This paper studies the problem of optimal parallel tracking control for continuous-time general nonlinear systems.Unlike existing optimal state feedback control,the control input of the optimal parallel control is int...This paper studies the problem of optimal parallel tracking control for continuous-time general nonlinear systems.Unlike existing optimal state feedback control,the control input of the optimal parallel control is introduced into the feedback system.However,due to the introduction of control input into the feedback system,the optimal state feedback control methods can not be applied directly.To address this problem,an augmented system and an augmented performance index function are proposed firstly.Thus,the general nonlinear system is transformed into an affine nonlinear system.The difference between the optimal parallel control and the optimal state feedback control is analyzed theoretically.It is proven that the optimal parallel control with the augmented performance index function can be seen as the suboptimal state feedback control with the traditional performance index function.Moreover,an adaptive dynamic programming(ADP)technique is utilized to implement the optimal parallel tracking control using a critic neural network(NN)to approximate the value function online.The stability analysis of the closed-loop system is performed using the Lyapunov theory,and the tracking error and NN weights errors are uniformly ultimately bounded(UUB).Also,the optimal parallel controller guarantees the continuity of the control input under the circumstance that there are finite jump discontinuities in the reference signals.Finally,the effectiveness of the developed optimal parallel control method is verified in two cases.展开更多
Smart manufacturing refers to optimization techniques that are implemented in production operations by utilizing advanced analytics approaches. With the widespread increase in deploying industrial internet of things(I...Smart manufacturing refers to optimization techniques that are implemented in production operations by utilizing advanced analytics approaches. With the widespread increase in deploying industrial internet of things(IIOT) sensors in manufacturing processes, there is a progressive need for optimal and effective approaches to data management.Embracing machine learning and artificial intelligence to take advantage of manufacturing data can lead to efficient and intelligent automation. In this paper, we conduct a comprehensive analysis based on evolutionary computing and neural network algorithms toward making semiconductor manufacturing smart.We propose a dynamic algorithm for gaining useful insights about semiconductor manufacturing processes and to address various challenges. We elaborate on the utilization of a genetic algorithm and neural network to propose an intelligent feature selection algorithm. Our objective is to provide an advanced solution for controlling manufacturing processes and to gain perspective on various dimensions that enable manufacturers to access effective predictive technologies.展开更多
A neural-network-based adaptive gain scheduling backstepping sliding mode control（NNAGS-BSMC） approach for a class of uncertain strict-feedback nonlinear system is proposed.First, the control problem of uncertain st...A neural-network-based adaptive gain scheduling backstepping sliding mode control（NNAGS-BSMC） approach for a class of uncertain strict-feedback nonlinear system is proposed.First, the control problem of uncertain strict-feedback nonlinear systems is formulated. Second, the detailed design of NNAGSBSMC is described. The sliding mode control（SMC） law is designed to track a referenced output via backstepping technique.To decrease chattering result from SMC, a radial basis function neural network（RBFNN） is employed to construct the NNAGSBSMC to facilitate adaptive gain scheduling, in which the gains are scheduled adaptively via neural network（NN）, with sliding surface and its differential as NN inputs and the gains as NN outputs. Finally, the verification example is given to show the effectiveness and robustness of the proposed approach. Contrasting simulation results indicate that the NNAGS-BSMC decreases the chattering effectively and has better control performance against the BSMC.展开更多
In order to develop predictive control algorithms for efficient energy management and monitoring for residential grid connected photovoltaic systems, accurate and reliable photovoltaic(PV) power forecasts are required...In order to develop predictive control algorithms for efficient energy management and monitoring for residential grid connected photovoltaic systems, accurate and reliable photovoltaic(PV) power forecasts are required.A PV yield prediction system is presented based on an irradiance forecast model and a PV model. The PV power forecast is obtained from the irradiance forecast using the PV model. The proposed irradiance forecast model is based on multiple feed-forward neural networks. The global horizontal irradiance forecast has a mean absolute percentage error of 3.4% on a sunny day and 23% on a cloudy day for Stuttgart. PV power forecasts based on the neural network irradiance forecast have performed much better than the PV power persistence forecast model.展开更多
Neural networks(NNs),as one of the most robust and efficient machine learning methods,have been commonly used in solving several problems.However,choosing proper hyperparameters(e.g.the numbers of layers and neurons i...Neural networks(NNs),as one of the most robust and efficient machine learning methods,have been commonly used in solving several problems.However,choosing proper hyperparameters(e.g.the numbers of layers and neurons in each layer)has a significant influence on the accuracy of these methods.Therefore,a considerable number of studies have been carried out to optimize the NN hyperpaxameters.In this study,the genetic algorithm is applied to NN to find the optimal hyperpaxameters.Thus,the deep energy method,which contains a deep neural network,is applied first on a Timoshenko beam and a plate with a hole.Subsequently,the numbers of hidden layers,integration points,and neurons in each layer are optimized to reach the highest accuracy to predict the stress distribution through these structures.Thus,applying the proper optimization method on NN leads to significant increase in the NN prediction accuracy after conducting the optimization in various examples.展开更多
