Recently we have proposed the simplified measuring method of flour particle size at 50 percent of cumulative weight that is established for measuring the crushed particle size of wheat quickly and simply. Therefore, we have estimated the size of particle from the flow direction length. So, it was considered whether we could approximate the particle projection form with some models. In this paper, we propose the method that the weighting coefficient for the paxticle projection form, flow direction length and aspect ratio of flour is presumed with using the suitable oval model.

抄録全体を表示

Signal processing and computation of chemical sensor responses is an important step in recognizing the target gas because a chemical sensor is not usually specific to a particular kind of gas. In this paper we present a chemical-sensing system using the Learning Vector Quantization (LVQ) algorithm to classify dynamic sensor array responses. This chemical sensing system consists of an array of piezoelectric quartz crystal microbalance (QCM) sensors, each coated with a different plasma organic film. Each sensor's resonant frequency decreases as gas molecules are adsorbed on the film. Because the sorption process depends on the gas-film interaction, the sensors show various dynamic responses according to their target gases. We explain the mechanism of the sensor, providing the dynamic response model whose coefficients are used in recognizing the target gas. We show that the sensor response is based on the adsorption of gas molecules at the surface of the film and the subsequent diffusion into the film. In order to extract the qualitative information about the target gas, the sensor responses are described by auto-regressive equations. Their coefficients are estimated sequentially by the Kalman filter algorithm and transformed to form a pattern vector, which is immediately classified by the LVQ. Identification tests of 14 vapors are performed using the system installed in a personal computer.

抄録全体を表示

Recently, the distance measurement system that can measure a big object such as a clay model or a die at high speed, in non-contact and with high accuracy is demanded. The distance measurement system which used a laser and triangulation has been used for such distance measurement until now. But it was difficult to reconcile the large distance measurement such as about 100mm and high resolution such as about 10μm because large focal depth is incompatible with a small beam diameter in illumination optics of conventional distance measurement system. So we develop the distance measurement system that used axicon lens for illumination optics to realize large focal depth and small beam diameter. In this paper, it is reported about principle of an axicon lens, experimental result using axicon lens, method of distance calculation and experimental result using this system.

抄録全体を表示

When we apply impedance control to execute any tasks, it is very important how to decide the impedance parameter to realize the desired task. If we can extract the impedance parameter from human teaching data as characteristic of the human skill, it is appropriate to use it for control because of the similarity between impedance control and human fingertips control. However, there often exists an unevenness in time and space in human data. Modeling with Hidden Markov Model (HMM) is one of the promising technique to construct an efficient model for time-variant data including unevennesses. HMM is capable of characterizing a doubly stochastic process with an underlying immeasurable stochastic process which can be measured through another set of stochastic processes. Therefore, the probabilistic modeling of certain time series data which includes unevennesses caused by the human is possible. We propose a method to model the impedance parameter sequence identified from human teaching data with HMM in order to extract an essential discrete model which expresses the human skill. In addition, some applications of the obtained model to robot control and skill evaluation are proposed.

抄録全体を表示

High-speed three dimensional (3-D) shape measurement is becoming an important issue in many application such as robot vision and automatic inspection of manufactured parts.In recent years, many studies have been made on high-speed 3-D shape measurement methods. A slit ray projection method and a coded pattern grid projection method are common approach to high-speed 3-D shape measurement. The slit ray method takes a long image acquisition time because of scanning the slit over measuring object. On the other hand, the coded pattern method can reduce the image acquisition time in comparison to the slit ray method. However, this approach requires identification of coded pattern, and it takes long decoding time. We proposed a new coding method of multi-slit ray for high-speed shape measurement. The proposed method divides measuring area by using multiple sensors without light pattern. In addition to this, we proposed a method of expansion of measuring range in depth by means of secondary slit ray. The main slit ray is used to calculate the distanse to object, and secondary slit ray is used to identify the main slit ray. The projected angle of secondary slit ray is differed from that of the main slit ray. The combined use of main slit ray and secondary slit ray expands the measuring range in depth to three times in comparison with single use of the main slit. The proposed method eliminates repeated scanning of slit ray, and realizes zero decording time compatible with high reliability.

