NN Weights Research Articles

Composite-Learning-Based Adaptive Neural Control for Dual-Arm Robots With Relative Motion.

This article presents an adaptive control method for dual-arm robot systems to perform bimanual tasks under modeling uncertainties. Different from the traditional symmetric bimanual robot control, we study the dual-arm robot control with relative motions between robotic arms and a grasped object. The robot system is first divided into two subsystems: a settled manipulator system and a tool-used manipulator system. Then, a command filtered control technique is developed for trajectory tracking and contact force control. In addition, to deal with the inevitable dynamic uncertainties, a radial basis function neural network (RBFNN) is employed for the robot, with a novel composite learning law to update the NN weights. The composite learning is mainly based on an integration of the historic data of NN regression such that information of the estimate error can be utilized to improve the convergence. Moreover, a partial persistent excitation condition is employed to ensure estimation convergence. The stability analysis is performed by using the Lyapunov theorem. Numerical simulation results demonstrate the validity of the proposed control and learning algorithm.

Dynamic evolutionary model based on a multi-sampling inherited HAPFNN for an aluminium electrolysis manufacturing system

It is technically difficult to accurately establish a dynamic adaptive model of an aluminium electrolysis manufacturing system (AEMS) because the system has many complex characteristics, such as multiple parameters, dynamic time variance, and a non-Gaussian distribution of process data. Inspired by the overall superiority of particle filtering theory in dealing with non-linear non-Gaussian problems, this paper presents a novel method based on a multi-sampling inherited hybrid annealed particle filter neural network (MSI-HAPFNN). Firstly, the neural network’s (NN’s) weights and thresholds are used as the state variables of hybrid annealed particle filter; Secondly, the hybrid proposal distribution obtained by sampling the above state variables is employed to replace the posterior proposal distribution in the standard particle filter (PF) algorithm as the importance density function, thereby adjusting the NN’s weights and thresholds in real time. Thirdly, the model achieves the features of multi-sampling and inheritance by introducing NN and PF weights, and using adaptive inheritance method. Therefore, this paper systematically proposes the theoretical construction framework and experimental procedure of MSI-HAPFNN. Furthermore, this article also introduces a genetic algorithm to thoroughly evaluate the prediction potential. The proposed model has been tested on the real-world system for aluminium electrolysis manufacturing and compared with several closely related frameworks. The experimental results show that the MSI-HAPFNN model can significantly improve the self-adaptive ability of the object system to working conditions and the prediction accuracy of power consumption, which is helpful in finding optimal design parameters in an AEMS.

Online optimal consensus control of unknown linear multi-agent systems via time-based adaptive dynamic programming

This paper considers the online optimal consensus control problem for unknown linear discrete-time (DT) multi-agent systems (MASs). Based on time-based adaptive dynamic programming (ADP) method, the control policies are designed by utilizing the current and recorded data of unknown MASs. The critic-actor NN frameworks are employed to approximate the performance indexes and optimal control policies, respectively. The NN weights are updated once at the sampling instant to produce real-time online control. Furthermore, the control policies are proved to effectively drive the MASs to achieve consistency and satisfy the Nash equilibrium. Finally, a numerical example is implemented to shown the feasibility of the control scheme.

Fixed‐time observer based adaptive neural network time‐varying formation tracking control for multi‐agent systems via minimal learning parameter approach

This study proposes a novel control scheme to investigate the time-varying formation tracking control problem for multi-agent systems with model uncertainties and the absence of leader's velocity measurements. For each agent, a novel fixed-time cascaded leader state observer (CLSO) without velocity measurements is first designed to reconstruct the states of the leader. Radial basis function neural networks (RBFNNs) are adopted to deal with the model uncertainties online. Taking the square of the norm of the NN weight vector as a newly developed adaptive parameter, a novel RBFNN-based adaptive control scheme with minimal learning-parameter approach and fixed-time CLSO is then constructed to tackle the time-varying formation tracking problem. The uniform ultimate boundedness property of the formation tracking error is guaranteed through Lyapunov stability analysis. Finally, two simulation scenario results demonstrate the effectiveness of the proposed formation tracking control scheme.