In this paper,an adaptive dynamic programming(ADP)strategy is investigated for discrete-time nonlinear systems with unknown nonlinear dynamics subject to input saturation.To save the communication resources between th...In this paper,an adaptive dynamic programming(ADP)strategy is investigated for discrete-time nonlinear systems with unknown nonlinear dynamics subject to input saturation.To save the communication resources between the controller and the actuators,stochastic communication protocols(SCPs)are adopted to schedule the control signal,and therefore the closed-loop system is essentially a protocol-induced switching system.A neural network(NN)-based identifier with a robust term is exploited for approximating the unknown nonlinear system,and a set of switch-based updating rules with an additional tunable parameter of NN weights are developed with the help of the gradient descent.By virtue of a novel Lyapunov function,a sufficient condition is proposed to achieve the stability of both system identification errors and the update dynamics of NN weights.Then,a value iterative ADP algorithm in an offline way is proposed to solve the optimal control of protocol-induced switching systems with saturation constraints,and the convergence is profoundly discussed in light of mathematical induction.Furthermore,an actor-critic NN scheme is developed to approximate the control law and the proposed performance index function in the framework of ADP,and the stability of the closed-loop system is analyzed in view of the Lyapunov theory.Finally,the numerical simulation results are presented to demonstrate the effectiveness of the proposed control scheme.展开更多
Wheeled mobile robots(WMRs) encounter unavoidable slippage especially on the low adhesion terrain such that the robots stability and accuracy are reduced greatly.To overcome this drawback,this article presents a neura...Wheeled mobile robots(WMRs) encounter unavoidable slippage especially on the low adhesion terrain such that the robots stability and accuracy are reduced greatly.To overcome this drawback,this article presents a neural network(NN) based terminal sliding mode control(TSMC) for WMRs where an augmented ground friction model is reported by which the uncertain friction can be estimated and compensated according to the required performance.In contrast to the existing friction models,the developed augmented ground friction model corresponds to actual fact because not only the effects associated with the mobile platform velocity but also the slippage related to the wheel slip rate are concerned simultaneously.Besides,the presented control approach can combine the merits of both TSMC and radial basis function(RBF) neural networks techniques,thereby providing numerous excellent performances for the closed-loop system,such as finite time convergence and faster friction estimation property.Simulation results validate the proposed friction model and robustness of controller;these research results will improve the autonomy and intelligence of WMRs,particularly when the mobile platform suffers from the sophisticated unstructured environment.展开更多
This paper develops a novel Neural Network(NN)-based adaptive nonsingular practical predefined-time controller for the hypersonic morphing aircraft subject to actuator faults. Firstly, a novel Lyapunov criterion of pr...This paper develops a novel Neural Network(NN)-based adaptive nonsingular practical predefined-time controller for the hypersonic morphing aircraft subject to actuator faults. Firstly, a novel Lyapunov criterion of practical predefined-time stability is established. Following the proposed criterion, a tangent function based nonsingular predefined-time sliding manifold and the control strategy are developed. Secondly, the radial basis function NN with a low-complexity adaptation mechanism is incorporated into the controller to tackle the actuator faults and uncertainties. Thirdly, rigorous theoretical proof reveals that the attitude tracking errors can converge to a small region around the origin within a predefined time, while all signals in the closed-loop system remain bounded. Finally, numerical simulation results are presented to verify the effectiveness and improved performance of the proposed control scheme.展开更多
Magnetically coupled rodless cylinders are widely used in the coordinate positioning of mechanical arms,electro-static paintings,and other industrial applications.However,they exhibit strong nonlinear characteristics,...Magnetically coupled rodless cylinders are widely used in the coordinate positioning of mechanical arms,electro-static paintings,and other industrial applications.However,they exhibit strong nonlinear characteristics,which lead to low servo control accuracy.In this study,a mass-flow equation through the valve port was derived to improve the control performance,considering the characteristics of the dynamics and throttle-hole flow.Subsequently,a fric-tion model combining static,viscous,and Coulomb friction with a zero-velocity interval was proposed.In addition,energy and dynamic models were set for the experimental investigation of the magnetically coupled rodless cylin-der.A nonlinear mathematical model for the position of the magnetically coupled rodless cylinder was proposed.An incremental PID controller was designed for the magnetically coupled rodless cylinder to control this system,and the PID parameters were adjusted online using RBF neural network.The response results of the PID parameters based on the RBF neural network were compared with those of the traditional incremental PID control,which proved the superiority of the optimization control algorithm of the incremental PID parameters based on the RBF neural network servo control system.The experimental results of this model were compared with the simulation results.The average error between the established model and the actual system was 0.005175054(m),which was approximately 2.588%of the total travel length,demonstrating the accuracy of the theoretical model.展开更多
Dynamic neural network(NN)techniques are increasingly important because they facilitate deep learning techniques with more complex network architectures.However,existing studies,which predominantly optimize the static...Dynamic neural network(NN)techniques are increasingly important because they facilitate deep learning techniques with more complex network architectures.However,existing studies,which predominantly optimize the static computational graphs by static scheduling methods,usually focus on optimizing static neural networks in deep neural network(DNN)accelerators.We analyze the execution process of dynamic neural networks and observe that dynamic features introduce challenges for efficient scheduling and pipelining in existing DNN accelerators.We propose DyPipe,a holistic approach to optimizing dynamic neural network inferences in enhanced DNN accelerators.DyPipe achieves significant performance improvements for dynamic neural networks while it introduces negligible overhead for static neural networks.Our evaluation demonstrates that DyPipe achieves 1.7x speedup on dynamic neural networks and maintains more than 96%performance for static neural networks.展开更多
