Despite the popularity of rational transfer function matrix models for continuous-time MIMO systems, associated problems like "bias" due to measurement noise with equation error minimization, and reducibility of the estimated models, prompt for alternatives. This paper considers the use of models parametrized in terms of time moments of impulse response of continuous linear finite-dimensional MIMO systems. A novel algorithm presented in this paper gives bias-free irreducible estimates. Relevant implementation issues are discussed and simulation results are presented.

View full abstract

The periodic autoregressive process is one of unstationary processes. For directly obtaining the coefficients of periodic AR processes, the circular Levinson algorithm was derived. In this paper, we construct backward periodic AR processes from the auxiliary coefficients used in this algorithm. A numerical example is also presented and the statistical property of the estimated coefficients of backward periodic AR processes based on a sample of finite size is derived.

View full abstract

The purpose of this paper is to propose a comprehensive design procedure for the ARIMAX type time series model including Box-Jenkins type mixed time series model. The design procedure contains various algorithms which may be useful for systematic time series model building, especially an effective algorithm to set up initial values for nonlinear optimization which is indispensable to the parameter estimation of time series model including MA part is given. Finally, using practical sales and advertising data of two companies, the effectiveness of the design procedure proposed in this paper is shown.

View full abstract

In this paper, a method of active control is presented for suppressing random vibration of class of flexible structures subject to both wind and seismic disturbances. It should be noted that inherently each of them has mutually different characteristics. Therefore, it is reasonable from the practical point of view to provide two types of controllers, one of which excludes the vibration due to wind disturbance and another suppresses the vibration due to seismic input. Following to the idea,the state variable is split into two parts-the one is driven by wind disturbance and the other by the seismic one. Consequently the optimal controls are obtained in the form of feedback and feedforward.

View full abstract

In this paper, a method of the active boundary control is proposed for suppressing the vibration of mechanically flexible structures subject to random disturbance. The mathematical model is described by the Euler-Bernoulli type partial differential equation. Reduced-order models of the system are described by the Itô stochastic differential equations. For the system, we employ the cost functional having the input derivative constraints in order to obtain the controls which are well smoothed enough to generate by practical actuator. By introducing an augmented vector and a new process which drives the control by its integration, we obtain the optimal control as the feedback form of integrated state and its own integration.

View full abstract

In this paper, under the LQG control performance criterion, a method for realizing the suboptimal control for one-dimensional nonlinear stochastic systems with finite horizon is developed by exhibiting a wide variety of sample runs. As preliminaries, the equation of determining state trajectory satisfying the performance criterion and its linearization scheme are reviewed. In order to realize finite horizon control for nonlinear systems, the suboptimal state trajectory emanating from the initial state is switched at time t_s into the one hitting the target. Decision of switching time is illustrated by results of simulation experiments. Finally, relative relation between the initial state level and the target state one is examined.

View full abstract

In this paper, algebraic matrix Riccati equations are analyzed that have nonnegative-definite quadratic as well as constant terms, employing the so-called algebraic method. Necessary and sufficient conditions are established for the existence of stabilizing solution, which is not necessarily nonnegative-definite unlike the case of standard equations arising in LQG problems. Nonnegative-definite stabilizing solution is shown to exist if and only if the system is asymptotically stable and the H _∞ norm of transfer matrix be less than one. For positive-definite solution, we need additional condition that the system be controllable. Next necessary and sufficient conditions are established for the existence of the so-called anti-stabilizing solution for both general and positive-definite cases. It is shown that stabilizing solution is minimal existence of nonnegative-definite stabilizing solution, all the other (hence nonnegative-definite) solutions with lattice structure are derived, together with the number of nonnegative-definite solutions.

View full abstract

This paper proposes a new method of identification of continuous time-delay systems from sampled input-output data. Identification of multi-input, multi-output systems where the time-delays in the individual input channels may differ from each other, is discussed. With the aid of a digital pre-filter, an approximated discrete-time estimation model is first derived at, in which the system parameters remain in their original form and the time-delays need not be an integral multiple of the sampling period. Then an identification method combining the instrumental variable (IV) method with the genetic algorithm (GA) is proposed, i.e the time-delays are selected by the GA, and the system parameters are estimated by the IV method. Simulation results show that our method yields excellent estimates even in the presence of high measurement noises.

