Identification of time-series signals using a dynamic neural network with GA-based training

We propose a dynamic neural network (DNN) that realizes a dynamic property and has a network structure with the properties of inertia, viscosity, and stiffness without time-delayed input elements, and a training algorithm based on a genetic algorithm (GA). In a previous study, we proposed a modified training algorithm for the DNN based on the error back-propagation method. However, in the previous method it was necessary to determine the values of the DNN property parameters by trial and error. In the newly proposed DNN, the GA is designed to train not only the connecting weights but also the property parameters of the DNN. Simulation results show that the DNN trained by the GA obtains good performance for time-series patterns generated from an unknown system, and provides a higher performance than the conventional neural network. This work was presented in part at the 10th International Symposium on Artificial Life and Robotics, 0ita, Japan, February 4–6, 2005     

A NEURAL NETWORK APPROACH TO ON-LINE IDENTIFICATION OF NONLINEAR SYSTEMS

Research investigating neural identification of dynamic systems has concentrated on off-line techniques. To be suitable for adaptive process control, on-line algorithms must be developed. This study investigates a modified back-propagation technique to achieve practical on-line capability. A technique denoted history-stack enhancement greatly improves the identification performance of the neural network. As a demonstration, a composite system of formidable but realistic nonlinear components was constructed and used to compare identification techniques including a recursive linear estimator and die new neural method. The results show that on-line neural identification is feasible for a wide range of formidable nonlinear characteristics individually found in industrial processes. Although performance is slower than with linear identification, the asymptotic accuracy of the neural technique is better for the nonlinear system being identified.     

Adaptive fuzzy APSO based inverse tracking-controller with an application to DC motors

This paper introduces the use of the adaptive particle swarm optimization (APSO) for adapting the weights of fuzzy neural networks (FNN) on line. The fuzzy neural network is used for identification of the dynamics of a DC motor with nonlinear load torque. Then the motor speed is controlled using an inverse controller to follow a required speed trajectory. The parameters of the DC motor are assumed unknown as well as the nonlinear load torque characteristics. In the first stage a nonlinear fuzzy neural network (FNN) is used to approximate the motor control voltage as a function of the motor speed samples. In the second stage, the above mentioned approximator is used to calculate the control signal (the motor voltage) as a function of the speed samples and the required reference trajectory. Unlike the conventional back-propagation technique, the adaptation of the weights of the FNN approximator is done on-line using adaptive particle swarm optimization (APSO). The APSO is based on the least squares error minimization with random initial condition and without any off-line pre-training. Simulation results are presented to prove the effectiveness of the proposed control technique in achieving the tracking performance.     

An adaptive recurrent fuzzy system for nonlinear identification

This paper describes the architecture and training procedure of a recurrent fuzzy system (RFS). The RFS is composed of a fuzzy inference system (FIS) and a delayed feedback connection. The recurrent property comes from feeding the FIS output back to the FIS input via an adjustable feedback parameter. Both the on-line and off-line training procedures based on the backpropagation-through-time (BPTT) algorithm have been investigated. The adjoint model of the RFS is obtained and used to compute the gradients. It is shown that the off-line training is insufficient to adapt to changes in system dynamics. So, an on-line training procedure is derived. In this procedure, a first in first out stack is used to store a certain history of the input–output data to perform a truncated BPTT algorithm. A quasi-Newton optimization method with a line search algorithm is used to adjust the RFS parameters. The performance of the developed RFS is demonstrated by applying to the identification of nonlinear dynamic systems. The simulation studies show that the proposed identification model has the ability to learn dynamics of highly nonlinear systems and compensate system uncertainties. The results are promising for the further application in the area of control and modeling.     