A neural network(NN) is a powerful tool for approximating bounded continuous functions in machine learning. The NN provides a framework for numerically solving ordinary differential equations(ODEs) and partial differe...A neural network(NN) is a powerful tool for approximating bounded continuous functions in machine learning. The NN provides a framework for numerically solving ordinary differential equations(ODEs) and partial differential equations(PDEs)combined with the automatic differentiation(AD) technique. In this work, we explore the use of NN for the function approximation and propose a universal solver for ODEs and PDEs. The solver is tested for initial value problems and boundary value problems of ODEs, and the results exhibit high accuracy for not only the unknown functions but also their derivatives. The same strategy can be used to construct a PDE solver based on collocation points instead of a mesh, which is tested with the Burgers equation and the heat equation(i.e., the Laplace equation).展开更多
Reinforcement Learning(RL)algorithms work well with well-defined rewards,but they fail with sparse/deceptive rewards and require additional exploration strategies.This work introduces a deep exploration method based o...Reinforcement Learning(RL)algorithms work well with well-defined rewards,but they fail with sparse/deceptive rewards and require additional exploration strategies.This work introduces a deep exploration method based on the Upper Confidence Bound(UCB)bonus.The proposed method can be plugged into actor-critic algorithms that use deep neural networks as a critic.Based on the conclusion of the regret bound under the linear Markov decision process approximation,we use the feature matrix to calculate the UCB bonus for deep exploration.The proposed method is equivalent to the count-based exploration method in special cases and is suitable for general situations.Our method uses the last d-dimensional feature vector in the critic network and is easy to deploy.We design a simple task,“swim”,to demonstrate the principle of the proposed method to achieve exploration in sparse/deceptive reward environments.Then,we perform an empirical evaluation on sparse/deceptive reward version gym environments and Ackermann robot control tasks.The evaluation results verify that the proposed algorithm can perform effective deep explorations in sparse/deceptive reward tasks.展开更多
Classification of speech signals is a vital part of speech signal processing systems.With the advent of speech coding and synthesis,the classification of the speech signal is made accurate and faster.Conventional meth...Classification of speech signals is a vital part of speech signal processing systems.With the advent of speech coding and synthesis,the classification of the speech signal is made accurate and faster.Conventional methods are considered inaccurate due to the uncertainty and diversity of speech signals in the case of real speech signal classification.In this paper,we use efficient speech signal classification using a series of neural network classifiers with reinforcement learning operations.Prior classification of speech signals,the study extracts the essential features from the speech signal using Cepstral Analysis.The features are extracted by converting the speech waveform to a parametric representation to obtain a relatively minimized data rate.Hence to improve the precision of classification,Generative Adversarial Networks are used and it tends to classify the speech signal after the extraction of features from the speech signal using the cepstral coefficient.The classifiers are trained with these features initially and the best classifier is chosen to perform the task of classification on new datasets.The validation of testing sets is evaluated using RL that provides feedback to Classifiers.Finally,at the user interface,the signals are played by decoding the signal after being retrieved from the classifier back based on the input query.The results are evaluated in the form of accuracy,recall,precision,f-measure,and error rate,where generative adversarial network attains an increased accuracy rate than other methods:Multi-Layer Perceptron,Recurrent Neural Networks,Deep belief Networks,and Convolutional Neural Networks.展开更多
Short-term forecast of urban traffic flow is very important to intelligent transportation. Although the conventional methods have some advantages, to some extent, in improving the traffic forecast's precision, it ...Short-term forecast of urban traffic flow is very important to intelligent transportation. Although the conventional methods have some advantages, to some extent, in improving the traffic forecast's precision, it is still hard to achieve high accuracy. In this paper, we propose a short-term traffic flow forecasting method, which is based on the hybrid particle swarm optimization-neural network(HPSO-NN) with error compensation mechanism.In HPSO-NN, the hybrid PSO algorithm is employed to train the structures and parameters of the feed-forward advanced neural network, while the error compensation mechanism is employed to improve the accuracy. HPSONN is used to forecast the vehicle velocity in Shanghai North-South Viaduct. Experimental results show that the HPSO-NN, compared with the auto-regressive and moving average(ARMA) model, can forecast traffic flow with a higher accuracy. What's more, we have also found that HPSO-NN with error compensation mechanism has better performance than that of HPSO-NN alone.展开更多
Conventional direct torque control (DTC) is one of the excellent control strategies available to control the torque of the induction machine (IM). However, the low switching frequency of the DTC causes high ripples in...Conventional direct torque control (DTC) is one of the excellent control strategies available to control the torque of the induction machine (IM). However, the low switching frequency of the DTC causes high ripples in the flux and torque that leads to an acoustic noise which degrades the control performances, especially at low speeds. Many direct torque control techniques were appeared to remedy these problems by focusing specifically on the torque and flux. In this paper, a state of the art review of various modern techniques for improving the performance of DTC control is presented. The objective is to make a critical analysis of these methods in terms of ripples reduction, tracking speed, switching loss, algorithm complexity and parameter sensitivity. Further, it is envisaged that the information presented in this review paper will be a valuable gathering of information for academic and industrial researchers.展开更多
This study proposes a scheme for state estimation and,consequently,fault diagnosis in nonlinear systems.Initially,an optimal nonlinear observer is designed for nonlinear systems subject to an actuator or plant fault.B...This study proposes a scheme for state estimation and,consequently,fault diagnosis in nonlinear systems.Initially,an optimal nonlinear observer is designed for nonlinear systems subject to an actuator or plant fault.By utilizing Lyapunov's direct method,the observer is proved to be optimal with respect to a performance function,including the magnitude of the observer gain and the convergence time.The observer gain is obtained by using approximation of Hamilton-Jacobi-Bellman(HJB)equation.The approximation is determined via an online trained neural network(NN).Next a class of affine nonlinear systems is considered which is subject to unknown disturbances in addition to fault signals.In this case,for each fault the original system is transformed to a new form in which the proposed optimal observer can be applied for state estimation and fault detection and isolation(FDI).Simulation results of a singlelink flexible joint robot(SLFJR)electric drive system show the effectiveness of the proposed methodology.展开更多