View full abstract

In the first place, this paper briefly examines state space models and their relationship to standard econometric models. Special discussions are critically made of similarities and differences concerning general concepts between the econometric and control engineering literatures. Applications of optimal control theory to structural, forecasting, planning and policy analysis are reconsidered. It is noted that crude model will sometimes give quite a creditable control strategy. Some important roles of certainty equivalence theorem are emphasized on its attractive use of simple and practical computational procedure associated with economic planning and policy design. On the other hand, it is noted that the separation theorem used in the control engineering literature is not necessarily in related to the application of economic stabilization policy problems. Furthermore discussions on data information and experimental possibility are compared with both literatures. As a result, the Kalman filtering plays a crucially more important role only in the product of control apparatus rather than it is applied to econometric models.

View full abstract

Retrieval algorithms for estimation of vertical profiles of stratospheric aerosols based on the visible data by ADEOS/ILAS in 1996 are described. Algorithms mainly consist of three parts. The first process is base-line estimation for observed transmittance spectrum. The form of Aλ^-4+Bλ+C, where λ represents wavelength, is adopted for our base-line. Next step is the inversion process which derives vertical profiles of total atmospheric extinction from ILAS transmittance data. The final step is the decomposition process from the obtained total extinction into each component of the atmospheric constituents, i.e. air molecules and aerosols. It is shown in model calculations with an ILAS simulator that our treatment provides the quick and accurate results within the expected domain of validity.

View full abstract

In the present paper, it is discussed how to treat with the analytical and numerical solutions of searchlight problem with faint background light on flat terrain bounded below by a diffuse reflector. It is elucidated that the angular distribution of radiation emergent from the top of the turbid slab of finite thickness with a point source on top is discussed by imbedding the problem within a family of parameters, allowing for the numerical solution via Monte Carlo method.

View full abstract

With the aims of exploring humankind's biaural localization ability as well as developing refined devices for environmental recognition in robotics, a practical method for simultaneous estimation of the range and bearing of a stationary random acoustical source in near-field is proposed by extracting new information about amplitude attenuation of propagated waves and combining it with the conventional relative time delay between the detected signals. The principle and effectiveness of the proposed localization method are presented, together with the results of numerical analyses and computer simulations, illustrating the usefulness and fundamental characteristics of this method under practical circumstances.

View full abstract

In this investigation, we derived dynamics of installed pipes' actions which is described by the stochastic differential equation from some assumptions based on construction works by the micro-tunnelling system. Firstly, we derive the model of installed pipes from the character of the micro-tunnelling system. We calculate the parameter from the geotechnical engineering. Finally, the simulation experiments are performed in order to check the proposed model.

View full abstract

Due to rapid increase of vehicle use, construction of large scale parking lots is requested especially in urban area for smooth traffic. To improve the efficiency of available land use, some parking lots are designed as a hybrid type combining surface (flat) and tower parking. Design of such a kind of hybrid type parking is quite complex. Simulation methodology is useful to aid in design for such large scale parking lot. In this paper, a simulation method PLISM-1 based on discrete event model is developed with the concept of blocks and networks for solving this problem. By this simulation method, the evaluation of given parking lot plan can be done easily both from static and dynamic viewpoints.

View full abstract

In this paper, from the the theoretical viewpoint of analyzing the stochastic systems, the new statistical methodologies for actual random fluctuations have been proposed based on the inverse Gaussian distribution function. By considering the flexibility of the inverse Gaussian distribution function, the orthogonally expanded expression of the probability function is first derived by use of the Schmit's orthogonalization technique, whose expansion coefficients reflect hierarchically the lower and higher order statistics of phenomena. Next, by adopting the Bayesian viewpoint, a computer-aided algorithm for estimating the unknown state of energy stochastic systems under random measurement noise is newly established. Finally, the effectiveness of theoretical results has been experimentally confirmed.