抄録全体を表示

A new particle sizing method using a dual-beam laser interferometer, which is composed of convergent and divergent beams, is presented. These convergent and divergent beams are used to illuminate a particle. When a particle passes through the intersecting region of these beams, interference fringes are formed by the scattered light from the particle, and the output signal is obtained from the light intensity passing through the slit put on the observation plane. As convergent and divergent beams are used for illumination, the wavefront curvatures of incident beams on the particle vary according to the particle position, and the spatial frequency of the interference fringes varies with the particle movement. The particle size is determined from its position where the visibility of the output signal reduces to zero. The experimental data obtained from glass spheres, whose diameters are approximately 100 to 200μm, are shown to confirm its operation. The measurement error of the particle size, which is caused by its passing position, is also discussed.

抄録全体を表示

By observing each output signal from experiments to recognize each of 6 gas sensors to measure the degree of indoor environmental pollution quantitatively, 7 kinds of simple gas, and one kind of complex gas were attempted. An independent recognition rule was established for each gas (8 rules). The result was determined from the position of a 6-dimension vector comprised of 6 sensor responses within the range determined by each rule, and a simple concentration index of the target gas was calculated from the input signal. As a result, 8 ranges were determined for each rule in 6-dimension vector spaces appeared to exist independently for the most part. However, negligible mistake in determination was confirmed in a low density area close to regular output in the determination experiment, information that can supplement the determination using a density guideline value should be obtained. The recognition rate without concentration correction was 99%. With concentration correction, 100% recognition rate could be obtained.

抄録全体を表示

In this paper, we study the nonlinear H_∞ control problem in robotic systems. Especially we consider the measurement feedback problem when velocitysignals are not available. So far, the solution of the nonlinear H_∞ state feedback problem in robotic systems has been proposed by B.S. Chen in 1994. We extend Chen's state feedback solution to the measurement feedback problem using only position signals. In this process, using Isidori's differential-game-like approach, we derive Chen's state feedback solution more easily and make the extension to the measurement feedback problem rather simple. And in the numerical simulation of 2-link manipulator system, we make sure that if disturbance attenuation level ‘γ’ is set appropriately small, the performance of our measurement controller is as equal as that of Chen's state feedback controller.

抄録全体を表示

The purpose of this paper is to synthesize the sliding mode control of high speed phase shifters in a power system. The sufficient conditions to reach a sliding plane are given by using an energy-type Lyapunov function. The control scheme is to regard a nonlinearity associated with the power torque-angle curve of AC generator as a variation of parameters based on a linearlized model of the system. By introducing the arbitrary large variation of parameters, a sliding mode controller is applied to the dynamical behavior of this nonlinear system. Moreover, using the Lyapunov function for reaching a sliding plane, the dynamic system becomes asymptotically stable, and thus, the proposed control method is robust control. The validity and usefulness of the controller are tested numerically on a single-machine infinite bus power system.

抄録全体を表示

Recently, it is considerd to be important to cooperate two manipulators effectively. In this system, problems related to handling, collision avoidance and so many others are must be considerd together. This paper proposes a method of planning joints paths for cooperative two manipulators. We consider one manipulator holds a object rigidly and the other executes specified tasks to the object. We give only relationship of endeffectors of these two manipulators, so each manipulator can get arbitrary configurations. In general, there are redundant degrees of freedom peculiar to the cooperative two manipulators. We use this redundancy and the potential function method to avoid collisions between the object and the link, the link and the link, the link and the obstacle, and so on. Finally, simulation results in 2D and 3D-workspace show the effectiveness of the proposed method.

抄録全体を表示

A regularization approach is investigated to determine an order of a transfer function model in the presence of input and output noises. By introducing multiple regularization parameters to the corrected least squares (CLS) estimation, we clarify that the system order can be determined by comparing each optimal regularization parameter, which minimizes the mean squares error (MSE) of the estimate, with a corresponding eigenvalue of the covariance matrix of the input-output data. An asymptotic MSE is given, and an analytical expression of the optimal regularization parameters is clarified. An iterative algorithm is proposed by using the only accessible input-output data, in which the system model order and the noise variances can be calculated simultaneously.

