functions. The strategy is to first construct a low-resolution representation of the point cloud, then detect and localize anomalies, and finally construct a high-resolution representation around these anomalies to enhance their characterization. The approach was verified on four cracked concrete specimens. The experimental results show that the proposed approach was capable of fitting the point cloud, and of detecting and fitting the crack. The results demonstrated data compression of 99.4%, 72.2%, 92.4% and 78.9% for the four specimens when using low resolution fit for crack detection. For specimens 1, 2 and 3, 97.1%, 42.5% and 63.9% compression of data were obtained for crack localization, which is a significant improvement over previous TLS based crack detection and measurement approaches. Using the proposed method for crack detection would enable automatic and remote assessment of structural conditions. This would, in turn, result in reducing costs associated with infrastructure management, and improving the overall quality of our infrastructure by enhancing maintenance operations.