A novel vibration-sensing device based on the giant magnetostrictive effect for a self-powered structural health monitoring system was developed. This device can produce a large voltage because the magnetostrictive rod, in principle, can generate a large variation in magnetic flux density, even when it is subjected to small and low-frequency vibrations of the bridge girder. Therefore, the device could function as a power generator to supply power to a storage battery, when it is not being used as a vibration sensor, to drive the electric circuit . A field test was conducted to check the power generation function and sensing function of developed vibration-sensing device. The experimental results confirmed that the amount of energy of 1mW was obtained by passing one car, and the natural frequency of the bridge girder could be detected by frequency analysis of the generate voltage even when subjected to the intermittent load of passing cars.

In this presentation, we present Deep Echo State Network （DeepESN）, which is suitable for time series prediction of sensor data during machining of machine tools. The recursive structure is built into each reservoir layer of DeepESN to improve the prediction accuracy of time series. In the experiment, we used Sunspot and MackeyGrass instead of machine tool sensor data, and compared the prediction accuracy of DeepESN with recursion structure, DeepESN without recursion structure, and LSTM.

Tire pressure monitoring system (TPMS) has been brought into the regulatory frameworks of many countries in the automotive market. However, a life of battery in commercial TPMSs must limited because of some safety risks and environmental problems. It is noted that energy harvesting approaches are promising to realize a batteryless TPMS. In this study, we developed a triboelectric energy harvester for mounting inside a tire to power a TPMS. The developed harvester can generate the output power more than 1 mW under the load resistance of 10 MΩ at the tire rotational speed of 50 km/h and the load force of 2kN. The 4.4 mF capacitors can be saturated at 3.5 V during approximately 70 s. Additionally, a commercial TPMS can be effectively powered by this harvester when it works.

In this paper, we will introduce a full-pixel real-time vibration imaging system for magnified images that can switch the viewpoint of high-speed vision equipped with a zoom lens and interactively change the viewpoint position using mirror-driven ultra-high-speed active vision with mouse pointer on the screen. This system can be remotely operated from a general-purpose PC, and the zooming-up viewpoint in the wide-angle field of view with a wide-angle camera can be moved at high speed by operating a mouse pointer. It can realize remote vibration monitoring using ultra-high-speed pan-tilt camera. The effectiveness of this system was confirmed by using actual vibration cases such as large machines.

Neural networks or deep learning have become synonymous with artificial intelligence (AI), and the idea that human intellectual work will soon be possible with AI has come to be born. But that doesn't mean that deep learning is perfect. For example, it is known that there are many problems such as "inference is a black box", "unexpected answer by overfitting", and "large-scale network and long learning". Bayesian inference can provide learning and inference that is completely different from neural networks. Therefore, it might be possible to overcome the problems of neural networks. In this paper, we discuss the application of Bayesian inference to the estimation of parameters of the probability distribution. It also discusses the effects of prior probabilities and the relationship between Bayesian inference and maximum likelihood.

Vibration data is used to detect anomalies in rotating equipment such as bearings, but the conventional RMS value and crest factor are not always the good features for anomaly detection because the detection threshold must be determined depending on the usage conditions. In this paper, we adopt a method to learn the optimum features and thresholds automatically by using an artificial neural network. Specifically, a VAE (Variational Auto-Encoder) that compresses measurement data into a latent space with a small number of dimensions is adopted, and the degree of deviation from the normal distribution in the latent space determined by VAE is used as the anomaly score. Unlike the ordinal VAE, linear processing is performed, and the number of dimensions is corrected according to the degree of correlation of the latent space axes. As the result, it is shown that anomaly detection can be performed more effectively than existing methods for synthetic data and actual measurement data.

An ultra-low power consumption wireless vibration sensor module system equipped with an abnormality determination function using a variable automatic encoder (VAE) has been developed as a new method of anomaly prediction solution for rotating equipment such as motors, pumps, and fans, which are widely used in factories. This system uses a cloud service to learn the vibration conditions of rotating equipment during normal operation, and after wirelessly transmitting the learning parameters to the sensor module, the sensor module itself calculates the abnormality score. In order to be able to transmit the learning parameters wirelessly via LPWA, a compression method using branch pruning and quantization has been established. In addition, it was confirmed that the monitoring by the anomaly score by VAE can detect the anomaly of the equipment more quickly than the conventional monitoring by the vibration characteristic evaluation index value, and this method was confirmed to be extremely useful.

