A unified design for feedback-feedforward control system to improve regulatory control performance

The PID controller is widely used in industries because of its simplicity and robustness. A simple approach to improve regulatory control performance is to combine both feedback PID and feedforward controllers. The feedforward controller enables early compensation of a measured disturbance before it can seriously affect the process. The conventionally derived non-ideal feedforward controllers are not often used in practice. The reason is that an ideal feedforward controller based on direct inversion of process model is often not physically realizable. Several non-ideal feedforward control designs have been proposed where some of them involve rather intensive tuning procedure to obtain good disturbance rejection. In this paper, we present a new systematic method for designing a combined feedback-feedforward control system. The proposed design method is easy to use and applicable to stable, unstable and integrating deadtime processes where the ideal feedforward controller is physically not realizable.     

Complementary compound set-point control by combining muscular internal force feedforward control and sensory feedback control including a time delay

This paper proposes a new set-point control method for a musculoskeletal arm by combining muscular internal force feedforward control with feedback control including a large time delay. The proposed method accomplishes robust and rapid positioning with a relatively small muscular force. In the positioning by the muscular internal force feedforward controller, a large muscular force is required to achieve good performance. On the other hand, in the positioning by the feedback controller including the large time delay, the system can easily fall into an unstable state. A simple linear combination of these two controllers makes it possible to improve the control performance and to overcome the drawbacks of each controller in a complementary manner. First, a two-link six-muscle arm model is considered as a musculoskeletal system in this study. Second, the new set-point control method, which consists of the feedforward control signal and the feedback control signal including the time delay, is designed. Third, the stability of the proposed method is investigated using the Lyapunov–Razumikhin method. Finally, the results of numerical simulations and experiments are presented to demonstrate the advantages of the proposed method.     

綫性复合自动控制系統的图解綜合法

本文主要讨论了复合系统(具有开环及闭环控制回路的系统)的综合方法.文中拟出了一种用系统的对数频率特性计算系统误差平方积分值的简易方法,用这种方法可以很方便地综合前馈网络.对于执行环节输出有积分限制条件和用断续回路作为前馈单元的情形,文中也讨论了它们的综合方法.     

基于深度前馈神经网络的多因子人体表面积计算模型

人体表面积(BSA)在临床医学上有着至关重要的作用,但现有BSA计算方法大多只使用身高和体重2个参数且采用匹配简单函数的方法来估计体表面积,临床上也认为现有的BSA计算方法误差较大。针对这些问题,提出一种BSA回归预测模型。该回归预测模型包含2个部分：首先,借助相关性和显著性分析选择相关性较高的体表面积影响因子;其次,利用人体数据训练深度前馈神经网络,构建回归模型。实验分别采取5-折交叉验证与测试集验证2种方法。首先,将深度前馈神经网络模型与传统人体表面积计算方法进行精度评估和结果对比分析;其次将深度前馈神经网络模型与3种模型进行精度评估和结果对比分析。在与传统方法对比中,深度前馈神经网络模型的决定系数高于2种传统方法的,且比传统方法提高了6%,误差与传统方法的相比降低了近一倍。在与3种模型的对比中,深度前馈神经网络的决定系数比其他模型的提高了至少2%,误差降低。一致性分析实验结果也显示,深度前馈神经网络95%一致性界限最小,一致性最好。总体来说,提出的回归预测模型可以得到更加精确的体表面积预测值。     

Architecture and statistical model of a pulse-mode digitalmultilayer neural network

A new architecture and a statistical model for a pulse-mode digital multilayer neural network (DMNN) are presented. Algebraic neural operations are replaced by stochastic processes using pseudo-random pulse sequences. Synaptic weights and neuron states are represented as probabilities and estimated as average rates of pulse occurrences in corresponding pulse sequences. A statistical model of error (or noise) is developed to estimate relative accuracy associated with stochastic computing in terms of mean and variance. The stochastic computing technique is implemented with simple logic gates as basic computing elements leading to a high neuron-density on a chip. Furthermore, the use of simple logic gates for neural operations, the pulse-mode signal representation, and the modular design techniques lead to a massively parallel yet compact and flexible network architecture, well suited for VLSI implementation. Any size of a feedforward network can be configured where processing speed is independent of the network size. Multilayer feedforward networks are modeled and applied to pattern classification problems such as encoding and character recognition.     

