Adaptive neural network feedback control of a passive line-of-sight stabilization system

An adaptive neural network full-state feedback controller has been designed and applied to the passive line-of-sight (LOS) stabilization system. Model reference adaptive control (MRAC) is well established for linear systems. However, this method cannot be utilized directly since the LOS system is nonlinear in nature. Utilizing the universal approximation property of neural networks, an adaptive neural network controller is presented by generalizing the model reference adaptive control technique, in which the gains of the controller are approximated by neural networks. This removes the requirement of linearizing the dynamics of the system, and the stability properties of the closed-loop system can be guaranteed.     

Robust Finite-Time H ∞ Control for Nonlinear Jump Systems via Neural Networks

This paper presents a neural network-based robust finite-time H _∞ control design approach for a class of nonlinear Markov jump systems (MJSs). The system under consideration is subject to norm bounded parameter uncertainties and external disturbance. In the proposed framework, the nonlinearities are initially approximated by multilayer feedback neural networks. Subsequently, the neural networks undergo piecewise interpolation to generate a linear differential inclusion model. Then, based on the model, a robust finite-time state-feedback controller is designed such that the nonlinear MJS is finite-time bounded and finite-time stabilizable. The H _∞ control is specified to ensure the elimination of the approximation errors and external disturbances with a desired level. The controller gains can be derived by solving a set of linear matrix inequalities. Finally, simulation results are given to illustrate the effectiveness of the developed theoretic results.     

Robust Smooth-Trajectory Control of Nonlinear Servo Systems Based on Neural Networks

The electromagnetic torque introduces ripples into the electromechanical motion system due to nonlinearities, such as uncertain changes of magnet field, load, and friction, which generate speed oscillations and deteriorate the tracking performance of servo system. Furthermore, the minimum time response and smooth trajectory tracking are cruces in servo control. In this paper, an available method is proposed to solve them by using neural networks (NNs) and a nonlinear smooth trajectory filter (STF) for the robust smoothing feedforward control of a class of general nonlinear systems. First, the online weight-tuning scheme based on Lyapunov function can guarantee the boundedness of tracking error by good performance of NNs modeling nonlinear functions. Second, a feedforward controller based on the output of nonlinear STF is designed to guarantee minimum time response and smooth trajectory tracking. Finally, as a example, the motion system can be equivalent to the two-order system under the linear closed-loop current control in view of the (d,q) mathematic model for PM synchronous motor, so that this robust smoothing control method using neutral networks can be applied into position servo control. Moreover, the validity and effectiveness of this control method are verified through simulations and experiments     

An MRAC-based nonlinear speed control of an interior PM synchronous motor with improved maximum torque operation

A model reference adaptive control (MRAC)-based nonlinear speed control strategy of an interior permanent magnet (IPM) synchronous motor with an improved maximum torque operation is presented. In most servo systems, the controller is designed under the assumption that the electrical dynamics are neglected by the field-oriented control. This requires a high-performance inner-loop current control strategy. However, the separate designs for a high-performance current regulator and a robust speed controller need considerable effort. To overcome this limitation, an MRAC-based nonlinear speed control strategy for the IPM synchronous motor is presented, considering the whole nonlinear dynamics. Nonlinear speed control is achieved by an input–output linearization scheme. This scheme, however, gives an unsatisfactory performance under the mismatch of the system parameters and load conditions. For the robust output response, the controller parameters are estimated by an MRAC technique in which the disturbance torque and flux linkage are estimated. The adaptation laws are derived from Lyapunov stability theory. In view of the drive efficiency, the motor has to provide the maximum torque for a given input. To drive the IPM synchronous motor under improved maximum torque operation, the estimated flux linkage is employed for the generation of the d-axis current command. The robustness and output performance of the proposed control scheme are verified through simulation results.     

