The motion and energy produced by the human body during sport are important factors in estimating athletic ability and in providing useful feedback in training. However, previous methods of measuring body motion and energy production have been too cumbersome to be used in daily training.
In this study, an accelerometer was used to estimate the motion and energy production of the body during rowing. The accelerometer, capable of measuring body acceleration in three dimensions, was worn on the back by an oarsman who carried out simulated rowing movements using a rowing ergometer. The rate of acceleration of the body during rowing showed distinct characteristics from which the body motion could be roughly estimated. The waveform of the acceleration made by a skilled oarsman was remarkably different to that made by a semi-skilled oarsman. Energy production during rowing was estimated by integrating the absolute value of the acceleration. The estimated results agreed well with power production gained from the performance monitor of the rowing ergometer.
Measuring acceleration is a simple and valuable method of estimating motion and energy production during sporting activities. It can also be used in biomechanical and kinesiological investigations.

抄録全体を表示

A high-resolution ultrasonic thickness measurement system for a thinner specimen has been developed. Thickness is obtained as the reciprocal number of peak frequencies which relate the period of ultrasonic multiple echo reflected from the specimen. By using FFT (Fast Fourier Transform) for the calculation of the frequency spectrum of the multiple echo, the resolution can be made easily higher. The resolution reaches 0.1μm at 0.25mm thickness measurement. Moreover, thickness can be measured in the case of a dull echo wave form because extraction of the peak frequency is easy.
In this paper, the principle of the system is studied first. The results are given below.
1. The frequency-based technique can measure a few times thinner thickness compared to the conventional time domain technique.
2. The sampling rate for analog to digital conversion of the received signal can be much lower than when the time domain technique is used.
3. Accuracy of the measurement is influenced by ultrasonic attenuation of the specimen.
Next, the experimental system is developed. To shorten the FFT calculation time, the sampling rate and the FFT points are lowered using a frequency conversion technique. The experiments show that the system measured a 0.25mm thickness with 0.1μm resolution using a 20MHz ultrasonic probe. In addition, in the case of a dull echo wave form reflected from a corner, thickness was measured.

抄録全体を表示

This paper discusses the fixed-endpoint optimal regulator for a single-input, linear, time-invariant, discrete-time system. The dynamic state feedback controller in the z-domain is induced from the fixed-endpoint optimal regulator which can be expressed by the state feedback with time-varying gains in the t-domain. It consists of the constant parameters which do not depend on the initial state of the system. The system is stable, and the state settles to zero within a finite time.

抄録全体を表示

Model matching is one of the direct methods by which the characteristics of a controlled system can be adjusted to the desired values. It is well known that poles can be assigned by state feedback, but zeros are invariant under state feedback. So, we have difficulty in constructing a model matching system if the system has unstable zeros.
By using two-delay input control and state feedback, Mita et al. has shown that invariant zeros can be assigned and a model matching can be realized.
In this paper, we give a different proof to the necessary and sufficient condition that the invariant zeros can be arbitrarily assigned. Through this proof, it is made clear which element of the feedback gain relates to zero assignments. It is also shown that invariant zeros and poles can be arbitrarily assigned if and only if the two-delay input control system has no zeros. By the proof using coordinate transformation, the straightforward algorithm is obtained for the determination of feedback gains for model matching.
Zeros are assignable under the same condition, if the state feedback is replaced by an observer and its feedback.

抄録全体を表示

This paper proposes a design method of a discrete time LQI control system by using a reference model. For the continuous time system, a design method of the weight matrices of the index performance for an LQI control system by using a reference model has been proposed, and this method enables us to design an LQI control system with desired transient responses easily. This method is extended to the discrete time LQI control system. But the relative order of a discrete time plant often is unity and this fact makes it difficult to transfer the continuous time method to the discrete time case directly. Therefore not only the weight matrices of the index performance but also the control system structure are designed by using a reference model. First, the design method for an SISO system is proposed, and this method is extended to an MIMO system. Numerical examples are presented to show the effectiveness of this method.

抄録全体を表示

Model reference and pole placement adaptive controls are typical and reliable schemes of present adaptive control, but these are difficult to be applied to multiinput and multioutput (MIMO) system since too much time is required to compute their complicated algorithms.
Simple adaptive control (SAC) is useful for practical implementation, especially for MIMO controlled systems, because of simplicity of its structure. However, its conventional discrete-time algorithm has theoretical ambiguity in the delay of computational time and the feedforward element inserted to make the controlled system almost strictly positive real (ASPR).
This paper proposes a new discrete-time algorithm, which removes the above ambiguity. The defects of the conventional algotithm are first explained and the fact is analysed that the insertion of the feedforward gain is equivalently transformed to the change of adaptive gains in the algorithm. The error equations are derived and verified to be ASPR. The stability of the algorithm is proved by using Lyapunov function and Kalman-Yakubovich lemma.
Finally, a simple example is simulated to confirm the validity of the algorithm.

