Authors: D. de Rigo; A. Castelletti; A. E. Rizzoli; R. Soncini-Sessa; E. Weber

Abstract: An approach to the integrated water resources management based on Neuro-Dynamic Programming (NDP) with an improved technique for fastening its Artificial Neural Network (ANN) training phase will be presented. When dealing with networks of water resources, Stochastic Dynamic Programming provides an effective solution methodology but suffers from the so-called "curse of dimensionality", that rapidly leads to the problem intractability. NDP can sensibly mitigate this drawback by approximating the solution with ANNs. However in the real world applications NDP shows to be considerably slowed just by this ANN training phase. To overcome this limit a new training architecture (SIEVE: Selective Improvement by Evolutionary Variance Extinction) has been developed. In this paper this new approach is theoretically introduced and some preliminary results obtained on a real world case study are presented.

Keywords: integrated water resources management,stochastic dynamic programming,neurodynamic programming,evolutionary algorithm

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.02172

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 2171-2171

Authors: Letitia Mirea; Ron J. Patton

Abstract: This paper investigates the development of the Adaptive Neuro-Fuzzy Systems with Local Recurrent Structure (ANFS-LRS) and their application to Fault Detection and Isolation (FDI). Hybrid learning, based on a fuzzy clustering algorithm and a gradientlike method, is used to train the ANFS-LRS. The experimental case study refers to an application of fault diagnosis of an electro-pneumatic actuator. A neuro-fuzzy simplified observer scheme is used to generate the residuals (symptoms) in the form of the one-stepahead prediction errors. These are further analysed by a neural classifier in order to take the appropriate decision regarding the actual behaviour of the process.

Keywords: fault diagnosis, fuzzy multiple-modelling, fuzzy hybrid systems, neural networks, dynamic modelling, neural classifier, actuator systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.02438

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: ---

Authors: None

Abstract:

Keywords:

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00001

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: c1-c1

Authors: Tohru Katayama; Tomas McKelvey; Akira Sano; Christos Cassandras; Marco C. Campi

Abstract: This report discusses problems and methodologies that lie in the broad scope of systems and signals, with special focus on modeling, identification and signal processing; adaptation and learning; discrete event and hybrid systems; and stochastic systems. A common theme underlying all these areas is that problems in control systems and signals are usually defined and best studied in the framework of stochastic approaches. Although there are common precepts among all these technologies, there are also many unique topics within each area. Therefore, the current key problems in each technology are explained, followed by a discussion of recent major accomplishments with trends, and finally some forecasts of likely developments are provided. The conclusion summarizes some general forecasts for the overall field of systems and signals.

Keywords: systems and signals,modeling,system identification,adaptive control,learning,discre event systems,hybrid systems,stochastic systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00002

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 1-1

Authors: Jozsef Bokor; Gary Balas

Abstract: Linear Parameter Varying (LPV) systems appear in a form of LTI state space representations where the elements of the A(ρ), B(ρ), C(ρ) matrices depend on an unknown but at any time instant measurable vector parameter ρ ∈ P. This paper describes a geometric view of LPV systems. Geometric concepts and tools of invariant subspaces and algorithms for LPV systems affine in the parameters will be presented and proposed. Application of these results will be shown and referenced in solving various analysis (controllabilty/observability) problems, controller design and fault detection problems associated to LPV systems.

Keywords: geometric control,invariant subspaces,vector space distributions,dynamic inversion,decoupling,observer design

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00003

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 2-2

Authors: Silvio Simani; Cesare Fantuzzi

Abstract: This paper addresses the identification of non-linear dynamic systems. A wide class of these systems can be described using non-linear time-invariant regression models, that can be approximated by means of piecewise affine prototypes with an arbitrary degree of accuracy. This work concerns the identification of piecewise affine model structure through input-output data acquired from a dynamic process. In order to show the effectiveness of the developed technique, the results obtained in the identification of both a simple simulated system and a real dynamic process are reported.