View full abstract

This paper is concerned with the mathematical formulation of stochastic parabolic systems with boundary conditions of subdifferential type. There are two methods to formulate systems with subdifferentials. One is by a strong solution, the other is a weak one. The formulation by the strong solution is effective only for a one-dimensional spatial region. Since the spatial region of the considered system is a multi-dimension, only the formulation by the weak solution is applicable to the considered system. In this paper, first, the system equation is given and the physical meaning of the subdifferential boundary condition is explained. The definitions of the weak and strong solutions are stated. And the relation between these two solutions is derived. Secondly, the existence theorem of the weak solution is given. Finally, it is shown that the weak solution has the maximum solution.

View full abstract

We study the identification problem of a spatially varying discontinuous parameter in stochastic hyperboli equations. The consistency property of M.L.E. for a discontinuous coefficient is proved by using the method of sieves. A typical hyperbolic equation does not have a suitable regularit property of the solution with respect to a parameter to support the consistency property of the parameter estimate. Introducing the parabolic regularization technique, we show that the derived finite-dimensional M.L.E. converges to the infinite-dimensional M.L.E. under some conditions. Numerical examples are demonstrated to check the theoretical results.

View full abstract

In the control system design, the robust control theory has been studied actively. However, in order to apply the robust control theory to real plants, it is necessary to evaluate model uncertainty as well as a nominal model. In this paper, a new and practical identification method for the model uncertainty is proposed. The identification method is based on the ordinary least-squares identfication method in cooperation with the decimation. The key idea of the method is a frequency division by the decimation. The frequency band of the plant is divided into two parts,that is, one corresponds to the nominal model and the other corresponds to unmodeled dynamics. Effectiveness of the method is examined through actual data that is collected thorogh identification experiments of a ground-based test model for a large space structure (LSS).

View full abstract

This paper proposes recursive bias compensated least-squares algorithms for a continuous-time SISO system. Two continuous-time regression models are introduced based on the pre-filtered data obtained by state variable pre-filter and operater transformation methods. Using limiting behavior of LS estimates, algorithms for removing biases are derived. Simulation results are also included.

View full abstract

As is well-known, the ordinary regression analysis method is confined to a simplified linear model, an additive error of the estimation based on the Gaussian property and a least squares error criterion. However, it is often that the actual phenomena can not be exactly expressed by this simple model owing to various types of complex social and human factors. In this study, after focussing on an energy variable of acoustic systems, a new regression analysis method is proposed by introducing a multiplicative noise model suitable to the energy stochastic system. The effectiveness of the proposed method is confirmed experimentally by applying it to the actual acoustic data.

View full abstract

A morphological computation process is formulated for direct object coding. A non-linear field interaction scheme is invoked for associating spatial distribution of contour points with distortion distribution of morphological features on a system of diffusion fields. Through this association. the computation process induces a probability space on a class of random patterns. The computation process is autonomously adapted to patterns encountered in the search for a fixed point associated with a priori constraints on conditional pattern probability. Through this process adaptation, random patterns are categorized into a system of object codes to be manipulated by a symbolic reasoning scheme.

View full abstract

This paper presents a vector field matrix [2] derivation on a recursive algorithm of six degrees of freedom contour matching through single-eye imagery. We have already reported an algorithm in which the position and orientation of an object is determined by matching the reference pattern to the edge pattern on the image. However, in that algorithm, the vector field matrix was given intuitively. Here we actually derived the vector field matrix.

First, a mathematical model of the potential field is formulated by the stochastic evolution equality in Hilbert space. The existence and uniqueness properties of the solution to the potential field equations are studied. Second, we consider the cost function and derive a pattern matching algorithm suitable for computers within the framework of the gradient method. Third, we derive the vector field matrix for mathematical security of the algorithm. Finally, a representative time result of the experiments is shown, to support the theoretical aspects developed here.

View full abstract

In this paper, we present a method to obtain the restored image from the degrated image which applies the Gibbs distribution. Geman-Geman[1] proposed the Bayesian restoration of images based on the Simulated Annealing (SA). Their restoration method has a merit that can obtain the well-restored image. On the other hand, it has demerits which are a very slow convergence and a requirement of large computational cost, to get the Maximum A Posteriori(MAP) estimate of the original image from the degraded image. For overcoming such disadvantage of their method, we propose a deterministic rule of updating estimated gray levels of pixels and edge (line) elements based on a new energy function. Finally, we perform some simulations using real image data to show the feasibility of our method.

