We operate the output of suction motor of automatic vacuum cleaner with input from outside. However,in convolution learning, while the shape of objects are distinguished,the other hand,the amount of dust are not determined. In order to change the output of the turbo machine,another processing is performed in addition to convolution learning. As a result of a created device using image recognition, it was possible to make a sorting on the condition. the basic knowledge for the experimental result of the image recognition is reported to the automatic vacuum cleaner referring to the above mentioned result.

This paper proposes methods for operating a robot with nonverbal voice using tongue positions as indices. We defined the voice by Mel Frequency Cepstral Coefficients (MFCC). MFCC are acoustic characters and they show the characteristics of the vocal tract such as the mouth and the tongue. Nonverbal voice used for operating a robot is distinguished with tongue positions in our proposed methods. These tongue positions correspond to the coordinate of the robot’s hand and it enables the operators to operate a robot intuitively. We investigated the feasibility of the methods through experiments of operating a kitchen robot. These methods can realize operating a robot intuitively in continuous values with voice and can be utilized for user-friendly system.

We aim to obtain design guidelines that will enhance the acceptability of telepresence robots in office environments. As an initial step, we conduct an interview survey on a case in which a manager has used a telepresence robot to communicate with his team members occasionally for a long period (about 3 years). In this paper, we examine office worker’s impression change on the telepresence robot. The survey suggests that (1) frequency of interaction with the robot and relationship with the manager affect the impression formation of the members, and (2) suspicious feeling on the robot is weaken over time, and the members gradually start feeling as if the robot is like the manager himself.

Human imitation is suggested as a useful method for humanoid robots achieving daily life tasks. Furniture operation has a feature that it can be described as a combination of five kinds of actions of turning, pulling, pushing, grasping and releasing. In this paper, we focused on learning from the human demonstration with interaction and proposed learning furniture manipulation system. We conducted a furniture/home appliance manipulation with the life-sized humanoid robot PR2.

A purpose of this study is to achieve a robot system following a target human with a smartphone. In order for the robot to stably follow a target, the proposed system includes global following based on Wi-Fi intensity and local following based on walking state matching. This paper introduces target human detection for local following based on vertical acceleration data measured by a smartphone and horizontal velocities of legs measured by a LRF installed in the robot. Also, this paper provides results of detection experiments. The results show cross-correlation between acceleration and velocities of legs can detect the target.

In order for a home robot to manipulate household objects such as furniture and home appliances, it is necessary to know the internal state of the target object and the interrelationship with other related objects of its manipulation. In this paper, a data set is proposed to describe information necessary for such household objects manipulation. In addition, a ROS-based motion planning system is constructed to implement from object recognition to motion generation using the data set. A housework simulation is performed for tidying up a chair to a table with a mobile manipulator.

We are considering business for enterprises using robot technology. It is to provide knowledge and skills from remote places , and operate robots in the field. The Remote Robot Service Platform consists of network appliance robots and a cloud service / content store. On this paper we will report on the concept.

This paper reports on performance improvement of Au/Si infrared photodetector using high efficiency light absorbing nanostructure and backside illumination. We developed an infrared photodetector using nano-cuboids formed on an n-typed Si. Using FTIR, we confirmed that the Au nano-cuboids has about 20% reflectance in the near-infrared region. Therefore, this nanostructure performed as a high efficiency infrared light absorber. The absorbed photo energy was transduced to photocurrent by a Schottky barrier formed at the interface between Au/n-typed Si. We confirmed that the high efficiency light absorbing nanostructure has several times more infrared light detection sensitivity than the conventional structure. By backside illumination, we confirmed infrared detection sensitivity more than 10 times than previous research.

We propose gas measurements using the surface plasmon resonance (SPR) photodetector. SPR occurs when an angle of incidence matched the resonant angle determined by the wavelength. Schottky barrier formed between Au and n-type Si converts electrons excited by SPR into photocurrent. A relationship between angle of incidence and the photocurrent provides spectrum of incident light. Since the SPR spectrometer does not need a long optical path to disperse light, it can offer small sized spectoscopy. First, we confirm a spectroscopic performance by comparing with a commercial NIR spectrometer. Then, we used the SPR photodetector to measure NIR absorption of ethanol gas.