抄録全体を表示

Quadratic Hurwitz stability is discussed for interval polynomials. Taking cues from the weak version of Kharitonov's theorem, we pose two questions: 1)does Hurwitz stability of interval polynomials imply quadratic Hurwitz stability? and 2) if not, is there any case where this implication holds? By focusing on interval polynomials of degree two, we show two examples, one of which negates the question 1) and the other supports the question 2). Based on these facts, we further discuss several aspects of quadratic Hurwitz stability property in the polynomial coefficient space.

抄録全体を表示

This paper describes a method for the design of discrete-time model matching control systems for a multivariable plant with bounded measurement disturbances. This method uses the free parameters of the controller. Simulation results are presented which confirm the validity of the design method.

抄録全体を表示

In this paper we first derive a nonlinear dynamical model for the control of overhead traveling cranes, which make three-dimensional motion. Based on the derived nonlinear model, we calculate the optimal control such that the swing of the load during transfer is minimized and that the given terminal condition is satisfied at the end of transfer. We use Sakawa-Shindo algorithm for calculating the optimal control. The algorithm was derived on the basis of Pontryagin's maximum principle and improved subsequently for bettering the convergence of iterative computations. Also, the algorithm guarantees monotone decrease of the cost functional at each iteration. It will be shown that the algorithm works well and yields the successful results.

抄録全体を表示

The model following control system (MFCS) gets output signals to follow desired ones. Recently design methods of the MFCS for a mixture of a nonlinear part and a linear part, containing inputs in nonlinear part, using stable zero assignment or nonlinear for inputs were investigated. This paper shows, a design method of the MFCS for general nonlinear system, that is nonlinear system for internal states and inputs. We define new internal states with containing internal states and inputs, and new inputs which are formally linear. The Boundedness of internal states can be shown by obtaining a whole state equation which inclues state variable filters and ex-amining a time derivative of a quadratic form of the whole state equation. If the following five conditions are satisfied, internal states are bounded. (1) A derivative from inputs of the state equation by an output matrix, and an input matrix of new inputs is constant. (2) An invariant zero polynomial of linear part is stable. (3) A positive real condition of the transfer function matrix is satisfied. (4) An inner product condition of the nonlinear function and states is satisfied. (5) A norm condition of the nonlinear function is satisfied. A genetic algorithm (GA) seeks the design parameters which satisfy the five conditions.

抄録全体を表示

This paper proposes a hybrid quasi-ARMAX modeling and identification scheme for nonlinear systems. The idea is to incorporate a group of certain nonlinear nonparametric models (NNMs) into a linear ARMAX structure. Particular effort is made to find a better compromise to the trade-off between the model flexibility and the model simplicity by using knowledge information efficiently. As the result, we obtain a model equipped with a linear. ARMAX structure, flexibility and simplicity. The effectiveness and usefulness of the proposed hybrid model are examined by applying it to identification of a variety of nonlinear systems.

抄録全体を表示

Exact model matching method can not be used for a system having unstable zeros since unstable pole-zero cancellations occur in the overall systems. In this paper, we will propose an approximate cancellation of anti-stable zeros instead of exact cancellations. For this, we will show a design method of approximate polynomials for the inverse of anti-stable polynomials using Taylor's expansion. Then, we will show a design of model following control systems for a plant having unstable zeros based on the pole placement scheme.

抄録全体を表示

In this paper, we consider an identification and a validation/update problem for a special class of model sets, called the Smallest Model Set. The sets consists of SISO discrete-time transfer functions with time-varying parametric uncertainties measured by the weighted ∞-norm. We will analyze the formulated problems by using a concept of a distance in a parameter space. The identification and update problems are shown to be reduced to a convex optimization problem, and the validation problem becomes a comparison of distances. In particular, the update problem is shown to have an explicit solution in a special case. Based on the results, we propose a validation/update algorithm in an off- and an on-line case, respectively.