抄録全体を表示

Since an interactor is defined as a polynomial matrix which cancels all zeros at infinity of a given plant transfer matrix by multiplying from the left, an interactor can be regarded as an alternative representation of the structure at infinity of a plant. This implies that there is a direct relationship between the structure at infinity of a plant and the structure of degrees of its interactor. In this paper, we will discuss this relationship. For this purpose, we define a regular interactor by an interactor whose row degrees coincide with the multiplicities of zeros at infinity of a plant. Then, it will be shown that an interactor is regular if and only if it is row proper. We will also give a procedure for calculating a regular interactor from a given transfer matrix.

抄録全体を表示

Space robots, which consist of a satellite base and a manipulator mounted on it, are expected to perform various tasks for construction and maintenance of space structures. Since the angular momentum of the space robot system is conserved, the motion of the system is subject to nonholonomic constraints. When the manipulator makes motion from an initial position to a desired position, variation of the base orientation depends on the trajectory of the motion owing to the nonholonomic property of the system. Since the satellite base is desired to have a constant orientation, we seek such a trajectory of the manipulator that a given motion is attained, that the base orientation is unchanged before and after the motion, and that the integral of the sum of squares of accelerations of the joint angles during motion is minimized.
First, the equations of motion of the space robot are derived, where the orientation of the satellite base is expressed in terms of Euler quaternions. We parameterize the motion of joint angles by expressing it in terms of the truncated Fourier series, and a nonlinear programming technique using sequential quadratic programming algorithm is applied to determine the optimal coefficients of the Fourier series. Some results of the numerical computation are shown.

抄録全体を表示

In previous paper, we extended the designing method proposed by Keel et al., which originated from robust eigenvalue assignment, to a system with time-varying uncertainties which are exhibited by the linear combination of time-varying parameters and time-invariant matrices specifying the structures of uncertainties. Moreover, we proposed a novel algorithm to obtain a state feedback that achieved a robustly stable closed-loop system.
In many practical cases, it may be impossible to measure all of the state variables in the system. It shall be required to stabilize the system by use of output feedback controller.
This paper presents a new method for robust stabilization of the above-mentioned system by means of output feedback controller with low-order compensators, in which the closed-loop eigenvalue assignment is relaxed, that is, discrepancies between the desired and the achieved closed-loop eigenvalues are permitted. Based upon the Gronwall's lemma, a sufficient condition with respect to the uncertainty tolerance is also prepared for the asymptotical stability of the closed-loop system. Furthermore, we state the relation between the robustness of the stability and the order of dynamic compensator. For example, designs of lateral autopilot system of a missile are shown.

抄録全体を表示

This paper proposes an approach to the quantitative evaluation of robotic compliant motion tasks based on the power consumed by interaction between parts handled by a robot and parts in a task environment. Power is defined as instantaneous work, and in a task such as assembly where a moving parts grasped by a robot makes contact with a parts in a task environment and acts against friction during the contact, it results from the work lost to friction. As an application of the technique for evaluating tasks according to power consumption, a method for tuning the compliant motion parameters in a robotic task is discussed. This method is an experimental approach to motion control planning, whose aim is to obtain optimized motion parameters for a task. It employs the response surface method as a schema for experimental design, because of the method's similarity to the process whereby humans tune parameters. The response surface method is known as an effective means of incremental local optimization. Finally, to confirm the effectiveness of evaluating tasks according to power consumption and to test the mechanism for tuning motion parameters, the proposed tuning method is applied to a plug insertion task. This is a very interesting task in the sense that it is not easy to determine whether a robot has completed such a task from force and position information, even though the task requires only simple operations. If the work done during the task process is used as a task evaluation index, the task can be easily validated, and the motion parameters can be tuned in the same way as by a trained human operator.

抄録全体を表示

The authors proposed the automatic voltage regulator (AVR) of synchronous generators using a self-tuning controller (STC) which has the adaptation for the dynamics of plants. For the closed loop system and the controller are stabilized by setting the weighting factor in the evaluating function, the control results are affected significantly by the value of the weighting factor. Then, the auto-tuning of the control parameters is desired. It is necessary to establish the stable control strategy by tuning the weighting factor in real time under the guarantee of the stable performances.
In this paper, the auto-tuning AVR of synchronous generators based on the generalized predictive control is proposed. The number of predictive steps are decided for the stable system by arranging the poles of the closed loop system. Then, the decision method of the weighting factor is introduced by assigning the pole of the controller in the stable region. The sequential least-square identification using the supremum trace gain method is adopted for the identification method. The no-load step response and the response for the closing and breaking of load are discussed by the simulation and the experiment using the microprocessor-based control system. Thus, the validity of the proposed AVR is confirmed.