一种基于改进遗传算法的模糊神经网络控制器及其在烧结终点控制中的应用

针对烧结过程这一复杂、多参数耦合的高度非线性系统,融合遗传算法、神经网络和模糊控制的优点,提出一种基于改进遗传算法的模糊神经网络控制方法,并应用于烧结过程终点控制．首先采用遗传算法对给定的模糊神经网络控制器结构参数进行离线优化,然后利用BP算法较强的局部搜索能力和对对象的适应能力,进一步进行参数的在线调整．同时,为解决传统遗传算法早熟和收敛速度慢的问题,从交叉和变异算子、适应度函数选取等方面对遗传算法进行改进．采用精英保留策略,提高了全局搜索性能和收敛速度．仿真结果表明,所提出的控制器优于常规的模糊神经网络控制器(Fuzzy Neural Network Controller, FNNC)．算法的实际应用效果良好,为解决烧结终点控制问题提供了一条新的途径．     

基于改进遗传算法寻优的神经网络PID控制及应用

根据神经网络PID控制器初值的选取影响系统控制性能的特点,提出了一种基于改进遗传算法寻优的神经网络PID控制方法.即先利用遗传算法对PID控制器参数离线寻优,将求出的参数值作为控制器的比例、积分、微分系数的初值,再进行神经网络PID控制.对一类液位过程的实时控制结果表明采用本方法的控制系统具有较好的稳定性和鲁棒性.     

Adaptive Statistic Tracking Control Based on Two-Step Neural Networks With Time Delays

This paper presents a new type of control framework for dynamical stochastic systems, called statistic tracking control (STC). The system considered is general and non-Gaussian and the tracking objective is the statistical information of a given target probability density function (pdf), rather than a deterministic signal. The control aims at making the statistical information of the output pdfs to follow those of a target pdf. For such a control framework, a variable structure adaptive tracking control strategy is first established using two-step neural network models. Following the B-spline neural network approximation to the integrated performance function, the concerned problem is transferred into the tracking of given weights. The dynamic neural network (DNN) is employed to identify the unknown nonlinear dynamics between the control input and the weights related to the integrated function. To achieve the required control objective, an adaptive controller based on the proposed DNN is developed so as to track a reference trajectory. Stability analysis for both the identification and tracking errors is developed via the use of Lyapunov stability criterion. Simulations are given to demonstrate the efficiency of the proposed approach.      

MIMO系统的MPSO-PID型神经网络解耦控制研究

针对多输入多输出系统(MIMO系统)多变量、非线性、强耦合的特点,提出采用改进粒子群优化算法(MPSO)对PID型神经网络的权值进行优化的方法,实现对MIMO系统的解耦控制.其中,与基本PSO算法相比,MPSO算法后期仍能保持种群的多样性和较大的搜索空间;PID型神经网络是一种3层前向神经网络,网络各层神经元个数、连接方式、连接权值的初值都是按PID控制规律确定的.通过仿真分析,该方法有很好的控制品质:跟踪快、鲁棒性强、解耦效果好,为实际应用中强耦合系统控制方法的改进提供了理论依据.     

基于动态函数连接神经网络的自适应逆控制系统辨识研究

自适应逆控制将系统扰动消除和动态响应性能独立分开控制,其性能的优劣取决于系统对象、逆对象及逆控制器模型辨识精度的高低。文中提出用动态函数连接神经网络来实现自适应逆控制系统对象、逆对象的同时在线建模和逆控制器的离线建模,并将模型参数的辨识转化为空间参数寻优。针对混沌初始化对已收敛种群结构的破坏性,提出用变参数混沌粒子群优化算法对神经网络权值进行全局寻优,通过仿真实验可以看出基于动态函数连接神经网络的建模误差小,辨识精度高;与当前的参考模型自适应控制方法进行对比分析,所提方法能取得较好的扰动消除效果,并能使系统的跟踪响应性能得到提高,从而验证了方法的有效性、可行性。     

基于状态延迟动态递归神经网络的机器人动态自适应跟踪辨识

对一种在Elman动态递归网络基础上发展而来的复合输入动态递归网络(CIDRNN)作 了改进,提出一种新的动态递归神经网络结构,称为状态延迟动态递归神经网络(State Delay Input Dynamical Recurrent Neural Network).具有这种新的拓扑结构和学习规则的动态递归网 络,不仅明确了各权值矩阵的意义,而且使权值的训练过程更为简洁,意义更为明确.仿真实验 表明,这种结构的网络由于增加了网络输入输出的前一步信息,提高了收敛速度,增强了实时 控制的可能性.然后将该网络用于机器人未知非线性动力学的辨识中,使用辨识实际输出与机理 模型输出之间的偏差,来识别机理模型或简化模型所丢失的信息,既利用了机器人现有的建模 方法,又可以减小网络运算量,提高辨识速度.仿真结果表明了这种改进的有效性.     

