In today's industrial robots, especially in the filed of unmanned carriers, visual information is indispensable. For the purpose of avoiding obstacles, recognition of the direction is more important than recognition of the minute shape of the obstacles.
In this paper we propose an optical 3-D range sensor for a moving robot, an unmanned carrier, etc. This is based on a spatial-filtering used for detecting a focus. In this system many spatial-filters are arranged at equal interval in an image plane so that azimuth's information could be detected as well as the distance. We take the accuracy of approximately ±3cm for the obstacle at distance of 60cm in the experiments.

View full abstract

We had developed a 2-dimensional and a 3-dimensional simulator with Finite Element Method (FEM). We have studied simulation for hyperthermia with those simulators, then some results were reported. But dealing with a 3-dimensional model, that 3-dimensional simulator needs a long time and a plenty of memory to compute electric field and temperature profiles. Therefore, a new 3-dimensional simulator is developed with Finite Difference Method (FDM) for hyperthermia. This simulator can compute rapidly with fewer memories. As to computation, we use ICCG (Incomplete Cholesky Conjugate Gradient) method or SOR (Successive Over-Relaxation) method. We can give priority to time by ICCG method, to memories by SOR method. For the examination of the accuracy of this simulator, the result compared to direct calculation, and the result from 2-dimensional simulation with FEM.

View full abstract

Recognition method for a large number of human faces is proposed; Category classification is executed based on KL coefficients of face images, and then an operator identifies the subject with eye by comparing the input face image with reproduced images of the selected category. We propose a new adaptive learning algorithm for the base vectors of KL expansion to be calculated for a large number of human faces. The vectors are successively adjusted for a new face image. Adaptive Resonance Architecture (ART 2) is used as a category classification. This method is applied to 514 face images obtained from 58 persons, and its availability is indicated.

View full abstract

A sensitive plant, Mimosa pudica, has a small and interesting sensor mechanism in the leaf which might be different from animal's one. To measure the action potentials of the plant. a wind pressure was used as a stimulus, and a stimulator was made that can change its strength continuously. Electrodes were set at pulvinule (P₁), sub-pulvinus (P₂) and petiole (P₃). Then a leaflet was stimulated by a nozzle of the stimulator and the waveforms of the action potentials were observed. It was reported in the previous paper that (1) the waveform of electrode P₁, changed in four steps with the pressure; non-oscillation type, oscillation type, simple pulse type, and three phase pulse type, and that (2) the action potentials at electrodes P₂ and P₃ appeared over a certain threshold level of the pressure. The experiment was performed under the surround illuminance of about 200lx.
The present paper deals with the experiment performed under higher illuminance condition of about 500 and 1, 500lx. The following conclusions are obtained: (1) The change of four steps is confirmed under these condition. (2) In the lower illuminance level, the peak of the action potentials are generally larger and their waveforms become more complex than those in the higher one.

View full abstract

In this paper, an equivalent circuit of a video tape recorder's moving-coil actuator for a dynamic track following system is indicated and its frequency response is investigated.
In order to obtain a fine tracking, it is necessary to suppress an actuator's mechanical resonance. It is known that daming depends on the driver's output impedance, so the actuator's frequency response was analyzed by using an equivalent circuit. As a result, the following relationship was obtained: Q_V=Q_A/P, where Q_V is the Q factor in voltage drive, Q_A the Q factor in current drive, and P the peak-to-minimum ratio of the input impedance.
Experimental results show this relationship and the same frequency response as analyzed one from DC to resonant frequency and above.

View full abstract

The purpose of this paper is to propose an improved version of inversion operation of Polygon Interval Arithmetic (PIA) and demonstrate the usefulness of it to evaluate value sets of control systems containing parameters whose values exist in given intervals but their exact values are not known. By using PIA, a polygon containing a value set as a subset is computed very fast. To get a good evaluation of the value set, it is required to express a given transfer function in a class of special forms. Two examples, which can be modified so that they belong to such a class, are presented. Since these expressions require to execute the inversion operation of PIA many times, to reduce the enlargement of the uncertainty caused by that, it is necessary to improve the definition of the inversion operation of PIA. Moreover, to implement the new inversion operation, an algorithm to decompose a non-convex polygons into convex polygons is proposed.

View full abstract

In the industrial motor drive systems such as industrial plants and industrial robots, a torsional vibration is often generated because of the elastic elements in torque transmission. This vibration makes it difficult to achieve quick responses of speed and may result in damage to the plant. Such system is simply modelled as the 2-mass system. Many researches with respect to the 2-mass system have been reported. The state feedback control is useful to suppress the torsional vibration. In many papers, however, it is not described in detail how to determine the feedback parameters in order to suppress the vibration and improve the dynamic characteristics considering the variations of plant parameters.
In this paper, we discuss the procedure of tuning feedback parameters in the state feedback control system for the 2-mass system. The dynamic characteristics such as not only load speed but also motor speed, motor torque and torsional torque are taken into account for the pole assignment. The effects of variations of plant parameters on the stability and the control performances are examined in detail. These examinations are carried out with respect to three types of 2-mass system.