Recent years,neural networks(NNs)have received increasing attention from both academia and industry.So far significant diversity among existing NNs as well as their hardware platforms makes NN programming a daunting t...Recent years,neural networks(NNs)have received increasing attention from both academia and industry.So far significant diversity among existing NNs as well as their hardware platforms makes NN programming a daunting task.In this paper,a domain-specific language(DSL)for NNs,neural network language(NNL)is proposed to deliver productivity of NN programming and portable performance of NN execution on different hardware platforms.The productivity and flexibility of NN programming are enabled by abstracting NNs as a directed graph of blocks.The language describes 4 representative and widely used NNs and runs them on 3 different hardware platforms(CPU,GPU and NN accelerator).Experimental results show that NNs written with the proposed language are,on average,14.5%better than the baseline implementations across these 3 platforms.Moreover,compared with the Caffe framework that specifically targets the GPU platform,the code can achieve similar performance.展开更多
This paper studies an online iterative algorithm for solving discrete-time multi-agent dynamic graphical games with input constraints.In order to obtain the optimal strategy of each agent,it is necessary to solve a se...This paper studies an online iterative algorithm for solving discrete-time multi-agent dynamic graphical games with input constraints.In order to obtain the optimal strategy of each agent,it is necessary to solve a set of coupled Hamilton-Jacobi-Bellman(HJB)equations.It is very difficult to solve HJB equations by the traditional method.The relevant game problem will become more complex if the control input of each agent in the dynamic graphical game is constrained.In this paper,an online iterative algorithm is proposed to find the online solution to dynamic graphical game without the need for drift dynamics of agents.Actually,this algorithm is to find the optimal solution of Bellman equations online.This solution employs a distributed policy iteration process,using only the local information available to each agent.It can be proved that under certain conditions,when each agent updates its own strategy simultaneously,the whole multi-agent system will reach Nash equilibrium.In the process of algorithm implementation,for each agent,two layers of neural networks are used to fit the value function and control strategy,respectively.Finally,a simulation example is given to show the effectiveness of our method.展开更多
This paper considers a massive single-input multiple-output(SIMO)system,where multiple singleantenna transmitters simultaneously communicate with a receiver equipped with a large number of antennas.Different from the ...This paper considers a massive single-input multiple-output(SIMO)system,where multiple singleantenna transmitters simultaneously communicate with a receiver equipped with a large number of antennas.Different from the conventional noncoherent transceivers which require a certain level of the statistical information on the channel fading,we propose a joint transceiver design method based on machine learning,requiring a limited number of channel realizations.In the proposed method,the multiple transmitters,the channel,and the receiver are represented with a deep neural network(NN),and an autoencoder is adopted to minimize the end-to-end transmission error probability.Besides,the relationship between the number of training samples and the transmission error probability is analyzed based on the confidence interval method.Simulation results show that the proposed NN-based transceiver achieves lower transmission error probability in typical scenarios,and is more robust against the channel parameters variation compared with the existing methods.展开更多
The problem of adaptive stabilization is addressed for a class of uncertain stochastic nonlinear strict-feedback systems with both unknown dead-zone and unknown gain functions.By using the backstepping method and neur...The problem of adaptive stabilization is addressed for a class of uncertain stochastic nonlinear strict-feedback systems with both unknown dead-zone and unknown gain functions.By using the backstepping method and neural network（NN） parameterization,a novel adaptive neural control scheme which contains fewer learning parameters is developed to solve the stabilization problem of such systems.Meanwhile,stability analysis is presented to guarantee that all the error variables are semi-globally uniformly ultimately bounded with desired probability in a compact set.The effectiveness of the proposed design is illustrated by simulation results.展开更多
基金supported in part by the National Key Reseanch and Development Program of China(2018AAA0101502,2018YFB1702300)in part by the National Natural Science Foundation of China(61722312,61533019,U1811463,61533017)in part by the Intel Collaborative Research Institute for Intelligent and Automated Connected Vehicles。
文摘This paper studies the problem of optimal parallel tracking control for continuous-time general nonlinear systems.Unlike existing optimal state feedback control,the control input of the optimal parallel control is introduced into the feedback system.However,due to the introduction of control input into the feedback system,the optimal state feedback control methods can not be applied directly.To address this problem,an augmented system and an augmented performance index function are proposed firstly.Thus,the general nonlinear system is transformed into an affine nonlinear system.The difference between the optimal parallel control and the optimal state feedback control is analyzed theoretically.It is proven that the optimal parallel control with the augmented performance index function can be seen as the suboptimal state feedback control with the traditional performance index function.Moreover,an adaptive dynamic programming(ADP)technique is utilized to implement the optimal parallel tracking control using a critic neural network(NN)to approximate the value function online.The stability analysis of the closed-loop system is performed using the Lyapunov theory,and the tracking error and NN weights errors are uniformly ultimately bounded(UUB).Also,the optimal parallel controller guarantees the continuity of the control input under the circumstance that there are finite jump discontinuities in the reference signals.Finally,the effectiveness of the developed optimal parallel control method is verified in two cases.
