We have developed an electric power assist suit (PAS) with soft power transmission by spring. This PAS has function of automatically following user movement and detecting assist timing by discriminant function. To improve accuracy of assisting timing detection of power assist suit, we constructed a system of motion discrimination by detecting PAS wearer’s motion. Results of experiments showed less misrecognition of motion and suitable assist for PAS wearer’s movement.

Hydraulic cylinders have been selected as actuator in earlier exoskeletons due to low weight and high force density. In this paper, we propose a knee exoskeleton using Hydraulic Artificial Muscle (HAM) and four-bar knee joint mechanism. The HAM has similar advantages to hydraulic cylinders, but lighter and compliance, making it suitable for wearable robot applications. To allow the device to better comform to human knee kinematics, a four-bar linkage mechanism was utilized. The human knee biomechanics is a combination of rolling and sliding on the condyle, which can be emulated using four-bar linkage. To adapt the HAM to the link mechanism, we proposed and Angled Bar as a modification to the existing mechanism. The Angled Bar adjusts the torque output and motion range of the proposed device.

As a problem of an aging society, there is physical dysfunction due to muscle weakness of the elderly. And the trunk muscle is one of the parts of the body that are very difficult to rehabilitate. This muscle is an important muscle that supports the whole body movement related to the basic movements of human beings and complex movements like sports. Therefore, as a solution to this problem, we propose a new technique for understanding the role of trunk muscles and working on them. In this study, we design a wearable assist device that focuses only on the rotation of the trunk.

In this study, we developed a lumbar assisting device 2 (LAD2) to alleviate the burden of workers who carry heavy loads. One of the current issues of the LAD2 is that it is unable to perform a natural lifting movement due to its simple control method. With regard to the mechanism, LAD2 will separate from the wearer’s back when his/her torso falls forward, and this makes the device’s wearability decrease. Therefore, in this paper, we analyzed the lifting movement by using the motion capture system to improve the device. On the control side, the natural lifting was measured, and we compared the bending angle in power assistance with the angle without power assistance to improve the control method. On the Mechanical side, the vertical and horizontal curved surfaces of the back of the wearar was measured to widen a contact area of LAD2 for the development of the next generation device (LAD3).

In this paper, we analyze the hand motions using myoelectric moments of the forearm. We wear a myoelectric sensor on the forearm to measure the myoelectric signals and then calculate the myoelectric moments from the signals. The myoelectric moment is an attribute by weighting the measured myoelectric data with location information. As a result, it is possible to obtain not only myoelectric size but also the gravity position of myoelectric. Various hand motions can be identified by calculating and classifying myoelectric moment when move the hand. The purpose of this research is how hand shape and motion can be identified by using myoelectric moment.

In laparoscopic surgery, force sensing for surgical devices is highly required. In this paper, we propose a force sensing method for a cable-driven multi-DoF forceps based on string vibration. The sensing method generates vibrations of multiple cables continuously with a rotating pin and measures the multiple cable tensions from the sound signals with a microphone. The method distinguishes the sound signal from which cable generates and identifies its frequency from a single generation of the vibration. Experimental results show the measurement of four cables.

Echography is an imaging diagnostic method that has advantages of low cost, minimally invasive, real time, compactness, and simplicity. It is indispensable equipment in every medical field but examination with echography greatly depends on the expertise and experience of an operator. In this study, we adopt ROS, a software platform for robots, to construct a control system for a new echography support system. ROS can perform system design, simulation, and control, and can easily do everything from robot design to control. In this paper, the optimal motion range of a support robot in ultrasonic inspection was determined by planning the motion trajectory by ROS. Then, we reproduced the actual ultrasonic inspection environment on a simulation and examined the probe scanning trajectory and range of motion without contacting the bed and the patient.

In this study, we investigated whether there is a difference in the absorbance of foot soles between diabetic patients and healthy subjects by irradiating visible and near infrared rays. We measured the absorbance of the big toe, first metatarsal skin and the heel using a probe and confirmed significant differences at 540nm to 580 nm, 1200nm to 1400 nm, and 1600 nm to 1800 nm. 580 nm is a wavelength at which hemoglobin in blood is easily absorbed, and the significant difference seems to be due to the decrease in the number of capillaries in the patient. The significant difference at 1400 nm may be due to dry feet of the patient. The significant difference between 1600nm and 1800nm may be due to blood glucose.