View full abstract

In this paper, we derive the optimal auxiliary input for fault detection of stochastic linear system in the frequency domain based on the time average of Kullback discrimination information between the system models corresponding to the normal and the fault modes. Auto-covariance sequence of the optimal auxiliary input is obtained by solving a mathematical programming problem. Numerical simulation results indicate that the proposed auxiliary input reduces the mean detection time without making much effects on the original system behavior.

View full abstract

The Robbins-Monro stochastic approximation procedure is a useful method for finding the root θ of the regression function f(x) = α recursively based on the noisy observations Y(x) of f(x). This procedure updates the estimate based on only the information of the current estimate, and hence the convergence to θ is generally slow. In this paper, a multi-stage stochastic approximation procedure is proposed to accelerate the convergence. The proposed procedure updates the estimate by using the information of the past estimates in addition to current one. This idea is also applied to nonlinear state estimation problem. Numerical examples illustrate its applicability.

View full abstract

In this paper, a higher order scheme of numerical approximation is presented for stochastic differential equations of Itô type, where a stochastic version of Runge-Kutta (RK) scheme has been developed to evaluate the coefficients of Itô equations. The iterative evaluation has been performed for both drift and dispersion functions. To achieve higher order accuracy the present RK scheme is organized by 4 kind terms. The numerical solution generated by the 4-stage explicit RK scheme has 1.5 order accuracy and the asymptotic efficiency. The results of simulation experiments are shown for supporting the validity of the approximation scheme.

View full abstract

In this paper, the author proposes a newly developed concept of multi-dimensional fuzzy random variables(MFRVs), which have intrinsically both properties of fuzziness and randomness and they are considered to be obtained as fuzzy perceptions of ordinary non-fuzzy random vectors.

First, the concept of MFRVs is introduced by using the set. representation method of fuzzy sets and some measurability properties are investigated. Secondly, applying the multi-valued logic, the expectations and the second statistical moments of MFRVs are considered and their reasonable definitions are proposed. Finally, the properties of set representations for expected MFRVs are investigated theoretically.

View full abstract

This paper studies associative memories using a dynamical neural network of the McCullough-Pitts model in which each class of given standard information patterns is assigned to a limit cycle. In particular, a method to construct a connection matrix of the neural network which assigns limit cycles as required is studied. Further, some examples are given to statistically evaluate the performance of the proposed associative memories.

View full abstract

In this paper, the identification of NARX and NOE models is studied. Parameters of a MLP ANN are estimated based on the RPE method. Four modifications for time-variant cases are considered; exponential forgetting, constant trace and bounded information algorithms, as well as the simple backpropagation algorithm. Several interpretations of the constant trace algorithm are examined, and a normalization scheme to simplify the comparison of the algorithms is presented. Algorithms are experimented using data from a simulated process and evaluated by their prediction capability with respect to the number of samples presented. Exponential forgetting shows best performance on the limited set of training data, whereas constant trace is regarded as most suitable for on-line identification purposes. NARX models show robust performance, although outperformed by NOE models in prediction capability.

View full abstract

In this paper, effects of random perturbations on chaotic systems are investigated. The stationary distribution of the chaotic system with additive random perturbations are theoretically discussed, concerned with a class of one-dimensional discrete dynamical systems and it is shown that if the distribution of chaotic solution processes is uniform, then, under the additive random perturbations with identically and independently distribution, the stationary distribution of randomly perturbed system is also uniform. Finally, two examples with numerical experiments show the validity and the usefulness of the results obtained in this investigation.

View full abstract

A computational method for the disbond detection in aircraft lap joints and in the adhesive joints between aircraft skin and reinforcing doublers is considered using thermographical data. An idea related to method of mappings applies to a parameter estimation problem for a 2-D heat diffusion system under noisy observations. A computational algorithm for estimating the length of disbond is proposed based on the domain decomposition method.