A method of measuring the position and movement of an object in a non-contact manner using Bayesian inference is considered. The position and movement of an object can be measured by processing the movement of the object captured as a moving image by image recognition by Bayesian inference. Since it can produce features different from those of conventional measurement methods, it can be applied in various ways. This paper describes the development of basic theories and algorithms for that purpose. Bayesian inference can provide learning and inference that is completely different from neural networks. Therefore, it may be possible to overcome the problems of neural networks.

High-sensitive magnetoresistance (MR) sensors have attracted much attention as an Internet of Things sensor. To improve the sensitivity, a new high-sensitive magnetic sensor which is referred to a symmetric response magnetoresistance (SRMR) sensor is proposed. This sensor uses giant magnetoresistance (GMR) elements which have symmetric field response. An ac magnetic field with frequency f is applied to the GMR elements in order to modulate a small detecting magnetic field. Due to the symmetric response of the MR element, a carrier signal of the ac magnetic field is converted to 2 f and the detecting signal is converted to f. By using a full bridge circuit, where two pairs of GMR elements have opposite phase of the applied ac field, the 2 f components can be eliminated at the differential output terminal of the bridge. Since this system converts the frequency range of the detecting magnetic field to higher one, its noise level, which is dominated by the 1/ f noise, can be decreased. The effect of noise reduction and carrier signal suppression were confirmed by a circuit simulation. Moreover, the SRMR sensor with the antiphase ac modulation bridge was fabricated. It was confirmed that the detectivity of the magnetic field was obtained around 75 pT at 60 Hz. Improvement of the noise reduction and the magnetic field concentrator system, a perspective of the sensitivity less than picotesla might be possible.

In this study, we designed, fabricated and evaluated a piezoelectric thin-film loudspeaker. To enhance the acoustic output in the low frequency range, the device design was optimized by finite element analysis. The prototype speaker was fabricated using the piezoelectric PZT thin film deposited on a stainless steel substrate using RF magnetron sputtering. The vibration displacement and output sound pressure of the completed device were measured as a function of frequency. The measured SPL increased in proportion to the frequency below the resonance. Above the resonance frequency, the SPL was constant independent of frequency. It achieved a sound pressure level of 65 dB at the resonance frequency of 726 Hz. SPL was calculated from the displacement and we confirmed that the calculated results of the frequency response of SPL were in good agreement with the measurement results.

In recent years, attention has been paid to the activity of Unmanned Aerial Vehicle such as disaster site surveys and home delivery using drones. In order to improve the performance of UAV, the application of biomimetics is considered. Dragonflies are the highest flight performance in all organisms. It is thought that various microstructures existing in dragonfly wings are important. In this paper is paid attention to a micro spike of about 100 μm moreover analyzed the numerical fluid using ANSYS FLUENT. As a result, it was confirmed that the flow velocity around the micro spike was reduced. From the above, it is considered that the drag coefficient changes by micro spike.

In this study, we fabricated flexible piezoelectric energy harvesters (PEHs) from Pb(ZrTi)O3 (PZT) thin films on stainless steel foils for wearable applications. The PZT thin films were deposited on Pt/Ti-coated stainless steel substrates of 30 μm by RF-magnetron sputtering. We measured output power by large bending deformation applied to PEHs. During periodic bending motion, the maximum output power of 150.4 μW was measured from the effective area of 380mm2 at a load resistance of 100 kΩ. Parallel connection of each PEH enabled the increase of output power sufficient for charging a battery and turning on a LED.

This study designed a walked robot which had a rotating motion of an artificial claw to imitate the beetle walking and confirmed the effect of the claw vertical rotating or not on the walking distance of the robot. The walking distance by the robot when the claw rotates and when the claw fixed was measured from the displacement of the robot's position. The walking distance by the robot with the rotating claw motion was longer than that by the robot with the fixed rotating claw motion. Therefore, we could conclude that the rotating motion of the claw is necessary for the smooth walking of the robot. This result was similar to the result of the claw vertical rotating or not on a beetle, but this cause was considered to be different. It was suggested that the result may be affected by the differences of the claw direction and the leg motion between a beetle and the robot.