View full abstract

In this paper we present a practical design method for SISO dead-beat control systems based on two-degree-of-freedom compensation scheme. We aim to design the control system satisfying the following specifications: (1) For reference input, the output is settled in finite steps, then the output error before the settling time and the input are controlled optimally. (2) For disturbance, the output is settled to zero in finite steps, then the output error before the settling time and the input are controlled optimally. (3) The disturbance is assumed to be added to the input as well as to the output. (4) The output is ripple-free after the settling time. The designed system turns out to be a disturbance suppressed ripple-free dead-beat control system with transient responses of input and output considered. The design problem is reduced to the linear optimization problem with respect to the parameters yielded by increasing the settling time and the solution is easily obtained.

View full abstract

Since the Kharitonov's theorem for the robust stability of interval polynomials was appreciated, the robustness of varied characteristics of uncertain systems has been extensively investigated. Though positivity of real polynomials is a fundamental condition in network theory and adaptive control systems, little results are known about its robustness. We study the robust positivity of real polynomials with uncertain coefficient parameters which are expressed by linear combinations of polynomials. “Positivity margins” are obtained by means of the discriminants and the inners of the matrices composed of coefficients of polynomials. Both of them are sufficient for the linear combinations of polynomials to be positive. They can be easily obtained by calculating mainly eigenvalues of matrices.

View full abstract

Typical schemes of adaptive control are model reference adaptive control and adaptive pole placement control, but these are difficult to be applied to multi-input and multi-output (MIMO) systems because too much time is required to compute their complicated algorithms. Simplified adaptive control (SAC) has recently been proposed, whose algorithm is simple and is easily implemented to real systems.
This paper examines the applicability of SAC to a real system, that is, a 3-axis miniature robot whose dynamics varies with its payload and attitude. The robot is first regarded as a MIMO system to which the MIMO algorithm is implemented. The experimental results illustrate satisfactory performance, and verifies the applicability of the MIMO algorithm to a real system.
Since the interaction between three axes of the robot is known to be small, the single-input and single-output (SISO) algorithm of SAC is next implemented to individual axis, and the interaction is regarded to be absorbed to σ-modification term of the algorithm. The experimental results clarify that the SISO algorithm works almost the same as the MIMO one for this controlled system.

View full abstract

The compensational motion by the upper part of the body for the biped walking robot at transition when the walking velocity varies is investigated. The proposed method in this paper calculates the compensational motion of the upper part of the body at the transition from the motions of that at the stationary states before and after the transition. The motions at the stationary states are derived from the calculation using ZMP (Zero Moment Point). If the number of steps at the transition is a few, the proposed method enables the robot to walk stably at the transition even when the walking velocity varies much. In addition the sole's size necessary for the stable walking is derived from the area of ZMP.

View full abstract

In this paper, an autonomous navigation based on a TV vision system on board a vehicle is proposed. Our method is to use fluorescent lamp array on the ceiling as a lighthouse for vehicle motion.
First, the experimental study of a vehicle control based on information from photo-sensors, set up on a TV screen, is worked out using a practical autonomous vehicle. As a result of this experiment, the validity of our method is confirmed. Next, numerical simulations for this control scheme are carred out in detail. Then, the contour map about the initial position on the floor which gives the limit of controllable or uncontrollable region is indicated.

View full abstract

Rotational speed detection method using a photo-electric rotary encoder is widely used in robotic systems. When a small-sized rotary encoder with low number of pulse per revolution is used, error of the speed obtained from counting the encoder pulses is increased at lower speed of revolution. This paper proposes a rotational speed detection method of high resolution, which can be applied to a small-sized rotary encoder of 100 pulse per revolution. This method utilizes, at low speed range, analog signal of the encoder output pulse, the waveform of which is approximately triangle. By examining relation between the angle of rotation and the encoder two-phased output waveform, we propose an algorithm of detecting precisely the rotation angle from the two-phased output signals. The algorithm is implemented using a one-tip microprocessor. Rotational speed detected by the proposed method is compared with the accurate speed obtained from the potentiometer. Error of the proposed method is less than 6.8% at the low speed revolution of 1.8_??_80 rpm, which shows the usefulness of the method.

View full abstract

The driving force of ultrasonic motors (USMs) is the frictional force between the ultrasonic vibrating stator and the rotor. A mechanical resonance of the stator is utilized for expanding the amplitude of stator vibration. The mechanical resonant frequency of USMs is variable according to the temperature rise of USMs and the change of load applied to them. In order to drive and control USMs stably, the driving frequency of a high-frequency power source must track or match the resonant frequency of USMs. Therefore, detection of resonant state on USM is required. This paper describes a resonance detecting method of the USM with a piezoelectric element sensor mounted on the stator. The phase difference between the generated voltage by this sensor and the applied voltage to the USM indicates the resonant state of the USM, which is derived from the fundamental equations of electromechanical conversion theory and piezoelectric effect. This effective resonance detecting method is experimentally demonstrated using a computer-based measuring system.