Keywords: hybrid models,dynamic system identification,parameter estimation,data processing,non-linear systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00004

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 3-3

Authors: K. Dadhe; S. Engell

Abstract: In this paper, the estimation of prediction intervals for multi-step-ahead predictions from dynamic neural network models is described. Usually, asymptotic methods based on linearizations are applied with the potential problem of large coverage errors and too optimistic prediction intervals. The potential sources of these problems are the negligence of the network parameter uncertainties and the non-normality of the error distribution. To overcome these restrictions, bootstrap methods are used here. New formulations are introduced to apply the bootstrap to nonlinear time series models with exogenous input. An explicit model of the error process considers the influence of different training data densities on the empirical error distribution. A Monte Carlo study illustrates the proposed methods.

Keywords: dynamic neural networks,prediction,bootstrapping

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00005

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 4-4

Authors: E. Abd-Elrady; T. Soderstrom

Abstract: Second-order nonlinear ordinary differential equations (ODE's) can be used for modeling periodic signals. The right hand side function of the ODE model is parameterized in terms of polynomial basis functions. The least squares (LS) algorithm for estimating the coefficients of the polynomial basis gives biased estimates at low signal to noise ratios (SNRs). This is due to approximating the states of the ODE model using finite difference approximations from the noisy measurements. An analysis for this bias is given in this paper.

Keywords: bias,discretization,identification,least squares,nonlinear systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00006

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 5-5

Authors: Bernt M. Akesson; Hannu T. Toivonen

Abstract: State-dependent parameter representations of nonlinear stochastic sampled-data systems are studied. Velocity-based linearization is used characterize sampled-data systems using nominally linear models whose parameters can be represented as functions of past outputs and inputs. For stochastic systems the approach leads to state-dependent ARMAX (quasi-ARMAX) representations. The models and their parameters are identified from input-output data using feedforward neural networks to represent the model parameters as functions of past inputs and outputs.

Keywords: neural network models,nonlinear models,sampled-data systems,stochastic systems,parameter identification

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00007

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 6-6

Authors: Jus Kocijan; Agathe Girard

Abstract: Identification of nonlinear dynamic systems from experimental data can be difficult when, as often happens, more data are available around equilibrium points and only sparse data are available far from those points. The probabilistic Gaussian Process model has already proved to model such systems efficiently. The purpose of this paper is to show how one can relatively easily combine measured data and linear local models in this model. Also, using previous results, we show how uncertainty can be propagated through such models when predicting ahead in time in an iterative manner. The approach is illustrated with a simple numerical example.

Keywords: systems identification,Gaussian process models,nonlinear systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00008

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 7-7

Authors: Ingela Lind

Abstract: The objective of this paper is to find the structure of a nonlinear system from measurement data, as a prior step to model estimation. Applying ANOVA directly on a dataset is compared to applying ANOVA on residuals from a linear model. The distributions of the involved test variables are computed and used to show that ANOVA is effective in finding which regressors give linear effects and what regressors produce nonlinear effects. The ability to find nonlinear substructures depending on only subsets of regressors is an ANOVA feature which is shown not to be affected by subtracting a linear model.

Keywords: system identification,nonlinear systems,structural properties,analysis of variance,linear estimation

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00009

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 8-8

Authors: A. Krzyzak; J. Z. Sasiadek; B. Kegl

Abstract: Nonlinear dynamic systems of Hammerstein type are identified from input and output measurements. Identification algorithms for a memoryless nonlinear part and for a linear dynamic part are proposed. Convergence and rates of convergence of the algorithms are investigated. The class of nonlinearities considered in the paper is very large and cannot be parameterized therefore nonparametric approach is used. The performance of identification algorithms is studied in simulation experiments.

Keywords: Hammerstein system identification,Hermite series,convergence

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00010

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 9-9

Authors: Shuanghua Bai; Jules Thibault; David D. McLean

Abstract: The technique of dynamic data reconciliation has been previously studied in the literature and shown to be an effective tool to better estimate the true values of process variables by using information from both measured values and process models. Real-time implementation of dynamic data reconciliation involves solving complex optimization problem, leading to large computation time. This paper presents a study on the use of a dynamic Autoassociative Neural Network (AANN) for dynamic data reconciliation. Once trained, the AANN can be directly used for online signal validation. Closed-loop performance of the AANN for both linear and nonlinear processes was evaluated using simulations of two storage tank processes. The AANN provided accurate estimates of measured values for the two processes studied in this investigation.