Training two-layered feedforward networks with variable projection method.

The variable projection (VP) method for separable nonlinear least squares (SNLLS) is presented and incorporated into the Levenberg-Marquardt optimization algorithm for training two-layered feedforward neural networks. It is shown that the Jacobian of variable projected networks can be computed by simple modification of the backpropagation algorithm. The suggested algorithm is efficient compared to conventional techniques such as conventional Levenberg-Marquardt algorithm (LMA), hybrid gradient algorithm (HGA), and extreme learning machine (ELM).     

Two adaptive control structures of robot manipulators

This paper proposes two simple adaptive control schemes of robot manipulators. The first one is the state feedback control which consists of feedforward from the desired position trajectory, PD feedback from the actual trajectory, and an auxiliary input. The second one is the feedforward/feedback control which consists of a feedforward term from the desired position, velocity, and acceleration trajectory based on the inverse of robot dynamics. The feedforward, feedback, and auxiliary gains are adapted using simple equations derived from the decentralized adaptive control theory based on Lyapunov's direct method, and using only the local information of the corresponding joint. The proposed control schemes are computationally fast and do not require a priori knowledge of the detail parameters of the manipulator or the payload. Simulation results are presented in support of the proposed schemes. The results demonstrate that both controllers perform well with bounded adaptive gains.     

Feedforward controller synthesis for linear time variant systems

This paper develops a framework for the synthesis of a feedforward controller for both discrete and continuous linear time variant (LTV) systems. The proposed method determines the structure of a feedforward controller, which is cascaded with the known time varying process, so that the ensemble behaves as a linear time invariant (LTI) system, satisfying certain design requirements. The obtained controller is simple, consisting in a parallel bank of first or second order LTI filters, followed by amplifiers with time variable gain. Simulation results are included to illustrate theoretical considerations. Copyright © 2008 John Wiley and Sons Asia Pte Ltd and Chinese Automatic Control Society     

Application of a mathematical model for the control and optimization of a distillation plant

The principles of optimizing process control are elucidated by weighing the merits of hill-climbing, feedback, methods against model, feedforward, methods. It is shown that the advantages of the two methods can be employed at both the optimizing and the stabilizing control levels of a chemical plant. A simple static model containing appropriately chosen non-linearities is being used in a feedforward control loop and updated regularly by feedback in an openloop computer control system, results of which are given.Finally, it is shown that such a model can also be used for statistical evaluation of computer control methods.     

Feedforward control with online parameter estimation applied to the Chylla–Haase reactor benchmark

The Chylla–Haase polymerization reactor is used as a benchmark problem to illustrate the potential of feedforward control design by extending the conventional cascade control in the framework of the two-degree-of-freedom control concept. To ensure an accurate temperature control, the feedforward control is adapted to different batch conditions and various products by an extended Kalman filter. The reaction heat and the heat transfer coefficient are estimated online based on simple physically motivated relations applicable to a wider range of batch reactors. Simulation results under model uncertainties show the effectiveness and accuracy of the adaptive feedforward control concept while maintaining the conventional feedback cascade control.     