This paper describes a method for evaluating the nano-bio reaction based on products according to experimental conditions analysis of electron microscopic images of nano-bio reaction products. This paper focuses on the spatial frequency components of the image that do not depend on the appearance of the image in order to evaluate and classify the images of the reaction products. Experimental results show the feasibility of the method by analyzing the microscope images of actual nano-bio reaction products.

In this paper, we report a film with Au nanospiral structures to behave a circular dichroic filter in visible light region. Rotating Si while cooling to -100 degree Celsius and irradiation of the thermal vapor of Au from an angle of 85 degrees created nano-spiral structures with a diameter of about 100 nm on the Si. The fabricated spiral structure was transferred to PDMS (Poly-dimethylsiloxane) to form a film. The transmitted light was measured and it was found that it has strong circular dichroism in the visible light region. This characteristic can be applied as a circular polarizing filter or an anti reflection filter.

Our study is based on the concept of contact fusion of droplets between vertically opposing substrates. This paper presents control of droplet-height for selective contact-fusion of droplets. Droplet-height is, for the first time, controlled by using electro-wetting-on-dielectric (hereafter EWOD). The shape of droplet such as droplet-height can be changed by electric field. It is possible to select contact/non-contact between facing droplets by controlling droplet-height. We could succeed in generating the difference in droplet-height by electric filed control. Applying 100V to electrode for EWOD, the height-difference of 4μL droplet amounted to 176m, initial height was 1200μm. This paper reports concentration control of substance in a droplet by selective contact fusion of droplets.

In this work, we have attached the developed MEMS tactile sensor to the tongue model of elastic urethane resin for oral care simulation. This sensor can detect both normal and shear loads as resistance change of the strain gauge on the micro-cantilever fabricated on a Si substrate and embedded in the PDMS elastomer. Output of tactile sensor applying normal load to the tongue model was measured and it is confirmed that output depends on hardness of the tongue model and the point where normal load is applied. Therefore, it is considered that contact information to tongue model including force strength and position can be obtained from output of tactile sensor.

Cells are heterogeneous and efficient collection methods are required to reduce the variations. In this research, the positions of cells were recognized by microscope observation and image analysis. We developed a technique to recover single cells from a surrounding after curing of photocurable gel by multi-point irradiation with a blue laser and DMD.

Understanding of cellular functions and phenomena will bring benefits in a medical field. The construction of three-dimensional cellular tissues in vitro contributes to the elucidation of their functions. This study aims to develop a 3D single cell printer based on micro-droplets containing a single cell. We fabricated an encapsulation channel, which is able to capture, encapsulate, and collect a single cell. The encapsulation channel was fabricated with a PDMS thin layer and PDMS channel, which was made by soft lithography. We captured and encapsulated single cells; based on this results, we also designed an advanced model of a microchannel.

Many people lose the ability of upper limb due to accident or disease, which greatly influences their activities of daily living (ADL). In order to improve their quality of life (QoL), we developed a wheelchair-based upper-limb exoskeleton controlled by Kinect-based motion recognition system with predictive algorithm based on Kalman filter to assist upper-limb function in ADL. The Kinect was employed to design a human motion recognition system as the control input of the device. According to the angles data of user’s upper limb detected by Kinect, we established a 3-dimension dynamic model to calculate the necessary torques of each joint, in which we considered about inertia of human. We were able to control the assistance device by a real-time track. In addition, the Kalman filter was considered to overcome the time delay issue in this study. By using the established mathematical model in the prediction model and the updated model, we can predictively supply the assistance for upper-limb based on Kinect sensor.

The gait ability of the elderly decreases and the risk of fall increases. Thus, we developed wearable assist devices which improved dorsal flexion and minimum foot clearance (MFC) of the elderly. This device uses upper body muscle as a power source instead of an actuator. Attached strings transfer muscle power from the upper body to foot for assisting walking. To improve the efficiency of assist pattern, string paths were optimized through joint torque simulation. This simulator made three optimal paths which apply dorsal flexion torque. As a result of walking experiments, one of three optimal paths improved elderly MFC.