Keywords: data reconciliation,dynamic neural network,controller performance

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00011

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 10-10

Authors: Zi-Jiang Yang; Hideto Iemura; Shunshoku Kanae; Kiyoshi Wada

Abstract: This paper considers the identification problem of continuous-time systems with unknown time delays from sampled input-output data. An iterative global separable nonlinear least-squares (GSEPNLS) method which estimates the time delays and transfer function parameters separably is derived, by using stochastic global-optimization technique to avoid convergence to a local minimum. Futhermore, the GSEPNLS method is modified to a novel global separable nonlinear instrumental variable (GSEPNIV) method to yield consistent estimates if the algorithm converges to the global minimum. Simulation results show that the proposed method works quite well.

Keywords: identification,continuous-time system,time-delay,instrumental variable method,stochastic approximation with convolution smoothing

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00012

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 11-11

Authors: Svante Gunnarsson

Abstract: System identification of a flexible mechanical system using decimated data is studied. It is illustrated how the use of decimated data can give erroneous results due to the inter-sample behavior of the signals, and an intuitive explanation to this phenomenon is proposed. The possible improvement by using alternative assumptions for the intersample behavior is investigated.

Keywords: identification,physical models,continuous time systems,flexible arms,sampling

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00013

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 12-12

Authors: Fumitoshi Sakai; Toshiharu Sugie

Abstract: The paper proposes a novel approach to identification of continuous-time systems from sampled I/O data. The coefficients of plant transfer functions are directly identied by applying an iterative learning control which enables us to achieve perfect tracking for uncertain plants by iteration of trials. Furthermore, one way to make the method robust against the measurement noises is shown. One of the merits of the proposed method is that it does not require time-derivative of I/O signals. In addition, it indicates us the estimation accuracy explicitly through tracking control experiments. Numerical examples are given to illustrate the effectiveness of the proposed method.

Keywords: continuous-time,system identification,iterative learning control

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00014

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 13-13

Authors: Anna Soffia Hauksdottir; Bergpor Aevarsson; Gisli Herjolfsson; Sven P. Sigurosson

Abstract: General closed-form expressions of linear continuous-time system responses, are derived. The system eigenvalues can be real and/or complex, and may be repeated. A recursive computationally attractive method is formulated by which the partial fraction expansion coefficients can be computed fast and accurately. The closed-form expressions include the numerator coefficients of the transfer function, a matrix containing the partial fraction expansion coefficients and the system's eigenvalues, and a vector containing the independent time-basis functions. Higher-order responses can easily be computed in closed form from the impulse response.

Keywords: impulse response,partial fraction expansion,closed-form expression,higher-order response

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00015

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 14-14

Authors: Jonas Gillberg; Lennart Ljung

Abstract: This paper treats identification of continuous-time output error (OE) models based on sampled data. The exact method for doing this is well known both for data given in the time and frequency domains. This approach becomes somewhat complex, especially for non-uniformly sampled data. We study various ways to approximate the exact method for reasonably fast sampling. While an objective is to gain insights into the non-uniform sampling case, this paper only gives explicit results for uniform sampling.

Keywords: continuous-time systems,parameter estimation,continuous-time OE

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00016

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 15-15

Authors: Kaushik Mahata; Minyue Fu

Abstract: A novel direct approach for modeling continuous-time stochastic processes is proposed in this paper. First the observed data is passed through an input-to-state filter and the covariance of the output state is computed. The properties of the state covariance matrix is then exploited to estimate the positive real spectrum of the observed data at a set of prescribed points on the right half plane. Finally, the continuous-time parameters are obtained from the positive real spectrum estimates by solving a Nevanlinna-Pick interpolation problem. The estimated model is stable. The analytical results are illustrated using numerical simulations.