抄録全体を表示

Recently, a design of model matching and adaptive control systems under the presence of measurement noise is reported by selecting a degree of characteristic polynomial of state observer. But this methods can be only used for the case where the plant has a stable inverse. In this paper, we will show that the idea is also applicable for a plant with input noise. This method can be carried out by simple calculation comparing with the design of Kalman filter or H_∞ control. So the proposed method is adequate for an adaptive control in which the controllers must be calculated on line.

抄録全体を表示

This paper investigates an algorithm to obtain a model set that includes all possible transfer functions for a given plant to be identified. Specifically, efficiency of this identification algorithm is characterized by a diameter of an obtained model set and properties of the diameter are studied. The study shows that a model set given by the algorithm of Zhou and Kimura, which is the first algorithm having the above feature, does not shrink in the sense of a diameter even if the length of a given input-output time series increases. In order to resolve this difficulty, extension of their algorithm is proposed. Moreover, for the noisy case, it is shown that a long input-output time series does not always result in a small-diameter model set. An optimal length of the series is obtained based on an upper bound of the diameter. Simulation results are given for illustrative purposes.

抄録全体を表示

In this paper, we discuss H_∞ control problems for descriptor systems via Linear Matrix Inequality (LMI) approach. First we give a stability condition that means exponential stable and impulse free for descriptor systems. And Bounded Real Lemma that contains stability and norm condition for descriptor systems can be expressed as matrix inequalities. Using these conditions, we give LMI conditions for the solvability of H_∞ control problem for descriptor systems. The results are natural extensions for a state space expression case. We also show the construction of proper controller and a relation of the controller order and solutions of the LMI.

抄録全体を表示

This study proposes an adaptive control method for obtaining a model and control for a nonlinear motion system. Since the motion systems driven by electromagnetic force have strong nonlinearity, it is neccessary to obtain the model of the nonlinearity to achieve an objective motion control. The control architecture consists of a feedback system to stabilize the motion system, learning controller to obtain the model of nonlinearity and to tune the feedback gain to linearize the motion system and inverse dynamics to offset the the motion system to make motion trajectory coincide with the objective trajectory. The obtained model is described as series of weighted basis functions, and the explicit learning rules of weights are derived by using Popov's hyperstable theory. The tuning rules of feedback gains are determined by using the obtained nonlinear models. Further, this paper shows the high performance of the proposed method by demonstrating an experiment of the tracking desired trajectory in the electromagnetic levitation and positioning system.

抄録全体を表示

It is useful for control engineers to initialize automatically the tuning parameters of adaptive control by using a priori measured data without any trial and error.
The adaptive law using time varying gain matrix has better convergence rate than the one using fixed gain matrix. In order to be sensitive to characteristic change of control object, it is useful to use variable dead zone width which changes according to magnitude of disturbance.
In this paper, we propose a scaling type law having variable dead zone width and automatic initialization using a priori measured data. We prove a theorem that gives some properties and convergence of the algorithm, and illustrate how to determine the variable width by using data obtained from a priori measurement and at each iteration instant. The variable width is yielded by αd₀ (k), of which constant α and time variant d₀ (k) are respectively determined by a priori measured data and data calculated at each iteration.
Furthermore, we apply the above algorithm to identification of an electro-hydraulic servo system and verify the effects.

抄録全体を表示

The head positioning servo control on magnetic disk drives has been required to achieve both precise positioning and fast access. To satisfy these requirements, a Mode Switching Control (MSC) structure that includes several different structure controllers and switching functions has been widely used. Because the servo controllers can already be designed optimally, the only remaining issue is “switching”.
The MSC is a kind of Variable Structure Control System with switching in only one direction. This means that the servo mode is handed over from one mode to another during a single servo action. Thus when the mode is switched from seek mode to track-following mode, the switching time is the initial time when the track-following mode begins. As a result, Initial Value Compensation (IVC), a method that takes into account the initial state values, can be applied. The authors previously proposed the design for the IVC and showed its effectiveness in disk drive servo controls.
In this paper, we propose a design method for determining the optimal combination of initial states for mode switching. Here the H₂-norm of a transfer function between initial values and control input, is defined as the performance index of the transient response after mode switching. By combining the IVC method with this method of determining the optimal switching conditions, it can be said the MSC design methodology has been completed. Several experimental results show the effectiveness of this method.