In this paper, we devised the algorithm to visualize brushed area. In rehabilitation medicine, doctors and therapists often take self-care tests to check physical abilities to live. Specifically bed brush skill that patients wipe their body is difficult to evaluate. We captured brush motion with IMU-based MoCap and proposed to visualize brushed area to evaluate the skill quantitatively. This paper detailed accuracy of contact body with the brush in visualizing system with DhaibaWorks and obtained optimum parameters to use for the algorithm of contact to body from the brush.

The goal of this study is to develop a robotic walker that can support walking rehabilitation instead of Physical Therapist (PT). Walking training is being conducted on patients with spinal and lower limb diseases with the assistance of PT. However, a number of PT is not enough comparing to the number of patients, and thus, training time becomes short. This cause reduce training effectiveness. In this paper, we propose the walking training device with assistive manipulator for constant rehabilitation. Working assist experiments by a PT was conducted to measure the walking assist force. From the experiment, parameters required to design the assistive manipulator were determined.

Falling while walking is a major problem among the elderly and disabled people. In this paper, we developed a suit that can support walking by assisting dorsiflexion motion of ankle with pneumatic gel muscles (PGMs). Although PGMs have high power and are lightweight, it is difficult to control by air pressure due to their physical characteristics. Therefore we proposed ON-OFF pattern control that use some PGMs and supply a constant pressure. We designed the assist force based on the EMG and the torque of ankle joint and evaluated the EMG and ankle joint angle while walking in the suit.

This paper aims to explore the relationship between walking and emotion. We employed the two-dimension walking condition map and two-dimension emotion map to confirm their tendency and relation. According to the individual differences, we propose the personal walking ideal-line on the walking condition map. Then, we conduct the experiment by shifting the position on the personal walking ideal-line on walking condition map to stimulate the subject’s emotion. We thus can confirm the tendency and relation for connecting the walking and emotion. To valid use this tendency, we proposed to use it in the assistive engineering. As the results, we recommend patient should walk on their own personal walking-ideal-line, they can freely control the emotion by changing walking pattern. In conclusion, this method can be applied for various assistance devices to keep the motivation until they recover.

We propose a joystick system with force feedback function for the rehabilitation of tongue mobility and tactile sensation. This system is operated with the tip of tongue and has adjustable operational stiffness by impedance control. Our system can give each user effective training contents through selecting impedance parameters. Furthermore when impedance characteristics during stick operation are adjusted, it stimulates user's tactile sensation and requires the user to react quickly in the movement or posture. Namely our proposed system has a potential to improve the mobility and tactile sensation. In this paper, we fabricated a prototype system and examined its possibilities of use in tongue training through several experiments.

Over the few decades, rehabilitation methods such as rehabilitation robot, electrical stimulation and non-invasive brain stimulation have shown to be alternative to traditional rehabilitation therapy which rely heavily on therapists. On the other hand, there have been a lack of research on integrating each element to create an integrated system. In this study, we conducted an initial study on the integrated system that could simultaneously utilize muscle activation induced by focused ultrasound (FUS) exposed to brain and direct muscle activation stimulated by functional electrical stimulation (FES). Whether a latency is compensated and integrated system stimulate targeted muscle is tested through experiments. As a result, we validated the effectiveness of the integrated system.

To regain the locomotive ability in daily living, many prosthetic knee joint units have been developed for transfemoral amputees. Until now, several prosthetic knees have been developed for stair ascent as commercial products, such as Genium® (Otto Bock, GmbH). Such microprocessor controlled knees are multifunctional, and they are able to realize many activities of daily living for transfemoral amputees. However, those prosthetic knees are very expensive, so they have not been widely adopted. The purpose of the present study was to develop a control method for transfemoral prosthetic knees that deals with such variation of gait pattern within subjects. To evaluate the newly proposed algorithm with the new threshold values, the database of the level walking was used. As the result of the evaluation, the accuracies of the gate detection for the stance phase and swing phase were both increased. However, the accuracy for the stance phase of stair ascending could not be enough evaluated because of absence of a database for stair ascent.