High performance environment-friendly organic friction modifier (OFM) additives are crucial for reducing energy loss and greenhouse gases due to friction and wear. In this study, we designed and synthesized a new type of TEMPO-based OFMs which contain a cyclic structure and free radical. In addition to friction and wear measurements, surface adsorption characteristics of the OFMs were measured with quartz crystal microbalance (QCM). We confirmed that Amide-TEMPO exhibits better friction and wear reduction effect than the conventional glyceryl monooleate. The friction coefficient correlated negatively with the adsorption equilibrium constant, but it showed no correlation with the adsorption density. The anti-wear performance showed no correlation with both adsorption equilibrium constant and adsorption density, suggesting that it is difficult to interpret the wear characteristics simply from the static adsorption characteristics measured by QCM.

The lubrication phenomena of lubricants confined in a nanometer-sized gap are different from those in the bulk state. Clarifying the mechanism of them requires the method to measure the lubricant flow in the nanometer-sized gap. Therefore, we proposed the method to measure the lubricant flow using single molecule tracking of quantum dots with the fluorescence microscope. A single quantum dot has the diameter of several tens of nm and can be applied to measure the lubricant flow in nanometer-sized gap. In this study, we tried to measure the squeeze flow of lubricants in micrometer-sized gap to verify the validity of our proposed method. Our obtained velocities of the squeeze flow were good agreement with those of theoretical model, that indicated the validity of our method.

2-methacryloyloxyethyl phosphorylcholine (MPC) polymer is a promising coating material for extending the useful life of artificial joints due to its biocompatibility and lubricity. Especially, the hydrated brush-like MPC polymer film showed a low coefficient of friction. However, the mechanism of the lubricity is not fully understood. Friction coefficients of the MPC polymer films showed dependency on the shear gap, which must be caused by the thickness dependence of the film's mechanical properties (viscoelasticity). However, due to the thin film thickness (around 100 nm), it was difficult to measure the viscoelasticity. Thus, this research aimed to measure the hydrated MPC polymer brush film's viscoelasticity and elucidate its gap dependency. We established a measurement method that realizes simultaneous measurement of shear viscoelasticity and film thickness.

This paper addresses the issue of force feedback in robotics-assisted microsurgeries. Since it is difficult to install tiny force sensors on micro surgical tools, a novel approach for estimating force vector from the deformation of the surgical tool is proposed. In the proposed approach, a surgical instrument that deforms according to the magnitude of the tool-to-tissue force is designed and machine learning is applied to make the nonlinear relationship between the deformation of surgical tool and the applied forces, so that the tool-to-tissue force can be estimated according to the deformation of the surgical instrument in actual use. The experimental results prove that the method can successfully estimate the applied force with high accuracy from the three-dimensional.

Due to the labor shortage, there is a strong need for automation of material handling and picking operations in logistics and manufacturing sites. Developed robot hand can grasps various items placed in various places, including small space. This makes it possible to automate the handling of various items in logistics and manufacturing sites. The robot hand is the size of a human arm and can switch between suction and pinching according to the situation of the item to be grasped. The robot hand has an independent controller and can be expected to be used in various industrial robot arms. In this paper, the advantages of the developed robot hand and the results of its application to a picking robot are presented.

In order to record tactile data sensed by technicians, we are developing a nail three-axis tactile sensor and a three-axis tactile sensor for robotics. The nail three-axis tactile sensor is for the technician’s finger tip to obtain three-axis tactile feeling applied to the fingertip using deep learning, which learns nail color change images. On the other hand, the three-axis tactile sensor for robotics is for a robotic fingertip to obtain three-axis tactile data, which are derived with image data processing for contact image between skin rubber-sheet and an acrylic core. In this presentation, we will introduce R&D of these tactile sensors and explain the strategy of recording and playback of tactile information.

To improve the tactile VR technology, we researched the haptic device technology that can provide the material feeling to the palm of hand. We adopted Velvet Hand Illusion (VHI), which is one of the tactile illusion phenomena, and tried to control this sensation. Since it is necessary to create the model of VHI mechanism, we tried to formulate it by applying law of pragnanz in Gestalt theory, which is famous for visual research, to VHI. In the formulation, we focused on the basic VHI stimulus in which two parallel lines (wires) reciprocated between the palms and formulated the VHI intensity by using the phase difference of two parallel lines movement and the speed difference of the two lines as explanatory variables. We conducted the psychophysical experiment and verified this formulation. It was found that our proposed model can explain the relationship between VHI intensity and the stimulus conditions when there was no velocity difference between the two lines. On the other hand, in the condition that included phase and velocity difference, our model explained the relationship between them in only large velocity difference condition. This result suggests that the phase difference and the velocity difference are interrelated, and we should clarify the relationship between them to improve the prediction accuracy of VHI in the future.