A Neural Network-Based Adaptive Backstepping Control Law With Covariance Resetting for Asymptotic Output Tracking of a CSTR Plant

A robust nonlinear adaptive controller merging a backstepping approach with neural networks is proposed for a nonlinear non-affine model. The work presented here is evaluated on a complex uncertain model of a continuous stirred tank reactor plant including an unknown varying parameter that enters the complexity model. By exploiting NN and adaptive backstepping approximation methods, an equivalent adaptive NN controller is constructed to achieve robust asymptotic output tracking control. The robustness to uncertainties as well as the lack of informative process data is the main enhancement of this work. This is attained through the implementation of the covariance resetting algorithm in the least square estimation of the NN weight tuning algorithm. The proposed novel control algorithm has been analyzed using Lyapunov analysis. In addition to excellent output trajectory tracking performance, the proposed approach has a profound benefit in terms of substantially lower control effort in comparison to the established work in the literature. In terms of applications in the petrochemical industry, lower control effort can translate to a more energy-efficient actuator, leading to lower costs over a long-run operation. The proposed method's feasibility for chemical process control was shown via numerical simulation.

Multimodal biometric cryptosystem for human authentication using fingerprint and ear

The multimodal biometrics is mainly used for the purpose of person certification and proof. Lot of biometrics is used for human authentication. In which ear and fingerprint are efficient one. There are three vital phases involved in the biometric detection which include the Preprocessing, Feature extraction and the classification. Initially, preprocessing is done with the help of median filter which lends a helping hand to the task of cropping the image for choosing the position. Then, from the preprocessed Finger print and ear image texture and shape features are extracted. In the long run, the extracted features are integrated. The integrated features, in turn, are proficiently classified by means of the optimal neural network (ONN). Here, the NN weights are optimally, selected with the help of firefly algorithm (FF). The biometric image is classified into fingerprint and ear if the identical person images are amassed in one group and the uneven images are stored in a different group. The performance of the proposed approach is analyzed in terms of evaluation metrics.

Online Nash-optimization tracking control of multi-motor driven load system with simplified RL scheme

Although the optimal tracking control problem (OTCP) has been addressed recently, only the single-input system is considered in the recent literature. In this paper, the OTCP of unknown multi-motor driven load systems (MMDLS) is addressed based on a simplified reinforcement learning (RL) structure, where all the motor inputs with different dynamics will be obtained as a Nash equilibrium. Thus, the performance indexes associated with each input can be optimized as an outcome of a Nash equilibrium. Firstly, we use an identifier to reconstruct MMDLS dynamics, such that the accurate model required in the general control design is avoided. We use the identified dynamics to drive Nash-optimization inputs, which include the steady-state controls and the RL-based controls. The steady-state controls are designed with the identified system model. The RL-based controls are designed using the optimization method with the simplified RL-based critic NN schemes. We use the simplified RL structures to approximate the cost function of each motor input in the optimal control design. The NN weights of both the identified algorithm and simplified RL-based structure are approximated by using a novel adaptation algorithm, where the learning gains can be optimized adaptively. The weight convergences and the Nash-optimization MMDLS stability are all proved. Finally, numerical MMDLS simulations are implemented to show the correctness and the improved performance of the proposed methods.

Asymptotic tracking control of uncertain nonholonomic wheeled mobile robot with actuator saturation and external disturbances

This paper deals with the asymptotic tracking control for the uncertain nonholonomic wheeled mobile robot system subjected to actuator saturation and external disturbances simultaneously. A dynamic system is introduced to deal with the actuator saturation, radial basis function neural networks (RBF NNs) are employed to approximate the unknown closed-loop system dynamics, and an adaptive sliding mode feedback term is used to compensate for the approximation error as well as external disturbances. Consequently, a novel adaptive neural controller is designed to guarantee the stability of the closed-loop system and the asymptotic convergence of tracking errors. Meanwhile, the convergence of NN weights is verified, which means that accurate approximation of the unknown closed-loop system dynamics can be obtained and the constant weights can be reused to perform the same or similar control tasks. Finally, simulation studies illustrate the effectiveness of the proposed scheme.