It is a problem that the elderly to spoil gait function affects the activities of daily living (ADL), and deteriorates the quality of life (QoL) . In order to enhance the effect of rehabilitation in the gait function, it is desirable to increase the amount of training. In this research, we develop a measurement system for a gait training device that enables effective rehabilitation utilizing IoT and measurement technology proactively, and aims to realize ”always rehabilitation” that consider daily lives except training time as rehabilitation. We modularized a device for measuring the upper limb load, acceleration and walking distance, developed a device for presenting to the user via the Web, and implemented it on a commercial gait training device. We confirmed that we can measure normally using each mounted module.

The purpose of this study was to develop a prosthetic knee joint unit suitable for pedaling by modeling pedaling using four-link mechanisms and examining the link length and joint torque that maximizes crank torque. As a result, the crank torque increased as the link length became shorter. The joint torque of the simulation switched the function of flexion and extension at the singular point of the mechanism. This was shown to be reproducible by a active prosthetic knee joint unit. And the joint torque of the simulation and the joint torque of the able-bodied person were different. Also, it was suggested that the relationship between knee joint angle and knee joint torque of able-bodied person can be reproduced using a passive mechanism.

Motivation is an important factor in rehabilitation. In this paper, in order to improve patient motivation in hand rehabilitation, we developed a prototype of game for hand rehabilitation using the LeapMotion sensor. From the experiments using the game, subjects gave good evaluations about the enjoyment of operating the game, whereas further studies are needed to investigate effects of hand rehabilitation by the game.

The engineering design (ED) is one of the important subjects for the quality accreditation of engineering education. In this paper, I propose a model of ED management process based on the study-of-failure framework, in order that ED learners may devise and verify design solutions which can meet clients’ requirements. Moreover, I propose some systems modeling methods which are effective for the interdisciplinary teamwork of mechatronic ED.

We have been developing a programming teaching-aid robot with intuitive motion instruction set using an embedded microcomputer and using it for introduction of programming in universities and national institute of technology. In school classrooms, students not only create programs individually but also make joint movements of robots. In this research, we made it possible to easily combine robots individually programmed by converting robot into IoT. This system seamlessly supports individual learning and cooperative learning.

Deep learning has been widely used and studied on natural language explanation generation, question and answer, image recognition, and the like are currently undertaken. Therefore, we will develop a system using deep learning so that students can easily handle image recognition.

For the snake robot, elements necessary for mechatronics such as Arduino, various sensors, motors, communication methods, etc. are used. Therefore, we will develop devices that students can learn mechatronics using the structure and system of the snake robot.

This paper introduces three educational activities using an underwater robot, conducted by the author. First, the educational underwater robot “Mark3TM” was used for programs for high school students, junior high school students, and technical college students in the underwater robot convention in JAMSTEC '16,'17, and '18. From the questionnaire survey, the conventions were favorable and Mark3TM had learning effects on STEM for the students. Second, elementary school and junior high school students attended an educational program using the robot in "Tokyo Bay Marine Experience Academy". Third, high school students developed the ROV "MOU" to survey the Ryusendo cave. From the above, underwater robots are found to be more effective than expected about STEM education.

In late years underwater investigation and marine resources development accelerate worldwide, and personnel training becomes the problem by the underwater equipment industry. The lack of talented persons may cause delay in international competition. It takes time and expense to train personnel to master the operating of underwater equipment. However, for example, at fishery high schools and university in Japan, there are few institutions conducting professional education on underwater equipment, and few books become the textbook. Therefore, we developed a learning kit which can learn from the basics about the handling of underwater equipment.

This paper shows our effort for robot engineer training workshops in the project for the robot technology development and support at the University of Aizu. In particular, the paper shows our RT middleware workshop, e-learning and hackathon trial to obtain knowledge for controlling the robot by OpenRTM-aist. Additionally, the paper describes about our effort for workshop that robot engineers learn knowledge for special fields such as electronic engineering and information science.