基金supported in part by the Science and Technology development fund(FDCT)of Macao(011/2017/A)the National Natural Science Foundation of China(61803397)。
文摘Smart manufacturing refers to optimization techniques that are implemented in production operations by utilizing advanced analytics approaches. With the widespread increase in deploying industrial internet of things(IIOT) sensors in manufacturing processes, there is a progressive need for optimal and effective approaches to data management.Embracing machine learning and artificial intelligence to take advantage of manufacturing data can lead to efficient and intelligent automation. In this paper, we conduct a comprehensive analysis based on evolutionary computing and neural network algorithms toward making semiconductor manufacturing smart.We propose a dynamic algorithm for gaining useful insights about semiconductor manufacturing processes and to address various challenges. We elaborate on the utilization of a genetic algorithm and neural network to propose an intelligent feature selection algorithm. Our objective is to provide an advanced solution for controlling manufacturing processes and to gain perspective on various dimensions that enable manufacturers to access effective predictive technologies.
基金supported by the National Natural Science Foundation of China(11502288)the Natural Science Foundation of Hunan Province(2016JJ3019)+1 种基金the Aeronautical Science Foundation of China(2017ZA88001)the Scientific Research Project of National University of Defense Technology(ZK17-03-32)
文摘A neural-network-based adaptive gain scheduling backstepping sliding mode control（NNAGS-BSMC） approach for a class of uncertain strict-feedback nonlinear system is proposed.First, the control problem of uncertain strict-feedback nonlinear systems is formulated. Second, the detailed design of NNAGSBSMC is described. The sliding mode control（SMC） law is designed to track a referenced output via backstepping technique.To decrease chattering result from SMC, a radial basis function neural network（RBFNN） is employed to construct the NNAGSBSMC to facilitate adaptive gain scheduling, in which the gains are scheduled adaptively via neural network（NN）, with sliding surface and its differential as NN inputs and the gains as NN outputs. Finally, the verification example is given to show the effectiveness and robustness of the proposed approach. Contrasting simulation results indicate that the NNAGS-BSMC decreases the chattering effectively and has better control performance against the BSMC.
文摘In order to develop predictive control algorithms for efficient energy management and monitoring for residential grid connected photovoltaic systems, accurate and reliable photovoltaic(PV) power forecasts are required.A PV yield prediction system is presented based on an irradiance forecast model and a PV model. The PV power forecast is obtained from the irradiance forecast using the PV model. The proposed irradiance forecast model is based on multiple feed-forward neural networks. The global horizontal irradiance forecast has a mean absolute percentage error of 3.4% on a sunny day and 23% on a cloudy day for Stuttgart. PV power forecasts based on the neural network irradiance forecast have performed much better than the PV power persistence forecast model.
文摘Neural networks(NNs),as one of the most robust and efficient machine learning methods,have been commonly used in solving several problems.However,choosing proper hyperparameters(e.g.the numbers of layers and neurons in each layer)has a significant influence on the accuracy of these methods.Therefore,a considerable number of studies have been carried out to optimize the NN hyperpaxameters.In this study,the genetic algorithm is applied to NN to find the optimal hyperpaxameters.Thus,the deep energy method,which contains a deep neural network,is applied first on a Timoshenko beam and a plate with a hole.Subsequently,the numbers of hidden layers,integration points,and neurons in each layer are optimized to reach the highest accuracy to predict the stress distribution through these structures.Thus,applying the proper optimization method on NN leads to significant increase in the NN prediction accuracy after conducting the optimization in various examples.