In this study, we propose a quantitative diabetic neuropathy diagnostic method which uses the foot plantar images on different hardness floor based on skin hardness of patients. The purpose of this study is to estimate the severity of the diabetic neuropathy based on the foot plantar image. We compared the foot plantar hardness measured by durometer with the area change rate for evaluating the hardness based on foot plantar image. A significant positive correlation was found between the area change rate and the hardness of big toe (0.347), first metatarsal head (0.546), and heel (0.512). Also, we compared the evaluation between the method of the change of the foot plantar image and classical, neurological, and hemodynamic assessments. A positive correlation (0.545) was presented between the proposed evaluation and the classical assessments.

Echography has drawback that it requires proficiency in probe scanning to obtain a desired echogram, and that the ultrasound probe is kept pressed down on the body surface, so physical burden on the inspector is large. We have developed a robot system that supports ultrasound examinations to reduce the burden on ultrasound examiners. One of the developed One of the developed systems is a compensation system for organ movement caused by body movement. However, this system can only adapt to two-dimensional movement of organs, making it difficult to follow three-dimensional movement. In this echogram processing system, a problem remained in the algorithm for extracting the contour of the organ. So, in this paper, we built an image processing system so that contours could be extracted from missing contours. Addition, we reconstructed the shape of the PET bottle by combining the transformation matrix and new image processing system.

Medical problems in Japan include the increase in the number of patients due to the super-aging society, the localization of doctors, and the concentration of patients in large hospitals. The relative decrease in the number of doctors has increased the burden on doctors. Echography has attracted attention because of its safety and real-time properties. However, during the diagnosis, a large number of patients are diagnosed a day, and the burden on doctors is large, and the majority of people suffer from musculoskeletal disorders. In this paper, we attempt to elucidate the causes of musculoskeletal disorders and the Cinderella hypothesis using a motion capture system and wireless myoelectric sensors. First, we performed frequency analysis focusing on low-frequency intensity and considered the results.

In recent years, the consultation rate of colonoscopy has been increasing. On the other hand, due to the uneven distribution of doctors in urban areas, measures to reduce the burden on doctors are required in depopulated areas. One of the countermeasures is to use a telemedicine system, but the current telemedicine system has a problem that it has a different operational feeling from existing tests, such as no haptic presentation. The purpose of this study is to develop a remote endoscopy system with a haptic presentation. In this report, basic study of haptic presentation especially on the physician's site has been discussed.

In recent years, the development of medical image diagnostic devices has made it possible to perform diagnosis using a large amount of medical data, but the increased burden on physicians has become a problem. Therefore, in this study, we applied deep learning to the ultrasound B-mode images of the liver to diagnose the presence or absence of diseases using a computer. The estimation of the liver parenchymal region in the ultrasonic image realized the Pixel accuracy of 0.88 and the conformance rate of 0.84. The classification of liver cirrhosis using ROI resulted in 0.88 classification results. According to the results of the study, liver cirrhosis is classified from a single image, but estimation from moving images is necessary for automation, and it is necessary to improve the selection method of ROI in order to improve accuracy.

In recent years, the demand for orthodontic treatment has been increasing, but the change in the moment due to the movement of the teeth during orthodontic treatment has not been measured and has not been quantitatively evaluated. The purpose of this study is to develop a three-tooth device with a built-in 6-axis force sensor and to evaluate orthodontic treatment quantitatively. We have developed a device that can change the angle of the central tooth using a stepping motor to reproduce the movement of the teeth during orthodontic treatment. We verified whether wires needed to be reattached during orthodontic treatment using this device. As a result, it turned out that a larger moment can be applied to the teeth by reattaching the wire during orthodontic treatment

In this work, we propose a holistic body machine interface solution for a standing mobility vehicle for lower-body impaired people, which integrates an autonomous docking system into an embodied control interface. The proposed solution aims to help the user to travel around hands-freely and transfer back to a rest surface safely and easily. An overall design of the proposed embodied control interface, autonomous docking system and the sharing control strategy are presented separately, the result of an initial experiment showed the e ectiveness of proposed autonomous docking system.