View full abstract

For the purpose of developing an efficient driving system of electric vehicles (EVs), a testing system using motors has been devised, which enables us to simulate the driving performance of EVs to be simulated. In this system, running resistance of the EV is realized by torque control of the motor operated as a generator. Nowadays, torque control is performed by using conventional PID controllers, and reasonably good results can be obtained. However, when using conventional PID controllers, it generally needs much effort and time to tune the PID gains.
In this paper, in order to avoid the above problems and improve the performance of torque controllers, a method is proposed by using neural networks. At first, the testing system and its conventional PID controller are explained. Then a neuro-controller is described to tune the PID controller gains. Finally, applying the neuro-controller to the real testing system, we examine the effectiveness of the proposed algorithm.

View full abstract

For recent surface mount devices, chip components have been becoming ever smaller and thinner, requiring the utmost care in their handling. The impact force occurring at the time of mounting with present surface mount devices placement systems can damage delicate components, reducing their service life. A contact operation which suppresses this impact force and a pressure operation which mounts components using a force corresponding to the component are necessary.
In this paper, we examined a technique in which the structure of the control system is simple and which applies satisfactory pressure even for high-rigidity subjects. The technique is applied to a surface mount devices placement system. This is accomplished using a speed control mode which adjusts the contact speed to suppress the impact force and a force control mode for pressure with no overshoot. In the force control mode, a satisfactory response for reference was accomplished for even high-rigidity subjects by utilizing a reference filter. In the speed control mode, motion is stopped momentarily immediately after contact to prevent the vibrations which occur due to the condition of switching to the force control mode; it was shown that there is a proportional relationship between the contact speed and the impact force which occurs during the stop operation. Further, contact and pressure operations at the optimum contact speed according to this proportional relationship were verified, showing the validity of this technique.

View full abstract

It has been made clear that a superconducting magnetic energy storage (SMES) is very effective for power system stabilization. The control methods proposed for power system stabilization by SMES are such as the pole assignment, the optimal control and so on, each of which, however, has its drawbacks.
This paper is concerned with the power system stabilization by neural network control of the active power of SMES. First, the optimal stabilizing control of the SMES power for the model power system is calculated for various power system operating conditions and fault conditions. Then, these optimal controls are used as the training data for the neural network. The neural network used is a multi-layer type with a feedback from the output layer to the input layer. The trained neural network is examined by untrained operating conditions and faults.

View full abstract

When we apply a multi-layered neural network based on the back-propagation algorithm to a particular problem, we must determine beforehand the suitable sized network for the problem. But it is a very difficult problem. Too small a network will not learn at all, while too large a network will be inefficient and worsen its generalization ability due to overfitting the training data.
In order to solve this problem, in this paper we propose a structuring method for obtaining the suitable sized network by multipling and/or combining hidden units in the multi-layered neural network. The result is a compact and efficient network that performs better than the original. Also we demonstrate the effectiveness of this method by appling it to two problems, i.e., to identify a logic function, and to divide a plane.

View full abstract

Immune system has powerful abilities such as memory, recognition and learning to respond to invaded antigens, and is expected to be applied to many industrial applications as well as neural networks. This paper proposes an algorithm imitating the immune system that is based on the somatic theory and the network hypothesis in order to make a plan for manufacturing systems. The former theory illustrates that somatic recombination and mutation contributes to increasing of the diversity of antibody. The others illustrates that mutual recognition network among the antibodies contributes to control of proliferation of clones.
This paper shows that the proposed algorithm based on these theories is capable of searching for a global solution not but local solutions through a resource allocation problem.

View full abstract

This paper deals with an optimal production ordering problem for a modified flowshop manufacturing process having several machines and a carrier. The carrier carries a product between the machines and the yards. Through the motion of the carrier, the succeeding part of the process gives an influence on the preceding part. This leads to a complicated constraint of the problem. Further, the objective function is defined for evaluating a constant rate of production as well as the total processing time.
For obtaining a good suboptimal solution, the genetic algorithm (GA) is applied to this problem. GA is a new method to solve combinatorial optimization problems by simulating the process of natural evolution. The individual description and the procedures for carrying out the operations of reproduction, crossover and mutation are shown. Particularly, our attention is paid to the characteristics of individual description and the effects of redundancy and lethal gene in the individual description.

View full abstract

A chip shooter is a machine which mounts chip electronic components on a PCB (printed circuit board). The time of mounting all components on a PCB with a chip shooter, is influenced by in what sequence of mounting components, and in what placement of tape cassettes. Because of huge amount of permutation, it is impossible to solve this problem in brute force method. Therefor we developed an algorithm for optimizing sequence of mounting and place-ment of tape cassettes. We optimize some random data by this algorithm, and reduce the time of mountng by 13.8% at the maximum. This paper describes the algorithm and evaluation of the algorithm.

View full abstract