A passive walker can walk down shallow slope in a natural gait and can exhibit a stable cycle only by gravity. In this paper, we develop some prototypes of passive walker with knees for robot education. We demonstrate better size and better material of the passive walker. Moreover, stability of the passive walker is demonstrated by experiments.

TPIP (TelePresence over IP system) board is used for a rescuer robot system at Rescuer Robot Cutest. We developed an educational material to learn control engineering using a mobile robot equipped with TPIP. It is possible to learn speed control and trajectory control of the robot by using an encoder sensor. Control engineering beginner can easily understand how to control the robot because speed control and trajectory control is designed by pole placement method. Self-position estimation of the robot by using the encoder sensor is occurred estimation error. We propped how to correct self-position by using a camera and a simple landmark. It is possible to correct self-position of trajectory control by using simple image recognition.

In this paper, we verified a microphone array sensor using four MEMS microphones. We tried three sound sources such as monotone, double tones and percussion. We constructed the development environment of a navigation system based on ROS for a microphone array and a mobile robot. Finally, our proposed system was demonstrated through some experiments that sound source creation and robot navigation.

For the starting of Programming education in elementary school, it is necessary to keep the people who has skills of technical programming, education and volunteer for children. On the other hand, there are enough senior people who has motivation for skill up or catching novel techniques as Lifelong Learning while already retired. In this project, the team for the volunteer for school support in programming education field has proposed. In this paper, the situations of elementary school education for programming were investigated. Based on the results, one type of educational environment and tools have been proposed. And it is reported that the proposed environment has been also adapted for programming education for senior.

In recent years, the need for piping inspection has been increasing for various deteriorated infrastructures. Various piping inspection robots have been studied and developed by several research and educational institutes and companies so far. However, most of these robots are too big or cannot pass the elbow pipe or the branch pipe. There is much demand for the pipe inspection with 3 to 5 cm in diameter. It is needed to design the piping inspection robot that compatible with the pipe with small diameter. In this report, we describe a downsizing design and evaluation of moving mechanism for piping mobile robot.

To investigate wetlands, our laboratory has proposed a double spiral mobile robot that moves without damaging plants as much as possible. In the previous study, the proposal and simulation of the turning motion on the flat ground were conducted. For the balance control, the normalized energy stability margin was utilized as a stability index. We applied turning motion on the flat ground to the sloping ground. As a conclusion, it was found that the turning motion on the slope is possible when the turning angle is 5 degrees and the inclination angle is 9 degrees, the turning angle is 6 degrees and the inclination angle is 9 degrees, the turning angle is 7 degrees and the inclination angle is 7 degrees, the turning angle is 8 degrees and the inclination angle is 5 degrees. Furthermore, we analyzed the phenomenon of a falling down motion. We found that when the value of NESM decreased, the robot slightly loosed balance, then gripper failed to grasp the spiral and the robot fell down.

Extensive labor is required for environmental survey. In order to support it, we have developed an autonomous vehicle and control method called Path-generating Regulator (PGR). PGR make the heading of the vehicle converge to the predefined orientation angel in the field. We utilized Lyapunov vector field approach to define the predefined orientation. We have derived new control law and applied it to four wheeled autonomous vehicles. We have showed examples of the control simulation using the proposed control theory. In this paper, we perform experiments using real vehicles and show efficiency of the control law.

This study aims to improve the stability of stair descent motion of crawler-type robots. In this paper, the authors verify the differences in performance for robots with and without sub crawler. Theoretical performance is verified by experiments using small prototype robots.

We are developing clustered convex type telescopic manipulators. Compared with the conventional linear actuators, our convex type manipulator can make long reach and are very light. They are also compact when they are reeled up. Even for these advantages, our previous models were limited on the power about 30N and stability: the structure, especially the tape reel direct driving mechanisms were complicated, and it needed high-level maintenance to keep its function. In this report, we will describe the new manipulator type J, which employs simplified roller driving for the tape, to improve the total performance. The power is about 100N and the maintenance is easy.