The purpose of this study is to develop a robot that assist minimally invasive total knee arthroplasty (TKA), regardless of surgeons’ experience. Because TKA requires advanced techniques for bone cutting. In TKA using a semi-automatic robot, surgeons determine cutting conditions based on their own experience. Therefore, in this study, we developed the algorithms that enables estimating in real-time bone cutting temperature affected by surgeons’ experience. The algorithm used a kind of machine learning method, Support vector machine. We also investigated the relationship between surgical conditions and the accuracy of temperature estimation by SVM.

The purpose of this study is to develop the colonoscope which has an advancing mechanism and a bending mechanism at the tip. The advancing mechanism is composed of two balloons and a bellows. The bending mechanism is composed of a silicon tube and four wires, and is installed inside the bellows.

Colonoscopy has high sensitivity, but the colonoscopic insertion is difficult, and the burdens on endoscopists and patients are large. In the developed device, both the advancing mechanism and the bending mechanism can be electrically controlled, and full automation is expected. As a result of inserting this device into the colon model, it succeeded in running straight, curved, and u-shaped sections.

In this study, we have developed training system for endoscopic forceps operation, that uses the training data previously obtained from a surgeon who has been certified by JSES. This training system enables trainees to experience the senses of forceps operation from the perspectives of vision and force. Although the effectiveness of the developed system, especially for the introductory education of forceps operation, has been already confirmed, the assessment method for the experimental result depended on subjective questionnaires with test subjects. Then, this study developed a tool which enables trainees to review the training and to find the reflection point of own forceps operation in the training. This tool visualizes the difference of trajectories of endoscopic forceps in three-axis directions and the spindle rotation direction among a trainee and the certified doctor. Our experimental results showed that subjects unfamiliar with forceps operation tended to shift the forceps in the depth direction.

To realize automated home rehabilitation for hemiplegic paralysis patients, a robotic device based on finger lifting exercise has been investigated. The device demands a patient to perform voluntary finger lift up exercise while detecting the degree of unwanted movements of the other fingers. The device analyze the movements and calculates the degree of recovery, which is provided to patients. In previous studies, only pushing forces of fingers are measured. To analyze the entire finger motor function, it is necessary to measure both flexor and extensor at the same time. In this report, we show an easy-to-use optical measurement device that monitors these muscular activities at the same time. The result of data analysis by the experiments with healthy and hemiplegic patients, it was shown that the proposed device and measurement method have capability of roughly measuring the status of paralysis.

This paper clarifies the characteristics of displacement mechanomyogram (dMMG) in human walking and verifies the performance of a multipoint dMMG interface, which we have been developed, measures the mechanical muscle activity in multiple muscular groups. A preliminary experiment with 10-m walk test measured the dMMG and integrated EMG (iEMG) from rectus femoris muscle in young and elderly subjects. In this experiment, we found that the dMMG of an elderly subject significantly increased compared with the results of a young subject, as well as the result of iEMG. From the result, we developed the prototype multipoint dMMG measurement interface and conducted an operation verification test of the developed interface. As a result, the system simultaneously detected the dMMGs in rectus femoris, vastus lateralis, and biceps femoris during knee extension and flexion in sitting position.

We have been developing a fingertip rehabilitation support device for hemiplegic patients. The device developed to date has not yet considered the hand posture. Depending on the usage of this device, there is a risk that paralysis may occur more sharply than expected if undesirable posture is taken. In this research, we focus on muscle tension and show the design that takes into account the hand posture. We also show the mechanism to lift a finger during rehabilitation which improves the usability of this device. The results of the experiments showed that the proposed design was appropriate.