抄録全体を表示

It is necessary to analyze an operator's thinking process and a operator team's intension forming process for preventing human errors in a highly advanced huge system like a nuclear power plant. Central Research Institute of Electric Power Industry is promoting a research project to establish human error prevention countermeasures by modeling the thinking and intension forming process.
The important is the future prediction and the cause identification when abnormal situations occur in a nuclear power plant. The concept of Multilevel Flow Modeling (MFM) seems to be effective as an operator's mental model which performs the future prediction and the cause identification. MFM is a concept which qualitatively describes the plant functions by energy and mass flows and also describes the plant status by breaking down the targets in a hierarchical manner which a plant should achieve.
In this paper, an operator's mental model using the concept of MFM was proposed and a nuclear power plant diagnosis support system using MFM was developed. The system evaluation test by personnel who have operational experience in nuclear power plants revealed that MFM was superior in the future prediction and the cause identification to a traditional nuclear power plant status display system which used mimics and trends. MFM proved to be useful as an operator's mental model by the test.

抄録全体を表示

Neural networks have been extensively investigated because of their favorable features, i. e. nonlinearity and learning ability. However there always arises a problem of determining their optimal structures. Even if we restrict ourselves to a three-layer neural network, the problem still remains: we must determine the number of hidden neurons, but its relationship to the performance of the network is not clear. Therefore there is no generally applicable methodology for determining the number of hidden neurons, and thus in practice it is usually determined based on trials and errors.
In this paper a structure design method for multilayer neural networks is proposed. Optimal structure or optimal number of hidden neurons is determined based on a redundancy test imposed on the neurons, where the neurons are tested for the linear dependency among them. The basic idea is that a smaller network, so far as it works as well as any larger one, is preferable in view of avoiding overfitting, making learning time shorter and deriving some insight into the inner structure of the target system. Once the network is trained and an allowable error bound is given, redundant neurons are eliminated, and the optimal network structure can be found automatically. The method basically requires only a single time of learning: repeated learnings for different network structures are not necessary moreover, in the course of the determination of the optimal structure, the weights are updated so as to maintain the network performance, and thus the weights relevant for the determined optimal structure can be obtained through the process without any additional learning effort.
The validity of the method is demonstrated by two simulations: optimal structure design for the neural networks representing static and dynamical input-output mappings.

抄録全体を表示

First, asynchronous and synchronous types of interconnected neural networks with discrete state transition are proposed in order to solve continuous optimization problems with convex and quadratic functions.
Next, a computing method is proposed in order to apply these neural networks to ordinary nonlinear functions. The neural computing method adopts an idea of the quasi-Newton method. In this method, Hessian matrix of the minimizing function is approximated by a positive symmetric matrix, a subsidiary quadratic programming problem having this matrix as a quadratic coefficient matrix is solved by the proposed neural network. After its stationary solution is obtained, the approximated matrix is updated to a new appropriate matrix by the quasi-Newton method formula. The updating procedure is equivalent to adjustment of the connective coefficients between neurons. In particular, as a revision of the above method, “Neural Pseudo-Quasi-Newton Method” is presented in which the updating of the matrix is performed after the number of transitions equal to the number of neurons.
Lastly, the neural pseudo-quasi-Newton method is applied to solve the learning problem of a three-layered neural network. This is the learning of three-layered neural networks by using inter-connected neural networks, i. e., “the learning of neural networks by neural networks”.

抄録全体を表示

The design of a machine which is composed of homogeneous mechanical units is described. We show the design of both hardware and control software of the unit. Each unit can connect with other units and change the connection by itself. In spite of its simple mechanism, a set of these units realize various mechanical functions. We developed control software of the unit which realizes “self-assembly, ” one of the basic functions of this machine. A set of these units can form a given shape of the whole system by themselves. The units exchange information about local geometric relation and cooperate in forming the whole shape through a diffusion-like process. There is no upper level controller to supervise these units, and the software of each unit is completely the same. Three actual units have been built to test the basic movements, and the function of self-assembly has been verified by computer simulation.

抄録全体を表示

In this paper, sufficient conditions are given for the existence of an unknown input observer for descriptor systems under the assumption of the observability in the sense of Rosenbrock. The realizability conditions are nonrestrictive and are checked practically for the system's parameters.

抄録全体を表示

This paper investigates the observer design problem for a class of nonlinear systems. The square integrable noise signal to the nonlinear systemsis is considered. It is shown that such an observer gain can be obtained by a solution of Riccati inequality that the observer error system is asymptotically stable with noise attenuation. It is also shown that the observer gain matrix obtained in this paper is equivalent to the solution of H_∞ problem with constant scaling matrix.

抄録全体を表示

The step response of a piezoelectric actuator is analysed experimentally in order to predict its creep movement. The programmed voltage calculated by the results of the analysis reduced the creep motion by one order of magnitude.

抄録全体を表示

A discrete-time preview-learning control method that does not require real-time calculations is proposed. This method was applied to a high-precision lathe for contour machining and 1.5μm straightness of a machined work piece was achieved.

抄録全体を表示

Our model of a thyroid gland regulating basal metabolism has been improved by adding three pictures of the endocrine system, the basal metabolism system and a thyroid gland. Moreover, change in the hormones and the biomaterials was showed by that of color. Hashitoxicosis, combination of Hashimoto's disease and Graves' disease could be simulated by using this model.

抄録全体を表示