Combined fuzzy based feedforward and bubble size distribution based feedback control for reagent dosage in copper roughing process

A combined fuzzy based feedforward (FBF) and bubble size distribution (BSD) based feedback reagent dosage control strategy is proposed to implement the product indices in copper roughing process. A fuzzy theory based feedforward compensator will be used to calculated the reagent dosage in advance to eliminate the influence of large disturbances according to ore grade and handling capacity. Since the bubble size is believed to be closely related to flotation performance and responds to changes in the reagent dosage, using BSD based feedback predictive control calculates the reagent dosage to stabilize the flotation running. Instead of simple statistic feature, the bubble size with non-Gaussian feature is characterized to be probability density function (PDF) by using B-spline. A multi-output least square support vector machine (MLS-SVM) based is then applied to establish a dynamical relationship between the weights of B-spline and the reagent dosage since the weights are interrelated and related to the reagent dosage. A multiple step based optimization algorithm is finally proposed to determine the reagent dosage. Experimental results can show the effectiveness of the proposed method.     

Realization of simple adaptive control by using parallel feedforward compensator

A problem is dealt with concerning a practical design scheme for a multi-input/multi-output simple adaptive control (SAC) system. The SAC method which was first derived as a sort of the direct model following adaptive control based on the command generator tracker (CGT) technique was fundamentally applicable to plants satisfying the so-called almost strictly positive real (ASPR) condition. However, the method can be applied to non-ASPR plants by using a parallel feedforward compensator (PFC), which makes the augmented plant including PFC be ASPR. Such a technique was proposed by Bar-Kana for plants satisfying the output feedback stabilizable condition. As an extension, the authors have proposed a simple and concrete PFC design procedure for SISO non-ASPR plants. The special feature of this proposal is that the method can be applicable to non-ASPR plants as the minimum phase system. The main aim of the paper is to give a practical design procedure of PFC for MIMO plants as a natural extension of the SISO case, so that an actual realization of MIMO SAC can be attained for a broad class of MIMO plants. The effectiveness of the proposed method is also confirmed through numerical simulations.     

压电陶瓷迟滞神经网络建模与线性补偿控制

针对压电陶瓷的迟滞非线性,本文首先进行实验测量得到压电陶瓷的位移迟滞数据;通过分析实验数据,引入线性方程实现压电陶瓷输入电压与输出位移关系的线性化,并建立了基于多项式拟合算法的神经网络迟滞模型;根据迟滞模型设计前馈控制器,分别采用了前馈开环和前馈结合PID的方法对压电陶瓷迟滞非线性进行补偿控制实验.实验结果表明,采用前馈开环控制,压电陶瓷位移主环迟滞减小了91.84%,位移次环迟滞减小了85.67%,位移跟踪的平均相对误差为2.97%;采用前馈结合PID控制,压电陶瓷位移主环迟滞减小了96.42%,位移次环迟滞减小了88.44%,位移跟踪的平均相对误差为2.04%.证明了该控制方法能有效地抑制压电陶瓷的迟滞非线性.     

唯输出条件下前馈网络参数的还原

将频谱理论的一个结论进行了推广,并巧妙地将其运用于前馈网络的分析中,由此得到一种关于前馈网络未知参数的攻击还原方法。即对于一个前馈网络,在非线性滤波函数和抽头集两个参数均未知的情况下,仅由足够长的前馈输出序列就能够实现这两个参数的求解还原。     

Robust output regulation by observer-based feedforward control

In this paper, a new method for the design of disturbance and reference feedforward controllers with disturbance observers is presented for systems, where the disturbance input location is unknown. The resulting controller achieves asymptotic tracking in the presence of the disturbances despite plant parameter variations and changing disturbance input locations, and is therefore robust. It is shown that the new controller design method is dual to Davison's approach for the application of the internal model principle. However, different from the latter one, the new design procedure allows a simple and complete prevention of controller windup by driving the disturbance observer with the saturated input. Furthermore, a straightforward separate reference channel design is possible in order to improve the tracking behaviour if the types of the reference and the disturbance signals differ. A simple example demonstrates the advantages of the new design method.     