For medical care and rehabilitation of hand and finger injury we developed a 3-D measurement and visualization framework and prototype system, which employs RGBD sensor and a deep neural network based visual detection and tracking method. By using this system for the actual medical treatment, doctors and medical staffs can record patient’s finger ability in quantitatively and perceive the effectiveness of the treatment intuitively. The experimental results for virtual testers are shown.

We propose a new rehabilitation assistance by light touch effect using UAV. The purpose of this research is to stabilize walking by providing a “light touch effect” with a small UAV. To confirm the light touch effect by floating object, we perform experiments to measure the sway of the center of gravity with/without flotages. During the experiment, balloons and a small drone were used as flotages. The swaying of the trunk of walking on flat and one leg standing were measured with balloons and a drone, respectively. As a result, it was confirmed that stability of trunk sway was improved by using balloons and UAV.

This study aimed to keep the motivation of users and achieve better effectiveness of assistance for rehabilitation and exercise. When these assistive devices are used to assist elderly people, they are prone to negative emotion and muscle fatigue. Thus, we proposed a walking assistive device that considered emotion and body condition. In our research, the deep neural network was used to develop two-dimensional emotion recognition (pleasure and arousal) by analyzing the users’ brain waves and heartbeats. Near-infrared spectroscopy (NIRS) was used to detect the users’ muscle fatigue by calculating the hemoglobin concentrations. As a result, the walking condition can be adjusted by the walking assistive device to maintain users’ positive emotion and nice body condition. On the contrary, the users’ current emotion and body condition can be promoted when the parameters of the assistant device are adjusted.

To enable the user move freely and stably without operating any control interfaces, we developed a motion prediction system based on logistic regression method for Kinect-based upper-limb assistive device. A probabilistic output is provided by classifying spatial vectors and performing the pre-trajectory with the highest probability, 85% of classification accuracy can be achieved.

Functional electrical stimulation (FES) is a type of neuromuscular electrical stimulation in which peripheral nerves are electrically stimulated to cause muscle contraction. Precise joint angle feedback control is necessary to develop FES rehabilitation systems that can help recover the motor function of patients with paralyzed limbs. Most research focus on the controller design for joint angle control. In this research, the role joint stiffness has in joint angle control is explored. First, a coactivation method based on values that correlate with the joint angle and joint stiffness is introduced. Implementing this coactivation method to a joint angle feedback control system enables the simultaneous control of joint angle and joint stiffness. To be specific, feedback control of joint angle and feedforward control of joint stiffness are realized. Experimental results show that joint stiffness is effective to supress disturbance, and also increases the tracking performance of the joint angle control.

Recently, robot assisted rehabilitation has drawn attention over the traditional rehabilitation techniques, owing to its intuitiveness, customizability and effectiveness. However, identification of the correct motion intention of the human is one of the key elements in robot assisted rehabilitation. Over the recent years, electromyography signals are used to identify human motion intention. However, during hand open/close motions both extensor and flexor muscle groups of the forearm show activation and it is necessary to separate the activation to identify the motion intention. In this study the effectiveness of different classification techniques in separating the muscle activation with different muscles configurations is evaluated. Results show with only two muscle sites, motion intention can be identified with a higher accuracy.

We proposed a functional electrical stimulation (FES) system using a clenching motion as trigger based on a relationship between clenching and grasping motion. In the proposed system, wearable device "earable™" was used to measure the clenching motion. This device is suitable for daily use because it does not require a sensor insertion into subject’s mouth and external holding. We indicated the proposed system could control a joint stiffness and a joint angle of a MP joint which are components of grasping motion.

Massage techniques in nursing have not only physical effects such as promoting blood flow to the patient and relieving pain, but also mental relaxation effects, and are often used as nonverbal communication means. Among them, hand massage is relatively easy to prepare and the effect of promoting blood circulation especially in the upper limbs is easily obtained. On the other hand, in order to acquire the skill of hand massage accurately, the skill of the target arm is various and the skill is difficult to learn visually. In the field of nursing education, special lecturers are required, and there are cases where the lecture time on massage is small. In order to obtain such difficult skills within a limited time for nursing education, it is necessary for the nursing students to have a system configuration that supports not only lectures but also self-learning. In this research, the massage learning support system is proposed to solve these problem by using a pneumatic flexible sensor device.