Affine modeling of nonlinear multivariable processes using a new adaptive neural network-based approach

This paper presents a new method for on-line identification of exact affine model for multivariable processes with nonlinear and time-varying behaviors. A self-generating radial basis function (RBF) neural network trained by growing and pruning algorithm for RBF (GAP–RBF) is utilized for deriving the affine model. The extended Kalman filter (EKF) is used for parameter adaptation in the GAP–RBF neural network. The growing and pruning criteria of the original GAP–RBF have been modified with the objective to enhance its performance in on-line identification. Simulation results on two nonlinear multivariable CSTR benchmark problems show an excellent performance of the proposed approach, incorporated with the modified GAP–RBF (MGAP–RBF) neural network, for affine modeling.     

改进的RBF神经网络在非线性系统中的应用

在RBF神经网络的各种学习算法中,最近邻聚类算法学习时间短、计算量小,不需要事先确定隐单元的个数,完成聚类所得到的网络是最优的,并且可以在线学习,是一种自适应聚类学习算法,非常适合非线性实时系统的应用。但常规最近邻聚类算法在实时性要求较高的系统预测中学习时间相对较长。针对这一问题,提出了系统离线学习时采用减聚类算法,在线学习时采用改进的最近邻聚类算法,并变步长修正聚类半径和限制学习样本数。在函数拟合实验中,这种改进算法明显缩短了RBF神经网络的学习时间,在钢包精炼炉电极系统的在线辨识中的成功应用进一步表明对最近邻聚类算法的改进是有效的。     

基于神经网络的自治水下机器人广义预测控制

针对自治式水下机器人高度非线性和时变性的特点,提出了一种基于神经网络的水下机器人广义预测控制策略．利用改进型Elman网络作为多步预测模型,在对网络学习算法进行改进的基础上,实现了Elman网络的在线学习,并提出了用于求解神经广义预测控制律的灵敏度公式．进行了具有神经网络在线学习功能和不具有在线学习功能的水下机器人的速度控制实验,并就预测控制效果进行了对比分析．实验结果表明,具有自适应学习功能的水下机器人速度控制法的精度要优于不具有在线学习功能的速度控制法,且当水下机器人动态特性发生变化时具有较强的自适应能力．     

Adaptive Feedback Linearization Using Efficient Neural Networks

For a class of single-input, single-output, continuous-time nonlinear systems, a feedback linearizing neural network (NN) controller is presented. Control action is used to achieve tracking performance. The controller is composed of a robustifying term and two neural networks adapted on-line to linearize the system by approximating two nonlinear functions. A stability proof is given in the sense of Lyapunov. No off-line weight learning phase is needed and initialization of the network weights is straightforward. The NN controller is tested on a standard benchmark problem.     

基于动态神经网络的非线性组合系统自适应观测器

基于动态神经网络，对一类非线性组合系统提出一种观测器设计方法．在观测器设计中，充分考虑了神经网络逼近误差项对观测器性能的影响，增加了鲁棒控制项，并设计了相应的参数自适应律，以保证良好的观测性能．神经网络的连接权值在线调整，无需离线学习．仿真结果表明了该方法的有效性．     

Distributed parameter system identification using finite element differential neural networks

Most of the previous work on identification involves systems described by ordinary differential equations (ODEs). Many industrial processes and physical phenomena, however, should be modeled using partial differential equations (PDEs) which offer both spatial and temporal distributions that are simply not available with ODE models. Systems described by a PDE belong to a class of system called distributed parameter system (DPS). This article presents a method for solving the problem of identification of uncertain DPSs using a differential neural network (DNN). The DPS, assumed to be described by a PDE, is approximated using the finite element method (FEM). The FEM discretizes the domain into a set of distributed and connected nodes, thereby, allowing a representation of the DPS in a finite number of ODEs. The proposed DNN follows the same interconnection structure of the FEM, thus allowing the DNN to identify the FEM approximation of the DPS in both 2D and 3D domains. Lyapunov's second method was used to derive adaptive learning laws for the proposed DNN structure. The identification algorithm, here developed in Nvidia's CUDA/C to reduce the execution time, runs mostly on the graphics processing unit (GPU). A physical experiment served to validate the 2D case. In the experiment, the DNN followed the trajectory of 57 markers that were placed on an undulating square piece of silk. The proposed DNN is compared against a method based on principal component analysis and an artificial neural network trained with group search optimization. In addition to the 2D case, a simulation validated the 3D case, where input data for the DNN was generated by solving a PDE with appropriate initial and boundary conditions over an unitary domain. Results show that the proposed FEM-based DNN approximates the dynamic behavior of both a real 2D and a simulated 3D system.     