基金supported in part by the Australian Research Council Discovery Early Career Researcher Award(DE200101128)Australian Research Council(DP190101557)。
文摘In this paper,an adaptive dynamic programming(ADP)strategy is investigated for discrete-time nonlinear systems with unknown nonlinear dynamics subject to input saturation.To save the communication resources between the controller and the actuators,stochastic communication protocols(SCPs)are adopted to schedule the control signal,and therefore the closed-loop system is essentially a protocol-induced switching system.A neural network(NN)-based identifier with a robust term is exploited for approximating the unknown nonlinear system,and a set of switch-based updating rules with an additional tunable parameter of NN weights are developed with the help of the gradient descent.By virtue of a novel Lyapunov function,a sufficient condition is proposed to achieve the stability of both system identification errors and the update dynamics of NN weights.Then,a value iterative ADP algorithm in an offline way is proposed to solve the optimal control of protocol-induced switching systems with saturation constraints,and the convergence is profoundly discussed in light of mathematical induction.Furthermore,an actor-critic NN scheme is developed to approximate the control law and the proposed performance index function in the framework of ADP,and the stability of the closed-loop system is analyzed in view of the Lyapunov theory.Finally,the numerical simulation results are presented to demonstrate the effectiveness of the proposed control scheme.
基金supported by the National Natural Science Foundation of China(61573078,61573147)the International S&T Cooperation Program of China(2014DFB70120)the State Key Laboratory of Robotics and System(SKLRS2015ZD06)
文摘Wheeled mobile robots(WMRs) encounter unavoidable slippage especially on the low adhesion terrain such that the robots stability and accuracy are reduced greatly.To overcome this drawback,this article presents a neural network(NN) based terminal sliding mode control(TSMC) for WMRs where an augmented ground friction model is reported by which the uncertain friction can be estimated and compensated according to the required performance.In contrast to the existing friction models,the developed augmented ground friction model corresponds to actual fact because not only the effects associated with the mobile platform velocity but also the slippage related to the wheel slip rate are concerned simultaneously.Besides,the presented control approach can combine the merits of both TSMC and radial basis function(RBF) neural networks techniques,thereby providing numerous excellent performances for the closed-loop system,such as finite time convergence and faster friction estimation property.Simulation results validate the proposed friction model and robustness of controller;these research results will improve the autonomy and intelligence of WMRs,particularly when the mobile platform suffers from the sophisticated unstructured environment.
基金supported by the National Natural Science Foundation of China (Nos. 52233014, U2241215)。
文摘This paper develops a novel Neural Network(NN)-based adaptive nonsingular practical predefined-time controller for the hypersonic morphing aircraft subject to actuator faults. Firstly, a novel Lyapunov criterion of practical predefined-time stability is established. Following the proposed criterion, a tangent function based nonsingular predefined-time sliding manifold and the control strategy are developed. Secondly, the radial basis function NN with a low-complexity adaptation mechanism is incorporated into the controller to tackle the actuator faults and uncertainties. Thirdly, rigorous theoretical proof reveals that the attitude tracking errors can converge to a small region around the origin within a predefined time, while all signals in the closed-loop system remain bounded. Finally, numerical simulation results are presented to verify the effectiveness and improved performance of the proposed control scheme.
基金Supported by Outstanding Young Scientists in Beijing of China(Grant No.BJJWZYJH01201910006021)Open Foundation of the State Key Laboratory of Fluid Power and Mechatronic Systems of China(Grant No.GZKF-202016)+2 种基金Henan Provincial Science and Technology Key Project of China(Grant Nos.202102210081,212102210050)Sub Project of Strengthening Key Basic Research Projects in the Basic Plan of the Science and Technology Commission of the Central Military Commission of China(Grant No.2019-JCJQ-ZD-120-13)Henan Provincial Fundamental Research Funds for the Universities of China(Grant No.NSFRF200403).
文摘Magnetically coupled rodless cylinders are widely used in the coordinate positioning of mechanical arms,electro-static paintings,and other industrial applications.However,they exhibit strong nonlinear characteristics,which lead to low servo control accuracy.In this study,a mass-flow equation through the valve port was derived to improve the control performance,considering the characteristics of the dynamics and throttle-hole flow.Subsequently,a fric-tion model combining static,viscous,and Coulomb friction with a zero-velocity interval was proposed.In addition,energy and dynamic models were set for the experimental investigation of the magnetically coupled rodless cylin-der.A nonlinear mathematical model for the position of the magnetically coupled rodless cylinder was proposed.An incremental PID controller was designed for the magnetically coupled rodless cylinder to control this system,and the PID parameters were adjusted online using RBF neural network.The response results of the PID parameters based on the RBF neural network were compared with those of the traditional incremental PID control,which proved the superiority of the optimization control algorithm of the incremental PID parameters based on the RBF neural network servo control system.The experimental results of this model were compared with the simulation results.The average error between the established model and the actual system was 0.005175054(m),which was approximately 2.588%of the total travel length,demonstrating the accuracy of the theoretical model.
基金supported by the Beijing Natural Science Foundation under Grant No.JQ18013the National Natural Science Foundation of China under Grant Nos.61925208,61732007,61732002 and 61906179+1 种基金the Strategic Priority Research Program of Chinese Academy of Sciences(CAS)under Grant No.XDB32050200the Youth Innovation Promotion Association CAS,Beijing Academy of Artificial Intelligence(BAAI)and Xplore Prize.