The therapeutic seal robot, named “ PARO, ” has been heralded as one of the most effective instruments for dementia care. Despite high expectations for PARO, many caregivers have struggled to achieve the expected results, since the usage of PARO largely depend on their knowledge and skills. Therefore, this research aims to clarify effective behaviors of caregivers during the usage of PARO. For this purpose, a preliminary experiment was conducted with two subject groups. One of the groups correspond to university students from faculty of health science who have enough knowledge and skills on dementia care. The other correspond to students from other faculties who have few knowledge and skills on them. For analyzing differences between the groups, their behaviors were visualized numerically using a RGBD sensor.

To support swallowing disorders is a serious problems for elderly persons because they are caused by various factors with advancing age. Mealtime assistant can give them suitable eating environment, students in nursing education must learn how to assist mealtime. Education environment regarding mealtime assistant, however, has various problems because of few instructors comparing with number of students, and few opportunities in clinical situation. We attempt to develop a mealtime assistant simulator for fundamental nurising skill education based on simulation education which is received a lot of attention as an effective educational process for medical and nursing students. In the previous work, we developed a tongue model composed of a tongue part with molded uretane gel, and a electronic circuit board with sensors to detect spoon motion. In the present work, we developed a lip model to detect stimulation by widthdrawed spoon.

Fall is a typical accident that occurs close by and is one of the major causes for the need for elderly care.

It is known that the sole and especially the toes play an important role in balancing and are deeply involved in walking and posture control. In addition, there have been many reports of reduced walking speed and walking stability and increased postural sway under dual task conditions.

Precociously, we developed trip risk monitoring system that can observe visual planar aspect while walking and MTC (Minimum Toe Clearance). This study evaluates trip risk under dual task conditions, utilizing the developed sensor systems

The aim of this study was to investigate body sway in 3 axial directions while walking under multitask conditions to assess the effects of task type and aging associated with the risk of falling. Body sway was measured with a single 3-axis acceleration sensor attached to the hip. 10 healthy young and 11 elderly individuals participated the experiment. The tasks were walking, visual, auditory and calculation, we combined these tasks and experimented. Significant differences in lateral body sway in condition 5 (visual, calculation and walking tasks) and sagittal body sway in condition 3 (auditory and walking tasks) were observed between elderly and young people. These results imply for elderly individuals that postural control is visually dependent and route changes based on sound prompts lead to excessive attentional load.

The aim of this study was to investigate body sway in 3 axial directions while walking under multitask conditions to assess the effects of task type and aging associated with the risk of falling. Body sway was measured with two IMU sensors attached to the hip and the shoulder. 10 healthy young and 11 elderly individuals participated the experiment. The tasks were walking, visual, auditory and calculation, we combined these tasks and experimented. Significant differences in except sagittal body sway in condition 2 (visual task) and all axis in condition 4 (calculation task) were observed between elderly and young people. These results imply that, compared to young people, additional tasks lead to excessive attentional load for elderly individuals.

We presented a biosignals detection method using neonate’s load information. Non-invasive monitoring is necessary for harmless and burdenless medical care of neonates. We have developed multi-biosignals monitoring device using quartz crystal resonator (QCR) load sensor, which has wide-measurement range. The detection of neonate’s biosignals from neonate’s load information is required to realize non-invasive monitoring system. However, the measurement of neonate’s heart rate is difficult since neonate’s load information is low signal-noise ratio due to the low amplitude of neonate’s biosignals. We proposed heart rate detection algorithm using Fast Fourier Transform (FFT) analysis. As a result, we succeeded in measuring neonate’s heart rate, the average and the standard deviation of the difference between the reference was 0.7 bpm and 3.5 bpm. Consequently, we confirmed the possibility of neonate’s heartbeats detection from neonate’s load information.

Pain sensation is an unpleasant sensory experience and an emotional experience, and it is difficult to quantify concretely because there are individual differences due to changes in age, gender, personality, cultural background, physical and mental situation, etc. In this study, the sensation of pain was reproduced with a qualitative expression. Eleven subjects were observed the reconstructed facial expressions, and qualitative and quantitative evaluations were conducted by questionnaire.