A neural network control system with parallel adaptive enhancementsapplicable to nonlinear servomechanisms

In this paper, we present a technique for using an additional parallel neural network to provide adaptive enhancements to a basic fixed neural network-based nonlinear control system. This proposed parallel adaptive neural network control system is applicable to nonlinear dynamical systems of the type commonly encountered in many practical position control servomechanisms. Properties of the controller are discussed, and it is shown that if Gaussian radial basis function networks are used for the additional parallel neural network, uniformly stable adaptation is assured and the approximation error converges to zero asymptotically. In the paper, the effectiveness of the proposed parallel adaptive neural network control system is demonstrated in real-time implementation experiments for position control in a servomechanism with asymmetrical loading and changes in the load     

实现标量混沌信号同步控制的神经网络方法

本文给出了一种利用线性输出神经网络实现标量混沌信号同步控制的方法。该方法利用线性输出神经网络构造被控混沌系统的模型，并基于Lyapunov理论与非线性系统控制方法，设计出神经网络权值变化规律与非线性反馈控制器，使神经网络模型的标量输出能大范围同步于给定的标量混沌信号。理论分析与计算机模拟结果都证实了这种方法的有效性。     

Real-time parallel adaptive neural network control for nonlinear servomechanisms—an approach using direct adaptive techniques

In this paper, a parallel adaptive neural network control system applicable to nonlinear dynamical systems of the type commonly encountered in many practical position control servomechanisms is developed. The controller is based on the use of direct adaptive techniques and an approach of using an additional parallel neural network to provide adaptive enhancements to a basic fixed neural network-based nonlinear controller. Properties of the proposed new controller are discussed in the paper and it is shown that if Gaussian radial basis function networks are used for the additional parallel neural network, uniformly stable adaptation is assured and asymptotic tracking of the position reference signal is achieved. The effectiveness of the proposed adaptive neural network control system is demonstrated in real-time implementation experiments for position control in a servomechanism with asymmetrical loading and changes in the load.     

Linear and Nonlinear Control Techniques for a Three-Phase Three-Level NPC Boost Rectifier

This paper deals with three control techniques for a three-phase three-level neutral-point-clamped (NPC) boost rectifier to study their relative performance. Linear, nonlinear, and nonlinear model reference adaptive control (MRAC) methods are developed to control power factor (PF) and regulate output and neutral point voltages. These controllers are designed in Simulink and implemented in real time using the DS1104 DSP of dSPACE for validation on a 1.2-kW prototype of an NPC boost rectifier operating at 1.92 kHz. The performance of boost converter with three control methods has been investigated respectively in steady state in terms of line-current harmonic distortion, efficiency, and PF and during transients such as load steps, utility disturbances, reactive power control, and dc-bus voltage tracking behavior. The linear PI controllers are characterized by reduced complexity but poor performance, whereas the nonlinear control technique has improved the converter performance significantly, while nonlinear MRAC exhibits much better performance in a wide operating range     

具有改进反馈环的NF-GMDH网络及其在混沌预测中的应用

具有反馈环的GMDH(Group Method of Data Handling with a feedback loop简称为GMDH—FL)网络只有三层，结构简单；而模糊GMDH神经(Neurofuzzy GMDH简称为NF—GMDH)网络可以同时利用系统的数据信息和语言信息。本文结合这两种网络的优点，利用改进自组织策略，提出了具有反馈环的NF—GMDH网络(简称NF—GMDH—FL)。针对该网络的第二次及其以后循环训练中有大量冗余组合和计算的缺点，本文进一步研究了具有改进反馈环的NF—GMDH(NF-GMDH with improved feedback loop简称为NF—GMDH—IFL)网络，并将其应用于混沌时间序列预测。通过仿真研究，证明其网络训练速度快，建模精度高，有比NF—GMDH模型和NF—GMDH—FL模型更优良的性能。     

Fieldbus based distributed servo control using LonWorks/IP gateway/web servers

Virtual device network (VDN) is an intelligent integrated form of a device (control) network and IP network. If a fieldbus based distributed control system (DCS) is implemented on a global VDN, efficiency and flexibility can be significantly improved. In this study DCS implemented on the LonWorks/IP VDN is investigated with an example of model reference adaptive control (MRAC) of a geared DC motor. In order to compensate for the network induced uncertain time delay inherently present on VDN, the modified Smith predictor based internal model controller was integrated to MRAC (MRAC–IMC). The effectiveness of the proposed control scheme was tested through experiment. The proposed control scheme exhibited the robustness to noise and external disturbances and the good tracking performance around zero velocity point occurring due to overshoot and stick friction. Result of this study suggests that sophisticated servo control of dynamic systems is possible from a remote client PC on VDN by properly compensating the network delay.     