In our research group, we are developing the electric movement support apparatus for children, who have difficulty in movement. The aim is to support the development of the cognitive ability and emotion of children, through their moving experience by themself, from childhood. In this study, first, we propose a multiuse driving unit and, using this unit, we produce some electric movement support apparatus, for example, the electric moving base for infants, the front wheel drive type powered wheelchair, the motorizing unit for the assistance use moving support apparatus, and the electric moving support apparatus to use with a standing posture. Furthermore, we demonstrate a prototype of the motorizing unit for the assistance use moving support apparatus, which is composed of the multiuse driving unit and commercial parts. The proposed approaches are useful for the discount of the production cost of the electric movement support apparatus and for the popularization of the early movement experience using electric powered apparatuses.

Communication information between subsystems is usually not disclosed in commercialized information equipment consisting of multiple subsystems. For this reason, reverse engineering of the communication information between subsystems is necessary to incorporate those subsystems as developed by a company into the information equipment. The use of some communication information has not been clarified. Therefore, to support the analysis of unknown communication information, this study extracted the similarities of unknown communication commands using word embedding. The model was evaluated using known communication information data, and its adaptability to unknown communication information was examined.

In recent years, deep learning has come to be used in various situations at production sites and its recognition and attention continue to rise. This is because the demand for automation and further efficiency has increased due to the background of the times such as labor shortages in that sites, and deep learning has become a powerful method to meet that demand. In this paper, we aim to improve the detection accuracy from the viewpoint of data augmentation in order to apply the existing object detection algorithm YOLOv3 for the industrial application of deep learning.

Most of the autonomous mobile robots that guide people through facilities and routes do so without considering the condition of the follower. This is one of the dangerous behaviors that a follower performs during guidance. In this paper, we focus on looking away behavior, which is one of the dangerous behaviors that a follower performs during guidance, and aim to develop a safe guidance algorithm that considers the looking away behavior of the follower during guidance. By attaching a depth camera to an autonomous mobile guide robot and acquiring the follower's behavior based on the depth information, we can realize an algorithm to alert the follower of the looking away behavior. In this paper, we measured the looking away behavior of the robot within the range where the neck can be rotated and confirmed that it is difficult to judge the looking away behavior from the posture because the body change is small.

FRR (Feed & Reverse Roller) paper separation mechanism has been commonly used in MFP (Multi-Function-product). The FRR mechanism has been designed using a paper feed operation diagram obtained by static force balance. MRR (Motored Reverse Roller) mechanism has been proposed in order to meet the demand for high productivity and low maintenance costs. In the MRR mechanism, the reverse roller is driven by a DC motor with a constant torque. However, The MRR mechanism designed by the conventional method has a problem that the desired separation performance cannot be satisfied in a high velocity and a space-saving layout. Therefore, it is necessary to consider not only static force balance but also dynamic characteristics. In this paper, with the aim of proposing design methods for the FRR and MRR, we propose a new evaluation index derived by a mathematical model that considers dynamic characteristics, and report the results of confirming the effectiveness of the evaluation index by experiments.

This paper describes basic study on sheet attitude during decelerated delivery of the nip transport mechanism. The nip transport mechanism with a pair of rollers is widely used in many devices, such as automated teller machines (ATMs), printing machines, etc. In these machines, a sheet is sometimes nipped by two pairs of rollers rotating at unequal speeds. There are two kinds of the conveyance between rollers with unequal speeds, namely, accelerated delivery and decelerated delivery. The sheet attitude during the accelerated delivery and the decelerated delivery can be changed. In this paper, we analyzed the sheet attitude during the decelerated delivery by the geometric modeling and experimental observation. As a result, we have clarified that there are two modes of the sheet behavior during the decelerated delivery and we have found the parameters that affect the sheet attitude.

Flat-belt is widely used for conveying the sheet-like medium such as banknotes and cards. In flat-belt conveying, as a trouble leading to operating rate reduction and maintenance cost increase of the device, there is a failure due to belt breaking. As a cause of the belt breaking, scratches caused by handling of the belt, abrasion due to dust contamination from the outside, hydrolysis in a high humidity environment, is known. However, there is a case where the flat-belt is broken rarely even under the conditions that carried out the measures to the above. In this report, focusing on the debris adhering to the surface of the conveying pulley, was subjected to a reproduction test of the flat-belt breaking. The flat-belt breaking mechanism obtained from the results will be described. In addition, an example of a countermeasure against the flat-belt breaking is shown.