In order to respond to the growing needs of nursing, it is necessary to correct the gap between practice and education. Endotracheal suctioning has been performed not only at hospital but also at home frequently, there is a need to develop a suction simulator that can quantitatively assess the techniques of suction catheter manipulation. In this paper, we described a tracheal model with measurement functions: the measurement of the pushing force of the catheter on the airway wall by the force sensor and the detection of the catheter tip trajectory history by the camera were implemented. These data from our tracheal model might be useful not only assessing the proficiency of catheter manipulation quantitatively, but also assessing the risk of tracheal mucosal damage and pain caused by inappropriate manipulation or excessive pushing force.

Transportation using wheelchair is one of the most frequent transportation methods at home, medical and nursing care facilities. Wheelchairs are designed to be operable by passengers but often require assistance, and wheelchair assistance is a heavy burden. In order to quantitatively evaluate the physical burden of the caregiver when operating the wheelchair, it is necessary to analyze the characteristics of the force applied to the wheelchair when the caregiver operates the wheelchair. On the other hand, the optimal posture index of the caregiver when operating a wheelchair is not known, and the operation posture differs depending on the caregiver. Therefore, in this study, we compare the posture and the mechanical characteristics of multiple caregivers to operate a wheelchair, and consider the posture of the caregiver.

Quantitative wound data is essential to support objective wound assessment, and it can contribute to the improvement of wound management. This study developed a system that estimates the depth and classifies the different tissue types simultaneously and continuously using impedance data from the skin. The system uses 50 frequency spectroscopy and 8 direction tomography to obtain two-dimensional area data. To confirm the validity of the estimation and classification we used meat phantoms with a hole or different tissues or both, and the accuracy was calculated for small 200 regions in the skin area surrounded by the sensor system. The system showed 76.0% accuracy for the depth estimation and 87.0% accuracy for the tissue classification. The study confirmed the possibility of simultaneous depth estimation and tissue classification, and further study is required to validate the wound assessment.

This study shows the possibility of plaque detection by near-infrared hyper spectrum imaging (NIR-HSI) and machine learning without plaque stain. While the plaque stain shows where the plaque is, the plaque stain leaves color and may cause anaphylaxis. Near-infrared images are obtained from the plaque cultured teeth by NIR-HSI. RGB images of stained teeth are obtained and superimposed on the near-infrared image. The G channel of the RGB image is corrected to label the spectral data plaque or no plaque. Learning and prediction were performed by Artificial Neural Network. The proposed method showed that plaque could be detected without staining about 70% accurate.

Currently, Japan is rapidly becoming aging society, and 28 % to total population are 65 years old and over. With increasing elderly population, the increase of nuclear family causes the increase of elderly households and the elderly living alone. A social demand to improve the substantial home care of the elderly is expanding. Numerous information technology (IT) base applications in home care have been proposed. Most of those applications are aimed at an intensive management in care. Contrary to this tendency, we believe that while high technology may be necessary for medical practice, mental support is more important in medical care. In this paper, we developed a patient monitoring bed as a system to monitor the patient's sleeping condition without invading their privacy. The system can grasp a patient body temperature as a physiological data and can display the posture of the patient using the temperature distribution data.

Analysis of nurse call is important to evaluate nursing management. However, the nurse call data include wide variabilities of individual patients, which sometime blind true differences between two groups, and massive zero calls, which distort true distribution of nurse calls. These problems may decrease analysis performance. Moreover, in medical statistics, to detect true difference related to interventions, the outcome values are often adjusted by fundamental parameters of the body characteristics such as age, sex and body weight. Thus, the adjustment mechanism is also important for nurse call analysis. We proposed a new statistic model to analyze nurse call data based on Bayesian statistics. The model includes hierarchical Bayes model, zero inflated Poisson (ZIP) distribution and linear combination of adjustment parameters. The nurse call dataset in an orthopedics ward was prepared, and the nurse calls of the patient who had undergone orthopedics surgery and the other surgery were compared. Introduction of hierarchical Bayes and ZIP improved fitness of the model to the dataset. However, the adjustment with age and sex was not so effective for model improvement. Therefore, our model including hierarchical Bayes and ZIP may be effective for nurse call analysis.