Novel relative relevance score for estimating brain connectivity from fMRI data using an explainable neural network approach

BackgroundFunctional integration or connectivity in brain is directional, non-linear as well as variable in time-lagged dependence. Deep neural networks (DNN) have become an indispensable tool everywhere, by learning higher levels of abstract and complex patterns from raw data. However, in neuroscientific community they generally work as black-boxes, leading to the explanation of results difficult and less intuitive. We aim to propose a brain-connectivity measure based on an explainable NN (xNN) approach. New MethodWe build a NN-based predictor for regression problem. Since we aim to determine the contribution/relevance of past data-point from one region i in the prediction of current data-point from another region j, i.e. the higher-order connectivity between two brain-regions, we employ layer-wise relevance propagation (Bach et al., 2015) (LRP, a method for explaining DNN predictions), which has not been done before to the best of our knowledge. Specifically, we propose a novel score depending on weights as a quantitative measure of connectivity, called as relative relevance score (xNN-RRS). The RRS is an intuitive and transparent score. We provide an interpretation of the trained NN weights with-respect-to the brain-connectivity. ResultsFace validity of our approach is demonstrated with experiments on simulated data, over existing methods. We also demonstrate construct validity of xNN-RRS in a resting-state fMRI experiment. ComparisonOur approach shows superior performance, in terms of accuracy and computational complexity, over existing state-of-the-art methods for brain-connectivity estimation. ConclusionThe proposed method is promising to serve as a first post-hoc explainable NN-approach for brain-connectivity analysis in clinical applications.

Neural-network-based output-feedback control with stochastic communication protocols

This paper is concerned with the neural-network-based (NN-based) output-feedback control issue for a class of nonlinear systems. For the purpose of effectively mitigating the phenomena of data congestion/collision, the stochastic communication protocols are favorably utilized to orchestrate the data transmissions, and the resultant closed-loop plant is represented by a so-called protocol-induced Markovian jump system with uncertain transition probability matrices. Taking such an uncertainty probability into account, a novel iterative adaptive dynamic programming (ADP) algorithm is developed to obtain the desired suboptimal solution with the help of auxiliary quasi-HJB equation, and the algorithm convergence is also investigated via the intensive use of the mathematical analysis. In this ADP framework, an NN-based observer with a novel adaptive tuning law is first adopted to reconstruct the system states. Then, based on the reconfigurable system, an actor–critic NN scheme with online learning is developed to realize the considered control strategy. Furthermore, in light of the Lyapunov theory, some sufficient conditions are derived to guarantee the stability of the zero equilibrium point of the closed-loop system as well as the boundedness of the estimation errors for critic and actor NN weights. Finally, a simulation example is employed to demonstrate the effectiveness of the developed suboptimal control scheme.

Optimization of RNN-Based Speech Activity Detection

Speech activity detection (SAD) is an essential component of automatic speech recognition systems impacting the overall system performance. This paper investigates an optimization process for recurrent neural network (RNN) based SAD. This process optimizes all system parameters including those used for feature extraction, the NN weights, and the back-end parameters. Three cost functions are considered for SAD optimization: the frame error rate, the NIST detection cost function, and the word error rate of a downstream speech recognizer. Different types of RNN models and optimization methods are investigated. Three types of RNNs are compared: a basic RNN, long short-term memory (LSTM) network with peepholes, and a coordinated-gate LSTM (CG-LSTM) network introduced by Gelly and Gauvain. Well suited for nondifferentiable optimization problems, quantum-behaved particle swarm optimization is used to optimize feature extraction and posterior smoothing, as well as for the initial training of the neural networks. Experimental SAD results are reported on the NIST 2015 SAD evaluation data as well as REPERE and AMI meeting corpora. Speech recognition results are reported on the OpenKWS’13 test data. For all tasks and conditions, the proposed optimization method significantly improves the SAD performance and among all the tested SAD methods the CG-LSTM model gives the best results.