The purpose of the research is to analyze vibration characteristic of Flexible Mono-tread mobile Track (FMT) traversing on rough terrain．In this paper，we conduct some experiments with two types of rough terrain, i.e., a crossing pitch environment and a random step environment using the FMT and a differential type tracked vehicle to compare the vibration characteristic.

The disaster response robots usually require task-specific complicated mechanism, and thus it takes long development time. In order to solve this problem, we believe that a mechanism that enables to cross over various uneven grounds by attaching to existing mobile robots is effective. Also, step climbing problem has been postulated to 2-dimensional problem so far. However, since it actually has 3-dimensional shape when a mobile robot negotiates a step in the diagonal direction, examination on 2-dimensional alone is insufficient. Therefore, we propose a 2DoF tail mechanism that enables to climb over to a 3D high step by attaching to an existing mobile robot. We develop a prototype with lightweight and high generating force by using wire differential mechanism and wire reduction mechanism. In addition, we demonstrate the effectiveness of the mechanism by the experiment using the prototype.

This paper introduces a simulation environment for Unmanned Aerial Vehicle to facilitate the developing and testing of an autonomous navigation and coverage path panning system for search operations. The system is composed of a simulation using Gazebo and Software In The Loop (SITL) ardupilot to create a virtual UAV controller and 3D environment. MAVLink communication protocol and ROS are used to interact with the virtual environment for telemetry and navigation control. The SITL ardupilot and the real UAV’s ardupilot are the same controller software. This guarantees similar characteristics and configuration in both the real and simulated environments. We successfully tested navigation using a CPP algorithm and visual feedback inside the virtual world. The images where processed by a popular object detection model. Finally, the resulting information was used generate navigation commands for autonomous navigation.

From the viewpoint of disaster prevention, it is expected the unmanned automation of daily inspection/health assessment diagnostic work for facilities/structures in places where it is difficult to let the inspectors going such as offshore plant and dangerous places. By adopting remote controlled/autonomous robot, regarding the occurrence of abnormality due to human factors, breakage of facilities due to aging of facilities, malfunction caused by them, accidents by increasing the frequency of periodic inspection, it becomes possible to prevent it in advance. In addition, introduction of the robot makes it possible to conduct inspection work even during operation, so it is expected that the availability factor of facilities in such environment will be improved. This paper described about the plant disaster prevention challenge of the World Robot Summit 2018 disaster robotics category held in October 2018. The competition concept and the competition rule were introduced. The analysis of the competition results were shown and considered based on competition results. Finally, the lessons learned was mentioned.

The vehicle for operating on disaster areas has many heavy restrictions. This means that the wheels have risks of the puncture at traversing the ground with debris, and are easy to stack by the tight influence like slipping or sinking at traversing loose soil, and so on. The special wheel, therefore, are required to overcome these severe tasks. To develop special wheel, variable rigidity mechanism using slit tread construction is proposed. Using this mechanism, the variable rigidity wheel which can be kept airless condition is developed. We report about variable rigidity mechanism in detail and running experiments using variable rigidity wheel.

Robotic surgical automation has the potential to stably provide good clinical outcomes and reduce fatigues, operation time and medical errors. In order to automate surgical tasks, autonomous environment recognition is required. In this paper, we proposed a segmentation method using deep convolutional neural networks for robotic microvascular anastomosis. In order to train the deep convolutional neural networks, enormous training data is required. However, manual annotation of ground truth mask is a time-consuming task. Therefore, we proposed transfer learning using simulated images. After our network is trained on a large data set of simulated images, our network is trained on a small data set of real images. In the experiment, the network trained with the transfer learning achieved high accuracy for robotic tools and blood vessels despite that we used a small number of real images. However, we need to solve the imbalanced data problem to improve accuracy for a surgical needle.

In our previous work, we have proposed a passive link mechanism with RCM and gravity compensation to provide a surgeon with intuitive operation of articulated forceps. In this paper, we explore the design principles of a hands-on type user interface for operating articulated forceps enabling simultaneous positioning and precise tip movement control. We develop several different prototypes of user interfaces and experimentally evaluate their usability in terms of intuitiveness and precision in tracing and pick-and-place tasks.