Recently, dementia patients have been increasing due to the population aging society. Accidents caused by dementia have become social problems. In the hospital, to prevent these accidents, the nurse’s round of beds is executed and bed side sensors are used. Although these ways have many problems that are burdens on nurses, malfunction of sensors, detection after the accident and so on. Therefore, in this research, we are developing a system to continuously monitor the movement of a patient. In this report, we developed a function that can detect invasion by outsiders such as nurses and visitors. As a result, we confirmed that outside invasion is able to detect. Moreover, patient recognition can continue to detect after outsider invasion.

This paper proposed a case study of robotization using OpenRTM-AIST for the cloud observation system as Internet of things. We target to construct a remote management system for agricultural settlement sewage treatment facility in this study. We have already constructed cloud based remote observation system for them, and we try to add some remote switch robot on power control panel and several analogue meter observation component. For the robotization, we design a set of robot systems using SysML, especially requirement diagram and component diagram, we also design an actual switch robot using 3D CAD, and we composed some control components on OpenRTM-AIST middleware. From the experimental result, we can provide robotized IoT system and it has been confirmed that multiple locations can be controlled simultaneously.

On Internet of Robotic Things (IoRT), various robot systems are connected through the Internet. We need a transforming gateway to exchange data between heterogeneous robot systems to build IoRT systems. Generally, the gateway functions are integrated into robot systems or IoRT platforms. However, the incorporation of gateway functions requires a lot of work at the time of implementation. On the other hand, RT middleware offers the gateway functions using RT components which include multiple types of communication interfaces. In this study, we prepared gateway services using the RT components. The services are offered by a platform integrating heterogeneous systems based on Docker containers and Kubernetes. If you use this platform, you can deploy the gateway functions to server machines or devices like robots more easily and quickly.

With the development of cloud computing technology and machine learning technology in recent years, attempts have been made to acquire the behavior of a mobile robot using deep reinforcement learning on the cloud. This paper describes a system for acquiring the behavior of a wheeled mobile robot using deep reinforcement learning on a cloud. With the proposed system, learning can be performed if there is an environment connected to a Web browser and the Internet without building a special environment when acquiring the behavior of a mobile robot. In addition, the result of trial experiments by the subject showed that the proposed system was applicable to the autonomous mobile system in the real environment.

In this system, three robots with different functions are networked, and the robots can cooperate by sharing the information obtained by each robot. This system uses RT middleware and robot service network protocols(RSNP) as common basic technologies. By using these technologies, robots can be easily networked. We have conducted field experiments using three different robots and one scenario for cooperation. The three different robots are two types of communication robots and mobile robot. In this case, the data obtained by each robot is transmitted to the server using RSNP, a system for sharing data with other robots was developed, and was verified the demonstration experiment at international robot exhibition 2019.

Robot development in robotics middleware such as ROS or RT middleware is attracts attention. In order to disseminate robotics middleware, it is necessary not only to develop functions but also to conduct dissemination activities such as workshops. Since 2005, a number of RT middleware tutorial have been conducted by AIST. In this paper, we send out questionnaires at RT middleware tutorials to improve the training course contents and grasp the contents of troubles, and consider the results of the answer of questionnaires.

In recent years, research and development of service robots has become active, so we are researching mobile robots that can be used for services such as security, guidance, and transportation. In this study, a web camera and a laser range sensor were installed on a mobile robot, and the movement of the robot was controlled using AR markers that included the position information. As a movement control system, the robot moves toward the target coordinates entered in advance, finds the target AR marker, and speaks guidance when a specified distance is reached. To facilitate such movement control task, RTM and ROS SLAM were used in combination, so that they can move smoothly to the target coordinates and speak immediately upon arrival. To confirm the effectiveness of the proposed system, an experiment was conducted in which the model room was moved a round.