Nonlinear systems identification using dynamic multi-time scale neural networks

In this paper, two Neural Network (NN) identifiers are proposed for nonlinear systems identification via dynamic neural networks with different time scales including both fast and slow phenomena. The first NN identifier uses the output signals from the actual system for the system identification. The on-line update laws for dynamic neural networks have been developed using the Lyapunov function and singularly perturbed techniques. In the second NN identifier, all the output signals from nonlinear system are replaced with the state variables of the neuron networks. The on-line identification algorithm with dead-zone function is proposed to improve nonlinear system identification performance. Compared with other dynamic neural network identification methods, the proposed identification methods exhibit improved identification performance. Three examples are given to demonstrate the effectiveness of the theoretical results.     

遗传算法自适应模糊神经网络控制

神经网络能够以任意精度逼近任意复杂的非线性关系,具有高度的自适应和自组织性,在解决高度非线性和严重不确定系统的控制方面具有巨大的潜力.但一般神经网络训练算法如BP算法训练速度慢,受初值影响大且易陷入局部极小点,该文提出了一种基于模糊神经网络的间接自校正控制系统,控制器以高斯隶属度函数的径向基函数（RBF）神经网络结构,利用改进的遗传算法（GA）对结构和参数进行同步优化,改进适应度函数指导搜索过程,在保证稳定情况下大大加快了收敛的速度.神经网络正向模型（NNP）利用弹性BP算法进行离线辨识,使得到的模型泛化性能好.     

Temperature decoupling control of double-level air flow field dynamic vacuum system based on neural network and prediction principle

Double-level air flow field dynamic vacuum (DAFDV) system is a strong coupling, large time-delay, and nonlinear multi-input–multi-output system. Decoupling and overcoming the impact of time-delay are two keys to obtain rapid, accurate and independent control for two air temperatures in two concatenate chambers of the DAFDV system. A predictive, self-tuning proportional-integral-derivative (PID) decoupling controller based on a modified output–input feedback (OIF) Elman neural model and multi-step prediction principle is proposed for the nonlinearity, time-lag, uncertainty and strong coupling characteristics of the system. A multi-step ahead prediction algorithm is presented for temperature prediction to eliminate the effects of time-delays. To avoid getting into a local optimization, an improved particle swarm optimization is applied to optimize the weights of the OIF Elman neural network during modeling. By using the modified OIF Elman neural network identifier, the DAFDV system is identified and the parameters of PID controller are tuned on-line. The experimental results for two typical cases indicate that the settling times are obviously shorten, steady-state performances are improved and more important is that one temperature no longer fluctuates along the other, which verify the proposed adaptive PID decoupling control is effective.     

Neural-network predictive control for nonlinear dynamic systems with time-delay

A new recurrent neural-network predictive feedback control structure for a class of uncertain nonlinear dynamic time-delay systems in canonical form is developed and analyzed. The dynamic system has constant input and feedback time delays due to a communications channel. The proposed control structure consists of a linearized subsystem local to the controlled plant and a remote predictive controller located at the master command station. In the local linearized subsystem, a recurrent neural network with on-line weight tuning algorithm is employed to approximate the dynamics of the time-delay-free nonlinear plant. No linearity in the unknown parameters is required. No preliminary off-line weight learning is needed. The remote controller is a modified Smith predictor that provides prediction and maintains the desired tracking performance; an extra robustifying term is needed to guarantee stability. Rigorous stability proofs are given using Lyapunov analysis. The result is an adaptive neural net compensation scheme for unknown nonlinear systems with time delays. A simulation example is provided to demonstrate the effectiveness of the proposed control strategy.