Performance analysis of a pipelined backpropagation parallelalgorithm

The supervised training of feedforward neural networks is often based on the error backpropagation algorithm. The authors consider the successive layers of a feedforward neural network as the stages of a pipeline which is used to improve the efficiency of the parallel algorithm. A simple placement rule is used to take advantage of simultaneous executions of the calculations on each layer of the network. The analytic expressions show that the parallelization is efficient. Moreover, they indicate that the performance of this implementation is almost independent of the neural network architecture. Their simplicity assures easy prediction of learning performance on a parallel machine for any neural network architecture. The experimental results are in agreement with analytical estimates.     

一种模糊神经网络控制器参数的混沌优化设计

通过模糊控制与神经网络相串联的方式构成模糊神经网络系统，然后提出一种基于模拟退火策略的混沌优化算法，将该算法引入模糊神经网络参数域中进行优化，实现混沌粗搜索与细搜索相结合优化目的，体现出具有更强的模糊神经网络参数全局最优解的搜索能力。采用该控制器对一个非线性对象进行控制。仿真实验表明，该方法能有效地实现模糊神经网络控制器参数优化，控制具有无振荡、超调小、调节时间短等优点，算法结构简单，容易实现。     

Noise tolerance of attractor and feedforward memory models

In short-term memory networks, transient stimuli are represented by patterns of neural activity that persist long after stimulus offset. Here, we compare the performance of two prominent classes of memory networks, feedback-based attractor networks and feedforward networks, in conveying information about the amplitude of a briefly presented stimulus in the presence of gaussian noise. Using Fisher information as a metric of memory performance, we find that the optimal form of network architecture depends strongly on assumptions about the forms of nonlinearities in the network. For purely linear networks, we find that feedforward networks outperform attractor networks because noise is continually removed from feedforward networks when signals exit the network; as a result, feedforward networks can amplify signals they receive faster than noise accumulates over time. By contrast, attractor networks must operate in a signal-attenuating regime to avoid the buildup of noise. However, if the amplification of signals is limited by a finite dynamic range of neuronal responses or if noise is reset at the time of signal arrival, as suggested by recent experiments, we find that attractor networks can outperform feedforward ones. Under a simple model in which neurons have a finite dynamic range, we find that the optimal attractor networks are forgetful if there is no mechanism for noise reduction with signal arrival but nonforgetful (perfect integrators) in the presence of a strong reset mechanism. Furthermore, we find that the maximal Fisher information for the feedforward and attractor networks exhibits power law decay as a function of time and scales linearly with the number of neurons. These results highlight prominent factors that lead to trade-offs in the memory performance of networks with different architectures and constraints, and suggest conditions under which attractor or feedforward networks may be best suited to storing information about previous stimuli.     

A discrete-time adaptive control scheme for robot manipulators

A discrete-time model reference adaptive control scheme is developed for trajectory tracking of robot manipulators. The scheme utilizes feedback, feedforward, and auxiliary signals, obtained from joint angle measurement through simple expressions. Hyperstability theory is utilized to derive the adaptation laws for the controller gain matrices. It is shown that trajectory tracking is achieved despite gross robot parameter variation and uncertainties. The method offers considerable design flexibility and enables the designer to improve the performance of the control system by adjusting free design parameters. The discrete-time adaptation algorithm is extremely simple and is therefore suitable for real-time implementation. Simulations and experimental results are given to demonstrate the performance of the scheme.     

基于MATLAB的BP网络数字图像预测器设计

针对数字图像预测编码系统的预测器设计问题,采用三层BP网络作为预测器,以邻域像素作为输入,隐含层、输出层激活函数分别采用双曲正切S型函数、线性函数.给出了网络训练算法原理,研究运用MALTAB工具设计预测器的方法.通过实验手段讨论了在给定均方差和训练时的数据量的情况下,隐含层神经元的数目与预测精度无关的结论,介绍了基于MATLAB的多种网络训练方法,最后设计的预测器具有结构简单、易于实现的优点;对两幅特征明显的图像进行预测,取得满意的效果.实验结果表明该设计方法是可行、有效的.