Adaptive on-line parameter identification of a steer-by-wire system

Adaptive compliance control strategy can be a significant advantage for control of steer-by-wire systems. Initially the method is proposed for robotic applications where the main concern is the interaction forces between the robot and its environment. There are several studies about cooperative working of a robot and a human. As long as the steering system is a part of the vehicle where driver interaction is involved, it is reasonable to think that compliance control strategies can be adapted to steer-by-wire systems. Compliance control is a model reference control (MRC) strategy where the measured external force/torque is used as an input to a reference model to calculate its output and where the real system is controlled appropriately to track the reference system output. If a sensor is available to measure the external force/torque, system parameters need not to be estimated. A constant gain feedback controller can be used in such a case. However, if the parameter variations of the system are not within certain bounds, a model reference adaptive controller (MRAC) is needed. In addition to this, examining the change in the dynamics of the system due to the compliance of the driver arms is not possible by direct MRAC, because the driver effect is considered as a disturbance in this strategy. Therefore, in this study, instead of estimating controller parameters using direct MRAC where the main concern is the tracking performance, it is considered to use indirect MRAC in which the system parameters are estimated to observe their variations in the presence of parametric uncertainty and disturbances and to further examine the change in the dynamics of the system due to the compliance of the driver arms forming a closed kinematic structure by constraining the steering wheel. Hence, a steer-by-wire experimental setup including driver interaction and vehicle directional control units has been developed and three well-known adaptive on-line estimation methods, which are output-error method, equation-error method and modified recursive least squares method are evaluated on the driver interaction unit. These three methods are compared in terms of computational complexity, convergence, stability and applicability to real vehicles.     

Performance benefits of hybrid control design for linear andnonlinear systems

This paper provides an overview of recent developments on design of hybrid controllers for continuous-time control systems that can be described by linear or nonlinear differential state equations. Hybrid controllers provide a generalization of classical feedback controllers for linear and nonlinear systems. The benefit of hybrid controllers, that they can be used to achieve closed-loop performance objectives that cannot be achieved using classical linear or nonlinear controllers, is emphasized. This paper introduces hybrid controllers in the form of a switching control architecture and provides a summary of recently developed control approaches that utilize this control architecture. We provide a conceptual framework for these results, identify limitations of the results, and discuss the current status of hybrid control design approaches     

Adaptive feedback linearization control of antilock braking systems using neural networks

Antilock braking systems are designed to control the wheel slip, such that the braking force is maximized and steerability is maintained during braking. However, the control of antilock braking systems is a challenging problem due to nonlinear braking dynamics and the uncertain and time-varying nature of the parameters. This paper presents an adaptive neural network-based hybrid controller for antilock braking systems. The hybrid controller is based on the well-known feedback linearization, combined with two feedforward neural networks that are proposed so as to learn the nonlinearities of the antilock braking system associated with feedback linearization controller. The adaptation law is derived based on the structure of the controller, using steepest descent gradient approach and backpropagation algorithm to adjust the networks weights. The weight adaptation is online and the stability of the proposed controller in the sense of Lyapunov is studied. Simulations are conducted to show the effectiveness of the proposed controller under various road conditions and parameter uncertainties.     

基于神经网络仿射非线性系统滑模自适应控制

研究了一类单输入单输出仿射非线性系统的自适应控制问题.采用反馈线性化方法设计控制器,用神经网络逼近系统中的未知非线性函数,并在神经网络权值的自适应律中引入权值误差的概念,以改善系统的动态性能.同时采用滑模控制方法设计补偿器,提高了系统的鲁棒性.理论分析及仿真结果表明,所设计的控制器,不仅能解决该系统的轨迹跟踪控制问题,...     

Learning algorithms for a class of neurofuzzy network and application

A class of neurofuzzy networks and a constructive, competition-based learning procedure is introduced. Given a set of training data, the learning procedure automatically adjusts the input space portion to cover the whole space and finds membership functions parameters for each input variable. The network processes data following fuzzy reasoning principles and, due to its structure, it is dual to a rule-based fuzzy inference system. The neurofuzzy model is used to forecast seasonal streamflow, a key step to plan and operate hydroelectric power plants and to price energy. A database of average monthly inflows of three Brazilian hydroelectric plants located at different river basins was used as source of training and test data. The performance of the neurofuzzy network is compared with period regression, a standard approach used by the electric power industry to forecast streamflows. Comparisons with multilayer perceptron, radial basis network and adaptive neural-fuzzy inference system are also included. The results show that the neurofuzzy network provides better one-step-ahead streamflow forecasting, with forecasting errors significantly lower than the other approaches.     