In a device dealing with flexible media such as paper, it is common to move the sheet in the plane direction by using a nip roll or a belt. In some devices that need to be moved out of the plane, a mechanism to hold the entire surface is used to prevent the seat from falling. The purpose of these mechanisms is to handle the flexible medium within the processing unit, and it is difficult to apply these mechanism to flexible media handling in a living enviroment. In this report, we present the results of the design of handling device that can handle paper documents in a living environment.

To clarify parameters of a crown roller and tapered roller that affect the automatic aligning of an elastic belt, we investigated an analytical method that can quantitatively predict the movement in the width direction. In this analysis method, we assumed that the belt movement in the width direction occurs owing to a velocity distribution in the belt. Then, the validity of the proposed analysis method was examined by comparing the calculation results and experimental results of the belt movement in the width direction. The comparison showed good agreement between the calculation results and experimental results. Furthermore, it was observed that the radius of curvature of the crown roller and taper angle of the tapered roller affect the movement in the width direction, while the velocity has no effect.

Various web products are often rolled into rolls for ease of storage and transportation. However, these wound rolls are based on experience in testing web thickness distribution, tension and nip loads at production sites. Therefore, it is important to predict defects in advance when the web is wound up. In this study, we conducted an internal stress analysis of the winding conditions and optimized the winding conditions in consideration of the thickness distribution and examined the conditions under which the winding roll does not malfunction.

Porous materials typified by mesoporous silica is expected to be applied in various fields such as adsorption, ion exchange, and engineering. In this research, we focused on mesoporous silica thin films and porous carbon thin films. For the mesoporous silica thin film, mass creation method was established the by introducing the roll-to-roll fabrication method. In this method, the pore size could be controlled by changing the molar ratio of the precursor solution, and the thickness of mesoporous silica film is controlled by changing the peripheral speed ratio.

Human sweat contains a variety of substances, some of which are correlated with sweat in the body. Therefore, measuring the concentration of certain substances in sweat can play an important role in health management. However, it has been suggested that pH affects the measurements in sweat, and it would be beneficial to know the pH in order to improve the accuracy of the measurements. In this study, we develop a flexible sensor device that can measure pH in sweat with high accuracy using CNTs (carbon nanotubes) as electrodes. CNTs were used as electrodes and masked with PDMS, a resin. CNTs were used as electrodes and masked with PDMS, a resin, to simplify and reduce the cost of the sensor fabrication by skipping the doping process. In order to further improve the sensor sensitivity, pH measurement was performed by changing the concentration of polyaniline, and 175~200 [mg/mL] was found to be the optimum concentration of polyaniline.

In this study, quartz crystal microbalance (QCM) was introduced into the highly-precise surface force apparatus, to simultaneously measure the surface force and resonant frequency shift during the contact. Two spherical probes of PDMS and BK7 were used as the counter materials. With the PDMS probe, the decrease of resonant frequency was observed in contact with the Au coated crystal. The resonant frequency shift showed a linear relationship with the surface force in the lifting up process, while it showed a linear relationship with the contact radius of JKR theory in the withdrawal process. On the contrary, with the BK7 probe, an increase of resonant frequency was observed. The resonant frequency shift showed a linear relationship with the contact area of Hertzian contact theory in the lifting up process, while it showed a linear relationship with the contact area of DMT theory in the withdrawal process.

Head smear on slider surface by the laser heating is a big issue in the heat-assisted magnetic recording. In our previous study, we investigated the hydrocarbon with high molecular weight such as alkens or alkanes generated from a DLC film heated to 350-400 °C. Our previous results suggested that the hydrocarbons decomposed from DLC film polymerized in the thermal plasma between the head and disk surfaces. In this study, we focused on the dependence of humidity for head smear generation, because the water molecules might affect the polymerization of the hydrocarbon smear. As the result, we confirmed the humidity in environment affected the smear amount of hydrocarbons.