Keywords: positive real spectrum,continuous-time,stochastic process,Nevanlinna-Pick interpolation

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00017

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 16-16

Authors: Magnus Evestedt; Alexander Medvedev

Abstract: Stationary properties of a recently suggested windup prevention scheme for recursive parameter estimation are investigated in the case of insufficient excitation. When the regressor vector contains data covering the whole parameter space, the algorithm has only one stationary point, the one defined by a weighting matrix. If the excitation is insufficient, the algorithm is shown to possess a manifold of stationary points and a parametrization of this manifold is given. However, if the past excitation conditions already caused the algorithm to converge to a certain point, the stationary solution would not be affected by current lack of excitation. This property guarantees good antiwindup properties of the studied parameter estimation algorithm.

Keywords: recursive estimation,Riccati equation,windup

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00018

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 17-17

Authors: Magnus Evestedt; Alexander Medvedev; Torbjorn Wigren

Abstract: The windup properties of a recently suggested recursive parameter estimation algorithm are investigated in comparison with a number of well-known techniques such as the Normalized Least Squares Algorithm (NLMS) and the Kalman filter (KF). An acoustic echo cancellation application is used as a benchmark for comparing the properties of different approaches. The basic performance of the method, both for white and colored input signal, appears to be similar to that of the KF and superior to the NLMS. When the energy in the input signal decreases, the algorithm performs best of all compared estimation schemes. Once the solution of the Riccati equation of the algorithm converged to a user defined point, it will stay there even if the input excitation is reduced. This explains the good anti-windup properties of the method.

Keywords: acoustic echo cancellation,recursive estimation,windup

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00019

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 18-18

Authors: Lim Sze Chieh; George Mathew

Abstract: Although partial response (PR) equalization employing the linearly constrained least-mean-square (LCLMS) algorithm is widely used in recording channels, there is no literature on its convergence analysis. Existing analyses of the LMS algorithm assume that the input signals are jointly Gaussian, which is an invalid assumption for PR equalization with binary input. In this paper, we present a convergence analysis of the LCLMS algorithm, without the Gaussian assumption. An approximate expression is derived for the misadjustment. It is shown that the step-size range required to guarantee stability is larger for binary data compared to Gaussian data.

Keywords: adaptation,adaptive equalization,constraints,LMS algorithm,mean-square error,partial response channels,recording channels,PR target

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00020

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 19-19

Authors: Y. Becis-Aubry; M. Boutayeb; M. Darouach

Abstract: This contribution proposes a robust recursive algorithm for state estimation of linear multi-output systems with unknown but bounded disturbances corrupting both the state and measurement vectors. A novel approach based on state bounding techniques is presented. The proposed algorithm can be decomposed into two steps: time updating and observation updating that uses a switching estimation Kalman-like gain matrix. Particular emphasis will be given to the design of a weighting factor that ensures consistency of the estimated state vectors with the input-output data and the noise constraints and that enforces convergence properties.

Keywords: state estimation,multivariable linear systems,recursive method,bounded noise,ellipsoids,stability

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00021

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 20-20

Authors: Alexey Bobtsov; Artem Kremlev

Abstract: The problem of designing an output feedback compensator for any biased sinusoidal disturbance is considered. In this paper, we will develop the approach presented in (Marino et al., 2003). In (Marino et al., 2003) a compensator of order (2n+6) is proposed, which solves the posed problem by using the adaptive observers developed in (Marino and Tomei, 1995; Marino et al., 2001). This problem is solved by a (n+4)-order compensator.

Keywords: adaptive algorithms,disturbance rejection,linear systems,decision feedback,output regulation,observers,single-input/single-output systems

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00022

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 21-21

Authors: W. E. Leithead; Kian Seng Neo; D. J. Leith

Abstract: When data contains components with different characteristics and it is required to identify both, standard Gaussian regression, based on a model with a single stochastic process, is inadequate. In this paper, a novel adaptation of Gaussian regression, based on models with two stochastic processes, is presented. In both the prior and posterior joint probability distributions, the Gaussian processes for the two components are independent. The effectiveness of the revised Gaussian regression method is demonstrated by application to wind turbine time series data.

Keywords: identification,Gaussian processes,independent priors,independent posteriors

Digital Object Identifier (DOI): 10.3182/20050703-6-CZ-1902.00024

Conference: Proceedings of the 16th IFAC World Congress, 2005

Location: , Czech Republic

Start Date: Mon Jul 04 2005 - End Date: Fri Jul 08 2005

Page Numbers: 23-23