Network-Based Robust$H_infty$Filtering for Uncertain Linear Systems

The problem of network-based robust$H_infty$filtering for uncertain linear systems is investigated. Different from the design of the traditional filter, the effects of the network-induced delay and data dropout on the performance of a filtering-error system are considered. The derived criteria for$H_infty$performance analysis of the filtering-error system and filter design are expressed as a set of linear matrix inequalities, which can be solved by using convex optimization method. Numerical examples show the effectiveness of the design method.     

Model Reference Adaptive Control Design for a Shunt Active-Power-Filter System

In this paper, model reference adaptive control (MRAC) is proposed for a single-phase shunt active power filter (APF) to improve line power factor and to reduce line current harmonics. The proposed APF controller forces the supply current to be sinusoidal, with low current harmonics, and to be in phase with the line voltage. The advantages of using MRAC over conventional proportional-integral control are its flexibility, adaptability, and robustness; moreover, MRAC can self-tune the controller gains to assure system stability. Since the APF is a bilinear system, it is hard to design the controller. This paper will solve the stability problem when a linearization method is used to solve the nonlinearity of the system. Moreover, by using Lyapunov's stability theory and Barbalat's lemma, an adaptive law is designed to guarantee an asymptotic output tracking of the system. To verify the proposed APF system, a digital signal controller (dsPIC30F4012) is adopted to implement the algorithm of MRAC, and a 1-kVA laboratory prototype is built to test feasibility. Experimental results are provided to verify the performance of the proposed APF system.     

Optimised PWM inverter control using artificial neural networks

A novel method of producing optimum switching functions for the voltage and harmonic control of DC-to-AC bridge inverters using neural networks is presented. Results obtained from an experimental implementation of a neural network-based inverter system are included. The implementation does not depend on any hardware configuration and can be modified without affecting the performance     

Neural network-based estimation of power electronic waveforms

Artificial neural network techniques are indicating a lot of promise for application in power electronic systems. So far, these applications are mainly confined to control, identification, and diagnostic problems, but the application in estimation is fairly new. The paper explores the application of neural networks for estimation of power electronic waveforms. The distorted line current waveforms in a single-phase thyristor AC controller and a three-phase diode rectifier that feeds an inverter-machine load have been taken into consideration, and neural networks have been trained to estimate the total RMS current, fundamental RMS current, displacement factor, and power factor. The performance of the neural network-based estimators has been compared with the actual values, and excellent performance is indicated. Neural network-based estimation has the usual advantages of very fast and simultaneous response of all the outputs, noise, and fault-tolerant performance and can be easily implemented in dedicated analog or digital hardware chips, which can coexist with digital signal processor (DSP) and/or application-specific integrated circuit (ASIC) chips. The estimation techniques can be extended to more complex waveforms in power electronics     

Performance-Improved Design of N-PID Controlled Motion Systems With Applications to Wafer Stages

A nonlinear filter design is proposed to improve nanopositioning servo performances in high-speed (and generally linear) motion systems. The design offers a means to adapt fundamental control design tradeoffs-like disturbance suppression versus noise sensitivity-which are otherwise fixed. Typically performance-limiting oscillations in the feedback system that benefit from extra control are temporarily upscaled and subjected to nonlinear weighting. For sufficiently large amplitudes, this nonlinear filter operation induces extra controller gain. Oscillations that do not benefit from this extra control (typically because they represent noise contributions that should not be amplified) remain unscaled and, as such, do not induce extra controller gain. The combined usage of linear weighting filters with their exact inverses renders this part of the nonlinear filter design strictly performance based. The effective means to improve servo performance is demonstrated on a short-stroke wafer stage of an industrial wafer scanner. Since the nonlinear filter design is largely based on Lyapunov arguments, stability is guaranteed along the different design steps.