Previously, we have proposed an algorithm to reconstruct the supremally reconstructable parts of the input and the state of a linear timeinvariant system S based upon an output information alone, and simultaneously the algorithm has introduced the so-called “contracted system” for the system S. The main purpose of this paper is, subsequently, to develop an algorithmic method constructing two “partial inverse systems”, called as an input partial inverse system and a state partial inverse system, that reconstruct the input and the state as precise as possible, respectively, in the case of the knowledge on the initial state of the system S also available. The method follows along the line of applying effectively the contracted system of S so that minimal order partial inverse systems are presented and their fundamental properties are investigated. Finally, a numerical example is given to illustrate the algorithmic procedure developed in this paper.

抄録全体を表示

In this paper it is attempted to identify more general group utility functions based on the concept of convex dependence, which is an extended concept of utility independence in multiattribute utility theory. Here the number of the member in the group is assumed to be two. It is shown that the convex dependence could play an important role in the group decision or societal decision, since under the concept of convex dependence it is possible for each individual (member of the group) to assess the different preference structures depending upon the utility level of the other individual. These preference structures can be expressed as the conditional normalized utility functions which are expected to be assessed by each individual from the standpoint of attaining better society. The group utility function is then identified by aggregating the conditional normalized group utility functions of the individuals where the existence of a benevolent dictator is postulated.

抄録全体を表示

This paper is divided into two parts: (i) to give a mathematical description for a class of nonlinear MA models, which is expected to be a good basic model in case of fitting the data whose outstanding feature lies in sporadically large values such as observed in geo-physical, turbulence data, etc.; and (ii) to propose a method for identification of unknown parameters in the proposed nonlinear MA model.
First, a class of nonlinear MA models is proposed whose nonlinear terms are described by a set of Hermite orthogonal functions. Secondly, a method is presented for estimating unknown parameters by using the moment method, and asymptotic properties of the proposed estimators are mathematically investigated. Finally, salient features of the proposed nonlinear MA model and the estimators of unknown parameters are discussed from both qualitative and quantitative points of view.

抄録全体を表示

This paper is concerned with synthesizing, from a standpoint of so called “indirect control approach”, a new scheme of designing an adaptive control for a class of a single-input single-output linear discrete systems in the model reference adaptive control problem. Employing Gram-Schmidt orthogonalization technique, an adaptive identifier is constructed to generate both the estimates of the parameter and the state of the plant, but requires no more knowledge than the input-output signals as available measurements from the plant. Those estimates are proven to converge to the real values of the plant without errors in finite discrete time and are used to change in each step the parameters of a compensator connected ahead with the plant in a cascade. The compensator yields an adaptive control to the plant and is supplied with a reference model input in addition with two different signals. One is produced by feeding back the state estimates of the plant and another by feeding back the state of the compensator. When the identification process completes, the revised close loop transfer function of the plant coincides with the reference model one, namely at this time, the exact model matching is phased out clearly. After then, the output error between the plant and the model converges to zero asympototically.

抄録全体を表示

A new method for designing multivariable model reference adaptive control system is presented. In many adaptive control approaches, at least the following two assumptions regarding the plant have to be made. (1) The number of plant poles and zeros is known. (2) The plant is the minimum phase. The proposed method can avoid these restrictions by using the adaptive controller designed for an autoregressive model with dead time, based on Lyapunov's direct method. Next, in order to illustrate the effectiveness of this model for the both minimum phase and non-minimum phase plants, the results of computer simulation are given. Finally, the method is successfully applied to the real plant testing the performance of refrigerant compressors.

抄録全体を表示

The development of theoretical approach to a large-scale, complex and multi-objective system synthesis is drawing increasing interest in various fields of engineering and other areas. In this paper, the synthesis problem of a class of multi-objective systems is formulated as a fuzzyy scalar problem introducing a concept of degrees of satisfaction.
For the problem, a method of obtaining an optimal candidate empirically is given on the basis of the fuzzy finite state system, by which a heuristic synthesis method for the multi-objective system is proposed.
An example is illustrated to show how the proposed method works.

抄録全体を表示

To make the measurement with high accuracy by a two knife-edge direct reading balance, its built-in weights must be calibrated at reasonable interval. But the usual calibration of built-in weights needs a number of highly accurate standard weights corresponding to its built-in weights or requires complicated procedures both in measurements and calculation depending on different combinations of the built-in weights of the respective types of balances.
We have developed a calibration method of direct reading balance based on built-in weights by means of the difference equation of 1st order. This method can be commonly applicable to all types of the direct reading balances, independently on the combination of the weights. Its main features are:
(a) The relative calibration of the weights can be easily made by using a difference weight of unknown mass and a set of rough weights at every figure of the balance indication.
For the absolute calibration, only one standard weight is needed.
(b) The precision of the calibration is estimated within 1.6 times the repeatability of measurements at one figure of the indications.
(c) The calculation for this calibration is quite simple.

