This study proposes a method for generating motion commands by imitation learning and prediction based on a transformer model using a force-based two-channel bilateral control system. This study aims to demonstrate that imitation learning and generation of motion are possible using two-channel bilateral control, which transmits force information only. The proposed method can realize imitation learning and motion generation using only force information without using position (trajectory) information. The experimental results show that the force information predicted by the transformer model can generate motion using the two-channel bilateral control-based control structure.

In recent years, motor technology has been utilized in a wide range of fields, such as robots and automobiles. When a motor is used for electric vehicles, it is required that the motor supplies different levels of torque in different situations. For example, a motor in an electric vehicle is required to generate a higher torque when the electric vehicle suddenly stop. The braking distance can be shortened by improving the maximum torque of the motor, at the time of a sudden stop.Conventional motors have a larger safety margin of heat-resistant to avoid the motor from breakage. Therefore, it is not possible to provide the maximum torque that the motor originally has. The purpose of this research is to increase the maximum torque of the DC motors. A temperature estimation formula was used to set parameters for each current value to be supplied to the motor. We measured the actual temperature of the motor and compared the measured value to simulation value that is calculated with MATLAB/Simulink to reduce the difference in temperature between measured value and simulation value within 5◦C. Furthermore, we built and implemented a current control algorithm that incorporates the temperature estimation for this motor, so that a larger current can be supplied to the motor as compared with conventional motors. As a result, the maximum torque of the motor can be improved by 5.5 times.In addition, because a high-speed response is required to avoid emergency situations in automobiles, optimal feedforward control in real time was performed.

Encoders are the most commonly used position measurement sensors in industrial machines and robots. Compared to potentiometers, encoders have advantages in terms of environmental resistance and cost, and their detection speed and resolution are comparable to those of potentiometers, so the use of encoders in the field of precision positioning continues to expand. Since the accuracy of motor control depends on the accuracy of angle and angular velocity obtained from the encoder, encoders with high reliability and high accuracy are required. The authors have developed a high-resolution and high-accuracy absolute rotary encoder that applies eccentric rotation. However, eccentric rotation causes vibration due to disproportionate force, which may lead to deterioration of accuracy and lifetime.In this paper, we proposed a new angle calculation method that calculates the absolute angle from the signal characterized by the use of magnets with different magnetic flux densities. With the proposed method, the absolute angle is obtained with the same accuracy as the conventional method.

Monitoring vital signs is critical for daily health management and early detection of diseases. In this paper, we introduce a system for measuring the respiration rate and heart rate with high accuracy using channel state information obtained from commodity Wi-Fi. Unlike conventional methods, this system achieves high accuracy by separating biometric information from the signal using variational mode decomposition, which filters adaptively. By testing this system in an indoor environment, we achieved an accuracy of 98.5% for respiration rate and 98.5% for heart rate.

In recent years, robots to replace human labor in industrial fields and extreme environments are required. Especially, teleoperation technologies are expected to play an important role. Therefore, this study proposes bilateral haptic feedback control for different working ranges based on force-information transmission. Conventional bilateral control systems for the different working ranges are based on a dimensional scaling method which utilizes position/velocity-forces hybrid control. In dimensional scaling, the position of the master system tracks the speed of the slave, and the velocity of the slave system tracks the position of the master system. The proposed method can achieve a dimensional scaling control with force feedback by transmitting only force information with the performance equivalent to the conventional method. The proposed method can reduce the amount of haptic data traffic because of the utilization of only one type of information. Simulations evaluate the proposed method with a comparison to the conventional methods.

When performing tasks using deep learning, the quantity and quality of data are important. However, in unsupervised anomaly detection, it is difficult to obtain a large amount of high-quality training data. Hence, data are typically extended via rotation, translation, and cutting. However, the typical data augmentation method may generate data far from the original image. To obtain various high-quality training data, we partially use AugMix, a data augmentation method that mixes the original image with images obtained via image processing, such as by rotating the original image. Using this data augmentation, new data can be generated while retaining the features of the original image as much as possible, and the diversity of training data can be enhanced, compared with performing the same image processing. Consequently, the effectiveness of the proposal method for improving the accuracy of unsupervised anomaly detection is confirmed.