Adaptive backstepping neural network control for three dimensions trajectory tracking of robotic airships

The robotic airship provides a unique aerostatic platform for various applications, and these applications require high-precise trajectory tracking. However, it is a challenging problem due to nonlinearity and uncertainty of airship dynamics. This paper proposes an adaptive backstepping neural network control (ABNNC) approach to address this problem. First, the kinematics model and dynamics model of the robotic airship are presented. Second, the control problem of trajectory tracking is formulated, and a trajectory controller is designed using backstepping approach. A radial basis function neural network (RBFNN) is employed to approximate the uncertain dynamics model of the airship, and an adaptive law is designed to update the NN weight in the processing of approximation. The ultimate boundedness of the tracking errors are proven based on the Lyapunov theory. Finally, simulations are presented to illustrate the effectiveness and high precision of the designed controller.

Adaptive stabilization control for high-order nonlinear time-delay systems with its application

This paper investigates the state-feedback stabilization problem in the smooth case for a class of high-order nonlinear systems with time delays. By generalizing a novel radial basis function neural network (RBF NN) approximation approach to high-order nonlinear systems, we successfully remove the power order restriction and the growth conditions on system nonlinearities. It should be pointed out that the knowledge of NN nodes and weights does not need to be known a priori and operate on-line, and the adaptive parameter is only one. Furthermore, without imposing any growth assumptions on system nonlinearities, we construct a smooth adaptive state-feedback controller which guarantees the closed-loop system to be semi-globally uniformly ultimately bounded (SGUUB). Finally, we apply the proposed scheme to a single-link robot system and a numerical example to demonstrate the effectiveness of the controller.

Adaptive optimal control of unknown nonlinear systems with different time scales

The adaptive optimal control of unknown nonlinear system with different time scales is considered in this paper. The commonly used singular perturbation theory (SPT) to solve this problem is based on the accurately reduced system model, which is extremely difficult to be obtained in practical application. To overcome this difficulty, an adaptive dynamic programing-based optimal control algorithm with the simplified actor–critic–identifier structure is developed. A different time scales dynamic neural network (DTSDNN) identifier with a novel updating law derived from a properly designed Lyapunov function is proposed to estimate the unknown system dynamics. Furthermore, the critic NN with an improved adaptive law considering the NN weight estimation error information is designed, which can achieve faster convergent speed compared with the commonly used gradient method. Lyapunov approach is used to guarantee exponential convergence to a bounded region in the neighborhood of the optimal control and uniformly ultimately bounded (UUB) stability of the closed-loop system. Two examples are provided to illustrate the effectiveness and applicability of the developed approach.

Observer-based adaptive optimal control for unknown singularly perturbed nonlinear systems with input constraints

This paper introduces an observer-based adaptive optimal control method for unknown singularly perturbed nonlinear systems with input constraints. First, a multi-time scales dynamic neural network U+0028 MTSDNN U+0029 observer with a novel updating law derived from a properly designed Lyapunov function is proposed to estimate the system states. Then, an adaptive learning rule driven by the critic NN weight error is presented for the critic NN, which is used to approximate the optimal cost function. Finally, the optimal control action is calculated by online solving the Hamilton-Jacobi-Bellman U+0028 HJB U+0029 equation associated with the MTSDNN observer and critic NN. The stability of the overall closed-loop system consisting of the MTSDNN observer, the critic NN and the optimal control action is proved. The proposed observer-based optimal control approach has an essential advantage that the system dynamics are not needed for implementation, and only the measured input U+002F output data is needed. Moreover, the proposed optimal control design takes the input constraints into consideration and thus can overcome the restriction of actuator saturation. Simulation results are presented to confirm the validity of the investigated approach.