抄録全体を表示

Wiegand wire generates a sharp voltage pulse in a coil wrapped around the wire when an applied magnetic field crosses a threshold intensity. Few papers have reported the characteristics of the output pulse from the point of practical applications.
Detailed study have been performed on the frequency characteristics of the amplitude of output pulse and of the threshold intensity of the applied magnetic field.
Although the amplitude of the pulse and the threshold value of the applied field depend to some degree on the exciting frequency, the experiments show that the fluctuations of the amplitude of the output pulse and those of the threshold value of the applied field are reduced when the maximum intensity of the applied field is three times as much as the threshold intensity of the field.
According to the results, the angular accelerometer using Wiegand wire has been made, which can measure angular speed and acceleration within an error of 1% in the wide range of angular speed.

抄録全体を表示

A new method of non-contact type temperature measurement of the running thin lines (e.g. yarns, fibers, and wires) is developed.
In this method, Scanning allows line vibration.
The measuring error caused by the emissivity which is less than unity doesn't occur by use of Zero-method.
The high temperature resolution and the narrow field of view of the radiometer increase the accuracy of measurement. The radiometer isn't required to have high stability.
A temperature controllable radiator, which is made with the same material of the running thin line, is settled at the back of the line. In a scan across the line, one pulse is produced as a radiometer output which level is proportional to the temperature difference between a line and a background radiator. Temperature of the background radiator should be controlled to fade the pulse. When the pulse disappears, temperature of the line is equal to that of the background radiator.
In the experiments, the radiometer whose spectral band is 8∼14 [μm] has 0.01[K] temperature resolution and 40 [mrad] angular field of view, and is settled at a distance of 1[m] from the 1.06[mmφ] wire. For a high emissive object (ε=0.955), the measuring error is ±0.2[K]. For a low emissive object (ε=0.204), the measuring error is ±1.0[K] in using a domy radiator which makes environmental temperature uniform.

抄録全体を表示

This paper describes an experiment performed to clarify the meaning of psychological goodness of shapes, which is expected to be an useful tool in a heuristic method of computer vision. Factor analysis is carried out to find main factors representing geometrical features of random shapes used in our experiment. A psychological experiment is then performed to determine the order of goodness of these shapes. The result of this experiment is related to the geometrical factors drawn from the factor analysis. Finally, we obtain the factor ‘Simplicity’ representing goodness of shapes and four geometrical indices which are closely related to the factor.

抄録全体を表示

A fiuidic sonic oscillator can be used not only as a pulse generator but also as a temperature sensor and a flowmeter. For practical applicatios of them, the oscillator suited to each application must be designed. In its design we must select the combination of a wall attachment device with a connecting line which can oscillate and its frequency satisfies the design requirement. Since the calculating method of the geometrical configuratins of the device necessary to construct the oscillator has already been proposed by the author and others, this paper deals with the frequency. It has been investigated in many before studies, but its calculating method properly considering the dynamic characteristics of the device has not been reported yet.
This paper proposes the calculating method of the frequency based on the analysis of the dynamic switching in the oscillator. The results obtained can be summarised as follows:
1) The contacting-both-walls type dynamic switching is analized theoretically by introducing the switching criterion. By this analytical method it becomes possible to predict the minimum switching control pulse width and so on for a given device.
2) The calculating method based on the above analysis can express the influences of the geometrical configurations of the device and the connecting line length and diameter on the frequency. Therefore, by this method the constructing parameters of the oscillator satisfying the characteristics required can be predicted without an experimental trial and error.

抄録全体を表示

This paper is concerned with theoretical analysis on the mathematical modeling of cerebellar functioning in human motor control system. Discrete type equations of the motion of two joints arm system were derived, and a new algorithm was formulated, for estimating the motor control system dynamics. The motor command was self-adaptively generated according and algorithm.
The learning processes of our cerebellar functioning model via practices were experimented through the simulated computer analysis. Then, it was verified that the pursuit pattern of the two joints arm system, by our modeling and algorithm, to the desired movement was steadly improved as the learning stages advanced. Furthermore, it was also verified that our modeling and algorithm responded optimally to the unexpected environmental disturbances.

抄録全体を表示

The aim of this project is to propose the concept of master/slave control with compensation for object motion to facilitate the manipulation of a moving object and to evaluate the compensation by experimentally comparing the performance of the operator.
An experimental system was built which consists of a master/slave manipulator, a moving table for the moving object, and a computer which controls both the manipulator and the table. A computer control scheme for master/slave and object motion compensation was developed in consideration of the kinematics and dynamics of the manipulator. The computation time in this scheme proved practical and permitted a system sampling of 50ms.
Experiments were carried out with human operators performing manipulation tasks in computer controlled master/slave. Their performance was compared in three situations: no object motion, compensation for object motion, and no compensation. The comparison of the compensation and no compensation situations showed that the compensation reduced the operation time by 26-41% in the peg moving task and increased the accuracy by two and a half times in rectangle tracing. In valve turning, however, a significant improvement was not observed.
Thus, it was concluded that compensation for object motion can improve the performance of the human operator significantly in certain kinds of tasks.

抄録全体を表示

A fast method for arithmetic modulo a polynomial over GF(2) is proposed. In this method, any polynomial can be calculated modulo any polynomial over GF(2) at high speed, while no restrictions are imposed on the dividend or the divisor polynomial. The calculation time is shown to be proportional to the logarithm of the order of the dividend polynomial.
The fast method is compared with the conventional division algorithm in terms of the calculation time, so that it becomes clear that the fast method is much faster than the conventional one.
The application to a delayed version of msequence of high degree is demonstrated.

抄録全体を表示