To realize heat-assisted magnetic recording (HAMR), which can largely increase the recording density of hard disk drives, it is crucial to ensure the reliability and durability of the head disk interface (HDI) that involves high temperature and high-speed shear. In this study, we performed reactive molecular dynamics simulations to clarify the effects of water and oxygen molecules on the decomposition of nanometer-thick perfluoropolyether D-4OH films that lubricate the HDI of HAMR. From the simulation results, we found that water molecules accelerate the decomposition of D-4OH and this effect is suppressed by the presence of oxygen molecules. We also found that high normal pressures and concentrations of water and oxygen molecules lead to fast decomposition of D-4OH molecules.

This paper considers the transmissibility reduction of central pattern generator (CPG) based control system for pneumatic anti-vibration apparatuses (AVAs). This control method is an attenuation method for the flow disturbance, which is the variation of compressed air pressure. A CPG-based controller generates control input, whose frequency is synchronized with the frequency of the flow disturbance. In the proposed method, the parameter in the CPG-based control system is tuned so that the air spring stiffness can be vanished. For this reason, the resonance frequency of the transmissibility can be reduced. It leads to the improvement of transmissibility. The effectiveness of the proposed approach is shown by simulation.

The purpose of this study is to scan the surface of a phantom imitating the human body with a probe equipped with a three-axis tactile sensor, and to estimate the inner texture from the acquired data. As an experimental method, a compact three-axis scanning device connected to a tactile probe scans a phantom containing imitated organs in two directions, and measures shear stress and pressure. From the experimental results, it was confirmed that the shear stress in the scanning direction can easily estimate the phantom inner texture. In addition, it was confirmed that the estimation could be made more clear by increasing the load at the tip of the probe.

Automatic puncture method into artificial blood vessel is discussed. An industrial silicon tube is substituted for blood vessel and is covered with artificial skin model which is made of soft and transparent gel. The three-dimensional (3D) location of blood vessel is estimated using two IR cameras applying stereo visualization technique. 3D plane equation which includes 3D location of blood vessel image on the skin captured by one camera and a focus point of the camera is obtained. A cross line of two 3D planes obtained from two cameras provides the 3D position of blood vessel. Because artificial skin model refracts the light from the blood vessel, the depth of the blood vessel from skin surface looks shallower than the actual one. Therefore, to modify the location of blood vessel model, the 3D plane equations are rotated around blood vessel images on the skin surface by considering the angle of refraction. The cross line of two rotated planes provides the actual position of blood vessel. Automatic puncture experiment is conducted using the robot system with a puncture needle and the experimental results show that the robot successfully punctures the blood vessel into the transparent skin model.

The safety and reliability of the vascular access puncture are fundamental for the success of hemodialysis treatment, which promises less burden on the patients. The dialysis needing requires practiced manipulation skills with clinical wisdom that hardly passed down descriptively, especially on their subjective sensory experiences. This research proposes a practical method for measuring needle posture during the dialysis puncture with a miniature inertial measurement unit (IMU) mounted as an extension at the end of the needle hub. As an initial stage of the research, two clinical experts and two novice candidates participated in our study. The method utilized accelerations and geomagnetic data from the IMU, then assess three-dimensional angular displacements during the puncture to the skin vessel model. The difference in the penetration angle and its trends during dialysis needling seemed to represent their amount of technical skillfulness. The proposed method expects to record and quantify puncture skills on-site, applicable to skill enhancement, improving vascular access puncture reliability.

In this paper, an artificial beating blood vessel model was constructed in order to reproduce the cardiovascular system with beating and to be embedded it in a medical phantom. The torque generated by the motor is proportional to the driving current, so the current was controlled to generate artificial beating blood flow. So, a drive waveform was created with a function generator, and the motor was driven via a current drive device to generate blood flow by artificial beating. Previously, blood was simulated with water for the purpose of building the entire system, but this time we will increase the viscosity of the liquid that simulates blood flow. This takes into account the effects of blood on the blood vessels, blood flow, and pulsation to determine the effects on the human body. A liquid that has a viscosity close to that of blood has a slower response and a smaller amount of pressure change than water, so it is necessary to take into account the viscosity of the liquid. In order to construct a cardiovascular model with realistic pulsations, it will be required to generate pulsations with viscosity added in the future.

In a micro world such as under a microscope, errors of deformation of tools impair the operability. Increasing the rigidity of the tools decrease the error, but it causes the tools to become large. Authors propose a compensation method to cancel the errors by reproducing negative elastic properties with a precision stage. . In this paper, we analyzed the deformation of the tool and indicated that the error caused by the deformation can be compensated. We reproduce arbitrary elastic properties, controlling precision stage according to the force information detected by a tactile sensor. As a result, not only positive rigidity but also negative one were reproduced. In addition, using a stepping motor for driving the stage, high torque and high precision controll is possible in low speed range.