抄録全体を表示

Generally, the conventional ship's inertial navigation systems (S-INS) needs about 4-hours to settle the initial alignment under movement of ship. This paper presents an effective approach to the rapid initial alignment on a strapdown S-INS using three fiber-optic gyros and three conventional accelerometers.
In this approach, a modified least square filter as pre-filter of the measured data disturbed by movement of ship is used to improve the accuracy and settling time of initial alignment.
It is confirmed by simulation that the accuracy and the settling time (convergence time to meet the required accuracy) are improved by using this approach.

抄録全体を表示

This paper describes the chaos-fractal analysis of moving gas fire flames observed by a fast image sensor(500 flames/s) and NTSC video camera. The analysis, which is extended to two and three dimensional cases, discloses that the movement is chaotic if it is observed within the Nyquist condition but that it results in such false patterns as 1/f or 0/f near the Nyquist due to probable aliasing.

抄録全体を表示

In this paper, the domestic gas discrimination system is proposed for practical use. The system is constructed by using a digital signal processor, and is able to use as an real-time, stand-alone system. The commercial base semiconductor gas sensor TGS823 is used and 3 gases i.e. inflammable gas, alcohol and cigarette smoke, are chosen as the subjects of the discrimination.
The results of discrimination experiments are shown to demonstrate the effectiveness of the proposed system.

抄録全体を表示

This paper presents a method of extracting rules from neural networks trained using a random optimization method (ROM) with deterministic mutation (DM). The DM is performed on the basis of the result of neural network structure learning. The ROM with the DM is utilized to reduce the number of network connections for iris data. The network connections survived after training represent rules to perform pattern classification for the iris data. The rules are then extracted from the neural network in which hidden units use signum output functions to produce binary values. It enables us to extract simple logical functions from the network. Simulation results for the iris data show this method can generate simple rules compared with conventional methods.

抄録全体を表示

The Cerebellar Model Arithmetic Computer (CMAC) proposed by Albus is a neural network that imitates the human cerebellum, and it is basically a table look-up technique for representing nonlinear functions. The CMAC can approximate a wide variety of nonlinear functions by learning. However, the CMAC requires a number of learning iterations to achieve the adequate approximation accuracy, because the CMAC learning algorithm is derived based on the Least Mean Square method.
In this paper, a CMAC learning algorithm based on the Kalman filter is proposed. Using this algorithm, the number of learning iterations is reduced, and the comparable approximation accuracy is achieved as compared to the system using conventional learning algorithm. Generally, the learning algorithm based on the Kalman filter requires much larger computational quantity than that required by algorithms based on the Least Mean Square method. For CMAC system equations contain a sparse matrix, the computational quantity of the proposed learning algorithm can be reduced. Furthermore, since two CMAC weights being in the far place from each other can be considered to have little correlation, the number of weights that should be updated for an learning iteration can be reduced. Computer simulation results for modeling problems are presented.

抄録全体を表示

This paper proposes a constraint oriented approach to solve practical multipurpose machine (MPM) flow shop problems with changeover time and limited transportation capacity. In practical MPM flow shop problems, the transportation capacity is limited. Owing to this limitation, interference phenomena peculiar to complex production lines occur, thus decreasing production capacity. The algorithm proposed in this paper solves practical MPM flow shop problems by explicitly using constraints to avoid interference. The validity of this method is confirmed by discrete simulations of a section mill finishing facility line. Furthermore, the method is applied to an actual section mill finishing facility line.