View full abstract

The main purpose of this paper is to provide an evaluation method associated with the use of statistical inference approach to policy rule and its simulation analysis. Namely, in stabilization policy problems done in the past, stochastic judgement is not developed in the adjustment between target and instrument vectors. In addition to the newly design analysis, we critically reconsidered methods and results of quantitative stabilization policy analysis in the previous works of combined control theory and macroeconomic policy problems. Before doing the above discussions, this paper also presents briefly expository implication and a configuration of structural analysis and policy design. Furthermore, we introduced the schematic diagram for discussing most aspects of econometric models investigating and investigated some evaluations of dynamic stabilization problems in linear stochastic models developed in the econometric and control engineering literatures.

View full abstract

This paper developed a new device which is able to identify the unknown water flow resistance of elastic tubes from a direct part of the measured water pressures by circulating pulsatile water flow in tubes with a roller pump. The device consists of two pressure sensors, A/D transducer board, one roller pump with constant revolutional speed, a low-pass filter to cut off high frequency noise, and microcomputer with a special software to identify the unknown parameters of the tube. The water flow resistance R is identified by the on-line nonlinear least squares method.

View full abstract

We present in this paper, a series of mathematic models for the performance evaluation of DNHR(Dynamic Non-Hierarchical Routing) networks with access control. Theoretical performance characteristics that show time congestion, call congestion and traffic congestion for call attempts between any pair of switching centers are determined.

View full abstract

While Kalman filter has found frequent applications to many signal processing problems, its variant namely, the square root information filter (SRIF) method that is more accurate, stable and efficient than the original Kalman filter algorithm received little attention. It is partialy because that standard SRIF requires matrix inversion and matrix multiplication which are computationally intensive. To solve this problem, a modified Faddeeva algorithm is utilized for matrix manipulation and it subsequently yields to a faster VLSI systolic arrays. In the paper, detailed analysis of SRIF is treated with an actual VLSI systoric array implementation.

View full abstract

In this paper we present two estimation algorithms of static relative positioning by GPS, where the vector between two antenna stations is determined. There are two important types of GPS observables; pseudo-ranges and carrier phases. The first algorithm here is to utilize the double difference of the carrier phases. And the second algorithm for relative positioning is shown based upon the difference of the estimates for point positioning by using the pseudo-ranges, specially, applying Dilution of Precision(DOP). Finally, The experimental results for estimation of the baseline length are shown by using real receiver data obtained at two static points.

View full abstract

This paper presents a practical method for simultaneously identifying a nominal model and its model uncertainty bound in frequency domain. We modify and extend the ordinary set-membership (SM) identification approach to a case in which an upper bound of the uncertain term is unknown and no prior information of the identified model is available. The upper bound of the model uncertainty in frequency domain is given by the proposed modified SM identification. The adequate low-pass filtering of the input-output data is effective to reduce the bias error in the estimated nominal model. It is clarified that the optimal band-width of the low-pass filter can be determined so as to minimize the estimated model error bound in frequency domain. The proposed scheme can be extended to the closed-loop identification based on the fractional representation.

View full abstract

In a previous paper the explicit solution of a Lyapunov equation of a single-input single-output discrete-time system was derived^[1]. In this paper, this result is used to introduce an explicit solution of a Lyapunov equation of a multi-input multi-output discrete-time system given by the matrix fraction description.

View full abstract

In the identification of system whose a priori information of structure is poor, adequate candidate models are difficultly set up. This paper presents the method in which candidate models are successively genarated and evaluated by using the search tree and then the optimal model may be obtained. The system described by a functional expansion may be identified, because the flexible search of models may be possible on the basis of hierarchical inference.

View full abstract

The paper develops an automatic measurement system which enables an accurate measurement of ship's attitude by use of four servo-type accelerometers appropriately located on the ship. Through the location of sensors, the heaving, rolling and pitching signals of the ship can be separated from each other with adequate linear combinations of the four sensor outputs. By introducing linear dynamic models of the three signals and using Kalman filters, on-line measument of the heaving, rolling and pitching of the ship is realized. A bank of Kalman filters are used to overcome an uncertainty of some parameters included in the dynamics.

View full abstract

The optimal transmission problem is considered for a set of correlated Gaussian signals through parallel channels with feedback. We assume that the signals are given by a solution of a multidimensional linear stochastic differential equation. Under a constraint on the total power of the signals, we will obtain the optimal gains for the set of channels which minmizes the steady-state estimation error.

View full abstract