This study aims to investigate relationship between texture perception and touching condition such as active and passive touch. Four different sensory evaluation experiments under various tactile conditions were conducted. In the first sensory evaluation experiment, subjects actively touched the surface of six different underwear fabrics (Active touch) and evaluated the texture by 5 different sensory words. At the same time, pressing force and fingertip velocity were measured. In the next experiment, subjects passively touched the same sample that moved under their finger (Passive touch) and evaluated the texture as same as the first experiment. In this experiment, three conditions were set (Passive (i)～(iii)) based on the average pressing force and fingertip velocity of all subjects in the first experiment: Passive (i) with the same force and velocity, Passive (ii) with the same velocity but with low pressing force, and Passive (iii) with the same pressing force but with slower speed. In both experiments, subjects evaluated all samples three times. Concerning touching motion at active touch experiment, pressing force as well as fingertip velocity decreased when subjects touched the samples that marked high score on “zara-zara” feeling. This result indicates the possibility of object feature affecting the finger movement. Comparing the evaluation result of comfort in active and passive touch experiments, there were samples whose evaluation varied greatly between these two.

The purpose of this study is to confirm that the motion discrimination system constructed this time is useful. In this study, the author thought that it is possible to identify the motion from the change in the force applied to the sole of the foot. The author constructed a system that automatically identifies movements by combining data measured with sensor shoes with multiple 3-axis tactile sensors and supervised machine learning. This time, 3-axis tactile sensors is placed on heel, hallux, thenar, and hypothenar in insole of sensor shoes. A machine learning model was constructed using 12 data obtained from 4 locations. In the experiment, daily movements of walking, climbing stairs, descending, climbing slopes, and descending were performed. Using the constructed machine learning, motion discrimination was performed by changing the combination of the two motions, and the accuracy was evaluated. As a result, it was confirmed that the motion discrimination system constructed this time is useful.

Compression garment has been used in sports in order to improve exercise performance, such as reducing fatigue. Various studies have been conducted to elucidate the mechanism improvement of exercise performance by compression garment, and various findings have been reported. However, since the compression pressure and exercise conditions vary from study to study, there is no certain view on the effect of compression garment on exercise performance, and the mechanism has not been clearly elucidated yet. In this study, as a fundamental study to elucidate the mechanism, we focused on the phenomena that compression garment can reduce muscle activity by reducing vibration of soft tissue. Therefore, we measure and analyze changes in the vibration propagation characteristics of the gastrocnemius muscle from the body surface when the plantar flexion force and the size of the compression socks were varied. From the results, it was found that wearing compression socks reduced the vibration propagation rate of the gastrocnemius muscle in the frequency range of 10 to 30 Hz in most subjects. It was suggested that wearing compression socks greatly suppressed the vibration of the lower leg during exercise.

In order to easily measure load on lower limb joints while walking, the inertial force of body part is calculated from acceleration measured by nine wearable inertial sensors attached to each body part without using motion capture and force plate, and the lower limb joint moments during single stance phase are estimated. However, the lower limb joint moments cannot be estimated only by the inertial sensors since the body in double stance phase has a statically indeterminate structure. Therefore, the purpose is to use deep learning to estimate the lower limb joint moments during the double stance phase using only the inertial sensors, but learning is performed with the correct answer value using the motion capture and force plates. This study uses simple RNN, which is a type of deep learning. As a result of changing the learning order for each joint, it was possible to estimate the load on the lower limb joints moments with a certain degree of accuracy.

This report brings a function of sending alert at wandering as an idea. The function uses record of latitude and longitude from GPS terminal with possible person of wandering. To realize the function, daily life movement of non-wandering are learned by machine learning for past location data from GPS terminal. Wandering or non-wandering is judged by GPS terminal real location data and machine learning. For the judgement, the machine learning is used random forest regression to aim to apply theory and develop system simply. The past daily life's movement locations of non-wandering are learned for possible persons of wandering, a movement location of 10 minutes later is predicted by developed program. This study compared a predicted movement location with a real location, when its gap is big, judge as wandering. Explanatory variables in the learning data set were considered by more than one case, for example, suitable number of nearest time points as input. In addition, Correctness of the wanderings detection by this function was considered.