抄録全体を表示

Neural networks have been often used for pattern recognition since the error back-propagation algorithm (BP) has been proposed. One of difficulties to use the network is how to design the structure of the network. For this difficulty, various structural learning algorithms have been proposed. It, however, is considered that these learning algorithms based on BP are only local search. Another way is genetic algorithm (GA) which is good at search of large space since it uses many populations. It, furthermore, is not necessary for the fitness function to be continuous and differentiable. The object of this paper is to design the structure of the layered neural network using GA. We propose new coding method and fitness function. The proposed method is applied to various Boolean problems and acoustic diagnosis of compressor. The results shows that the proposed method is effective for designing the structure of the network.

抄録全体を表示

In this paper, a new random optimization method is presented. The proposed method is called RasID which is an abbreviation of Random Search with Intensification and Diversification and it can search for a global minimum based on the probability density function of the searching, which can be modified using information based on success or failure of the past searching to execute intensified and diversified searching. A principle of the RasID is such that searching for the parameters is strengthened when the probability of finding good solutions is high, and is weakened when the probability is low. In other words, when the searching is success, the probability density function is modified so that the search is done within a short range to the direction of decreasing the criterion function in order to achieve a certain improvement of the criterion function. On the other hand, when the search is failure, the searching range is expanded and the direction of the searching is adjusted to include even those which make the criterion function worse in order to search the parameters with a wide range. The RasID is superior to the commonly used random search method because the RasID is a kind of adaptive random search utilizing the past experiences of the searching as mentioned above. From the simulation results of realizing many kinds of nonlinear functions and controlling a nonlinear crane system using Neural Networks (N.N.), it is clarified that the RasID is superior in performances to the Gradient Method.

抄録全体を表示

In this paper, a global optimization algorithm of the unknown multimodal objective function under noisy observations is proposed. Our algorithm is constructed based on the learning performance of the variable hierarchical structure learning automata. And the numerical experiments are carried out in order to test the efficiency of the proposed algorithm. The results obtained imply that the proposed global optimization algorithm is useful for finding out a global minimum of the unkonwn multimodal objective function under noisy observations.

抄録全体を表示

A new error back propagation method which is called EBP learning algorithm has been proposed by the one of the authors for a supervised learning of multilayer neural networks. This method is free from a gradient method which is used in the well known BP method or its improved versions. The method is composed of the following two stages.
A) Error back propagation: Determination of fictitious teacher signals of hidden layers from the output error of a neural network in the backward direction.
B) Weight parameters determination: Weight parameters of a neural network should be determined to minimize the fictitious error using fictitious teacher signals of each hidden layer.
An exponential weighted least squares (EWLS) method which is well known in the field of signal processing is proposed in this paper to use for a determination of weight parameters of B). Simulation results using a proposed EBP-EWLS algorithm for a learning of a sinusoidal function are presented. The results show the advantge of learning numbers for success and 100% convergence rate for initial values of weight parameters of a neural network.

抄録全体を表示

This paper describes disparity of driver behavior, which is a cause of traffic accidents, and proposes a method of quantification of the disparity with entropy. It analyzes disparity of driver behavior approaching a signalized intersection during a yellow signal period with data of positions and speeds of vehicles as well as distinction of stopping at or passing through the intersection, captured with a roadside vision-based instrument. The data show that there is a region of disparity, where vehicles of the same speeds at the same positions may stop at the intersection or may pass through it. The disparity region is identified as an intersection of two regions, each of which is defined with two dimensional Gaussian distribution for stopping vehicles versus passing vehicles. The disparity region is closely related to the conventional theoretical dilemma and option zones. The disparity in the region is quantified with entropy with respect to the speeds and the distinction. The analysis with data captured in Tokyo area suggests that disparity in the distinction is more dangerous than that in the speeds if a value of the entropy corresponds to a degree of danger.

抄録全体を表示

We can obtain the positional information of a pillar to a robot using only one image by CCD camera. In the process, the vertical line of a pillar is used in its Hough transformation and the position of an end-point of the image and an approximated method of computation are applied. The results of some experiments are shown.

抄録全体を表示

The present paper considers robust servo problems where time-varying uncertainties are involved. Taking an augmented system, we apply a generalized version of the guaranteed cost control. The servo controller is expressed in terms of a solution to the associated Riccati inequality.

抄録全体を表示