In recently, the number of traffic accidents caused by automobiles is on decreasing trend, but caused by drivers, such as dozing driving, has increased, relative to the number of traffic accidents. Many studies report that drowsiness occurs during driving, and development of a system to sense quickly from the viewpoint of automobile driving support technology is required. In this research, the authors developed a biometric measurement system adopting capacitive coupled electrodes which can be sensed by the driver unconsciously and under unconstrained condition. In this paper, the authors explain the measurement principle using capacitive coupled electrodes, digital filter, RRI detection algorithm, body movement detection and this interpolation algorithm in Real-time. The authors have conducted RRI measurement experiments using the proposed system and showed that it is possible to measure RRI with high accuracy even when compared with highly reliable physiological measurement systems.

This research proposes a force control for a vehicle robot with an inverted two-wheeled/stable traveling mode. The conventional inverted two-wheeled vehicles have advantages such as the capability of pivot turn compared to the vehicle robots with three or four wheels. Additionally, if an arm (manipulator) is mounted on the vehicle, multi-functionalization of the mobile robot is expected. However, the inverted two-wheeled vehicles have disadvantages such as statical instability and deterioration of the main tasks and inverted-pendulum control due to mutual interference because of the presence of the underactuated joint. Therefore, this research proposes a force control as one of the multi-functionalization by a mobile robot with the advantages of both two-wheeled and stable traveling mode. This research introduces a variable compliance control for the pitch angle direction, which enables the flexible and stable transition between a two-wheel mode and stable traveling mode. The proposed method realizes a stable contact motion to the environment with achieving a soft landing and stable inverted traveling movements at the transition from the two-wheel mode to three-wheel mode. The simulations evaluate the proposed method.

Recently, a field of the servomotors is enlarged in robotics, automobile, and medical applications. In addition, requirements of the servomotors have been changed. Among them, demands of motors for downsizing, high accuracy and high functionality are increasing. However, it is difficult to achieve downsizing of the sensors for controlling the motor and the high accuracy and intelligent functions simultaneously. Therefore, the authors propose a magnetic encoder with realizing small size and high precision. This sensor can easily achieve reduction of the size and high resolution. The accuracy of the encoder can be doubled compared with the conventional structure by changing the number of poles of the magnet from the two poles to the four poles. However, it becomes difficult to detect the absolute angle by the conventional method. Therefore, the authors use the eccentric rotation of the sensor magnet, for the novel absolute angle detection methods. This research describes the new method of calculating the absolute angle and the experimental results for the accurate measurements of the magnetic absolute encoder. Accuracy of the developed magnetic encoder reached 14-bit. The proposed structure realizes the absolute angle simultaneous detection.

A method of stable contact motions for force tracking control based on the concept of resonance ratio control is proposed in this research. The similarity in the structure for admittance control for contact motions and two mass resonance systems is derived. From the similarity, it is found that the concept of the resonance ratio control is applicable to force tracking control. Based on the concept of resonance ratio control, an admittance controller for contact motion with the environmental disturbance (EnvD) compensation is proposed. By using the proposed method, vibration suppression in contact motion is realized, even if the environment has no, or insufficient, damping (decay). The response characteristics of contact motion for force tracking control can be determined arbitrarily by the proposed method.

In this research, a novel structure of macro - micro bilateral control for reduction of quantization noise and interference in the modal space is proposed. Macro-micro manipulation is needed for various fields such as minimally invasive surgery and investigations. Scaled bilateral control is one of key technology for micro manipulation. However, performance of scaled macro-micro bilateral control is deteriorated because of enlarged quantization noise and interference between force and position control. In this research, Kalman Filter base State Observer is extended and constructed in the common and differential modal space in order to reduce effects of enhanced quantization noise which is enlarged by scaling factor. In addition, interference and disturbance are eliminated by using a structure based on a disturbance observer. Validity of proposal is confirmed by simulations and an experiment.

As an academic field which deals with transmission and reproduction of human tactile information, there exists real-world haptics. Bilateral control is an important technology of real-world haptics to transmit tactile information between operators and environment. It is desirable for master or slave systems to work autonomously in order to adapt themselves for motions of operator or environment. In addition, varying the number of driven systems contributes energy conservation and wastes reduction according to change of connection between subsystems. In this paper, a bilateral control system is proposed which realize adaptability to environmental variation of motion and energy conservation by varying system configurations. The variance of system configurations is derived from expression of human fingers as a dynamic model of informationally connected system based on ability to oppos the thumb. ability to oppos the thumb is a necessary function of human fingers and an element which makes the connection vary dynamically. The model is applied to three-fingered bilateral control system. By using proposed method, flexibility to variation of environmental movement and higher concentration to the main tasks and energy conservation are obtained for the bilateral teleoperation. Experimental results show validity of the proposed method.