文摘Dynamic neural network(NN)techniques are increasingly important because they facilitate deep learning techniques with more complex network architectures.However,existing studies,which predominantly optimize the static computational graphs by static scheduling methods,usually focus on optimizing static neural networks in deep neural network(DNN)accelerators.We analyze the execution process of dynamic neural networks and observe that dynamic features introduce challenges for efficient scheduling and pipelining in existing DNN accelerators.We propose DyPipe,a holistic approach to optimizing dynamic neural network inferences in enhanced DNN accelerators.DyPipe achieves significant performance improvements for dynamic neural networks while it introduces negligible overhead for static neural networks.Our evaluation demonstrates that DyPipe achieves 1.7x speedup on dynamic neural networks and maintains more than 96%performance for static neural networks.
基金Project supported by the National Natural Science Foundation of China(No.11521091)
文摘A neural network(NN) is a powerful tool for approximating bounded continuous functions in machine learning. The NN provides a framework for numerically solving ordinary differential equations(ODEs) and partial differential equations(PDEs)combined with the automatic differentiation(AD) technique. In this work, we explore the use of NN for the function approximation and propose a universal solver for ODEs and PDEs. The solver is tested for initial value problems and boundary value problems of ODEs, and the results exhibit high accuracy for not only the unknown functions but also their derivatives. The same strategy can be used to construct a PDE solver based on collocation points instead of a mesh, which is tested with the Burgers equation and the heat equation(i.e., the Laplace equation).
文摘Reinforcement Learning(RL)algorithms work well with well-defined rewards,but they fail with sparse/deceptive rewards and require additional exploration strategies.This work introduces a deep exploration method based on the Upper Confidence Bound(UCB)bonus.The proposed method can be plugged into actor-critic algorithms that use deep neural networks as a critic.Based on the conclusion of the regret bound under the linear Markov decision process approximation,we use the feature matrix to calculate the UCB bonus for deep exploration.The proposed method is equivalent to the count-based exploration method in special cases and is suitable for general situations.Our method uses the last d-dimensional feature vector in the critic network and is easy to deploy.We design a simple task,“swim”,to demonstrate the principle of the proposed method to achieve exploration in sparse/deceptive reward environments.Then,we perform an empirical evaluation on sparse/deceptive reward version gym environments and Ackermann robot control tasks.The evaluation results verify that the proposed algorithm can perform effective deep explorations in sparse/deceptive reward tasks.
文摘Classification of speech signals is a vital part of speech signal processing systems.With the advent of speech coding and synthesis,the classification of the speech signal is made accurate and faster.Conventional methods are considered inaccurate due to the uncertainty and diversity of speech signals in the case of real speech signal classification.In this paper,we use efficient speech signal classification using a series of neural network classifiers with reinforcement learning operations.Prior classification of speech signals,the study extracts the essential features from the speech signal using Cepstral Analysis.The features are extracted by converting the speech waveform to a parametric representation to obtain a relatively minimized data rate.Hence to improve the precision of classification,Generative Adversarial Networks are used and it tends to classify the speech signal after the extraction of features from the speech signal using the cepstral coefficient.The classifiers are trained with these features initially and the best classifier is chosen to perform the task of classification on new datasets.The validation of testing sets is evaluated using RL that provides feedback to Classifiers.Finally,at the user interface,the signals are played by decoding the signal after being retrieved from the classifier back based on the input query.The results are evaluated in the form of accuracy,recall,precision,f-measure,and error rate,where generative adversarial network attains an increased accuracy rate than other methods:Multi-Layer Perceptron,Recurrent Neural Networks,Deep belief Networks,and Convolutional Neural Networks.
基金the National Natural Science Foundation of China(Nos.71361130012 and 61221003)the Specialized Research Fund for the Doctoral Program of Higher Education(No.20120073110017)
文摘Short-term forecast of urban traffic flow is very important to intelligent transportation. Although the conventional methods have some advantages, to some extent, in improving the traffic forecast's precision, it is still hard to achieve high accuracy. In this paper, we propose a short-term traffic flow forecasting method, which is based on the hybrid particle swarm optimization-neural network(HPSO-NN) with error compensation mechanism.In HPSO-NN, the hybrid PSO algorithm is employed to train the structures and parameters of the feed-forward advanced neural network, while the error compensation mechanism is employed to improve the accuracy. HPSONN is used to forecast the vehicle velocity in Shanghai North-South Viaduct. Experimental results show that the HPSO-NN, compared with the auto-regressive and moving average(ARMA) model, can forecast traffic flow with a higher accuracy. What's more, we have also found that HPSO-NN with error compensation mechanism has better performance than that of HPSO-NN alone.
文摘Conventional direct torque control (DTC) is one of the excellent control strategies available to control the torque of the induction machine (IM). However, the low switching frequency of the DTC causes high ripples in the flux and torque that leads to an acoustic noise which degrades the control performances, especially at low speeds. Many direct torque control techniques were appeared to remedy these problems by focusing specifically on the torque and flux. In this paper, a state of the art review of various modern techniques for improving the performance of DTC control is presented. The objective is to make a critical analysis of these methods in terms of ripples reduction, tracking speed, switching loss, algorithm complexity and parameter sensitivity. Further, it is envisaged that the information presented in this review paper will be a valuable gathering of information for academic and industrial researchers.