Real-Time Implementation of a Discrete Reduced Order Neural Observer: Linear Induction Motors Application

AbstractThis paper deals with a discrete-time neural observer for flux estimation of a linear induction motor, which is based on a Recurrent High Order Neural Network (RHONN). The RHONN weights are tuned on-line, with no off-line learning phase required, using an extended Kalman filter based algorithm. The observer stability and boundedness of the state estimation and NN weights are proven using the Lyapunov approach. Knowledge of the system model is not required. The applicability of this observer is illustrated by real-time implementation for flux estimation of a linear induction motor benchmark.

Adaptive neural dynamic surface control for a class of uncertain nonlinear systems with disturbances

In this paper, adaptive dynamic surface control is investigated for a class of uncertain nonlinear systems with unknown bounded disturbances in strict-feedback form. Dynamic surface control technique is connected with radial basis function neural networks (RBFNNs) based control framework to avoid the explosion problem of complexity. The composite laws are constructed by prediction error and compensated tracking error between system state and serial–parallel estimation model for NN weights updating. Using Lyapunov techniques, the uniformly ultimate boundedness stability of all the signals in the closed-loop systems is guaranteed. Simulation results illustrate the superiority of the proposed scheme and verify the theoretical analysis.

Consensus-based distributed cooperative learning control for a group of discrete-time nonlinear multi-agent systems using neural networks

This paper focuses on the cooperative learning capability of radial basis function neural networks in adaptive neural controllers for a group of uncertain discrete-time nonlinear systems where system structures are identical but reference signals are different. By constructing an interconnection topology among learning laws of NN weights in order to share their learned knowledge on-line, a novel adaptive NN control scheme, called distributed cooperative learning control scheme, is proposed. It is guaranteed that if the interconnection topology is undirected and connected, all closed-loop signals are uniform ultimate bounded and tracking errors of all systems can converge to a small neighborhood around the origin. Moreover, it is proved that all estimated NN weights converge to a small neighborhood of their common optimal value along the union of all state trajectories, which means that the estimated NN weights reach consensus with a small consensus error. Thus, all learned NN models have the better generalization capability than ones obtained by the deterministic learning method. The learned knowledge is also adopted to control a class of uncertain systems with the same structure but different reference signals. Finally, a simulation example is provided to verify the effectiveness and advantages of the distributed cooperative learning control scheme developed in this paper.

Adaptive control of nonlinear uncertain active suspension systems with prescribed performance

This paper proposes adaptive control designs for vehicle active suspension systems with unknown nonlinear dynamics (e.g., nonlinear spring and piece-wise linear damper dynamics). An adaptive control is first proposed to stabilize the vertical vehicle displacement and thus to improve the ride comfort and to guarantee other suspension requirements (e.g., road holding and suspension space limitation) concerning the vehicle safety and mechanical constraints. An augmented neural network is developed to online compensate for the unknown nonlinearities, and a novel adaptive law is developed to estimate both NN weights and uncertain model parameters (e.g., sprung mass), where the parameter estimation error is used as a leakage term superimposed on the classical adaptations. To further improve the control performance and simplify the parameter tuning, a prescribed performance function (PPF) characterizing the error convergence rate, maximum overshoot and steady-state error is used to propose another adaptive control. The stability for the closed-loop system is proved and particular performance requirements are analyzed. Simulations are included to illustrate the effectiveness of the proposed control schemes.

Face Recognition Classifier Design Based on the Genetic Algorithm and Neural Network

In the course of the face feature match, many classifiers have been designed. The neural network is usually selected as a classifier because of its validity and universality, whereas its training time, training epochs, and its convergence, all are not satisfied to us. It is often influenced by the author’s experience. In the case, a collaborative genetic algorithm and neural network is presented as a new face recognition classifier. The one thing is to train the NN weights by the GA until the stopping criterion is met, and the next thing is to use the BP algorithm to continue to train the network. The training time and training epochs have been improved in the experiment of the face recognition on ORL face database. The simulation shows the validity of methods.