文摘This study proposes a scheme for state estimation and,consequently,fault diagnosis in nonlinear systems.Initially,an optimal nonlinear observer is designed for nonlinear systems subject to an actuator or plant fault.By utilizing Lyapunov's direct method,the observer is proved to be optimal with respect to a performance function,including the magnitude of the observer gain and the convergence time.The observer gain is obtained by using approximation of Hamilton-Jacobi-Bellman(HJB)equation.The approximation is determined via an online trained neural network(NN).Next a class of affine nonlinear systems is considered which is subject to unknown disturbances in addition to fault signals.In this case,for each fault the original system is transformed to a new form in which the proposed optimal observer can be applied for state estimation and fault detection and isolation(FDI).Simulation results of a singlelink flexible joint robot(SLFJR)electric drive system show the effectiveness of the proposed methodology.
基金the National Key Research and Development Program of China(No.2017YFA0700902,2017YFB1003101)the National Natural Science Foundation of China(No.61472396,61432016,61473275,61522211,61532016,61521092,61502446,61672491,61602441,61602446,61732002,61702478)+3 种基金the 973 Program of China(No.2015CB358800)National Science and Technology Major Project(No.2018ZX01031102)the Transformation and Transfer of Scientific and Technological Achievements of Chinese Academy of Sciences(No.KFJ-HGZX-013)Strategic Priority Research Program of Chinese Academy of Sciences(No.XDBS01050200).
文摘Recent years,neural networks(NNs)have received increasing attention from both academia and industry.So far significant diversity among existing NNs as well as their hardware platforms makes NN programming a daunting task.In this paper,a domain-specific language(DSL)for NNs,neural network language(NNL)is proposed to deliver productivity of NN programming and portable performance of NN execution on different hardware platforms.The productivity and flexibility of NN programming are enabled by abstracting NNs as a directed graph of blocks.The language describes 4 representative and widely used NNs and runs them on 3 different hardware platforms(CPU,GPU and NN accelerator).Experimental results show that NNs written with the proposed language are,on average,14.5%better than the baseline implementations across these 3 platforms.Moreover,compared with the Caffe framework that specifically targets the GPU platform,the code can achieve similar performance.
基金supported by the National Natural Science Foundation of China(Nos.61773241,61973183)the Shandong Provincial Natural Science Foundation(No.ZR2019MF041).
文摘This paper studies an online iterative algorithm for solving discrete-time multi-agent dynamic graphical games with input constraints.In order to obtain the optimal strategy of each agent,it is necessary to solve a set of coupled Hamilton-Jacobi-Bellman(HJB)equations.It is very difficult to solve HJB equations by the traditional method.The relevant game problem will become more complex if the control input of each agent in the dynamic graphical game is constrained.In this paper,an online iterative algorithm is proposed to find the online solution to dynamic graphical game without the need for drift dynamics of agents.Actually,this algorithm is to find the optimal solution of Bellman equations online.This solution employs a distributed policy iteration process,using only the local information available to each agent.It can be proved that under certain conditions,when each agent updates its own strategy simultaneously,the whole multi-agent system will reach Nash equilibrium.In the process of algorithm implementation,for each agent,two layers of neural networks are used to fit the value function and control strategy,respectively.Finally,a simulation example is given to show the effectiveness of our method.
基金The work was supported in part by the Key Area R&D Program of Guangdong Province with Grant No.2018B030338001by the National Key R&D Program of China with Grant No.2018YFB1800800+2 种基金y Natural Science Foundation of China with grant NSFC-61629101by Guangdong Research Project No.2017ZT07X152by Shenzhen Key Lab Fund No.ZDSYS201707251409055.
文摘This paper considers a massive single-input multiple-output(SIMO)system,where multiple singleantenna transmitters simultaneously communicate with a receiver equipped with a large number of antennas.Different from the conventional noncoherent transceivers which require a certain level of the statistical information on the channel fading,we propose a joint transceiver design method based on machine learning,requiring a limited number of channel realizations.In the proposed method,the multiple transmitters,the channel,and the receiver are represented with a deep neural network(NN),and an autoencoder is adopted to minimize the end-to-end transmission error probability.Besides,the relationship between the number of training samples and the transmission error probability is analyzed based on the confidence interval method.Simulation results show that the proposed NN-based transceiver achieves lower transmission error probability in typical scenarios,and is more robust against the channel parameters variation compared with the existing methods.
基金supported by the National Natural Science Foundation of China (60704013)the Special Foundation of East China University of Science and Technology for Youth Teacher (YH0157134)
文摘The problem of adaptive stabilization is addressed for a class of uncertain stochastic nonlinear strict-feedback systems with both unknown dead-zone and unknown gain functions.By using the backstepping method and neural network（NN） parameterization,a novel adaptive neural control scheme which contains fewer learning parameters is developed to solve the stabilization problem of such systems.Meanwhile,stability analysis is presented to guarantee that all the error variables are semi-globally uniformly ultimately bounded with desired probability in a compact set.The effectiveness of the proposed design is illustrated by simulation results.
