In the NEDO project "Analysis and Intervention Technology of Functioning of the Elderly by Using Care Robots as Probes", some appropriate support service with robotic care equipments is tried considering and designing, by developing some robotic equipments with IoT system and analyzing some measurement data with AI technology. In this project, firstly, we developed the prototype of "Resyone Plus", one of the robotic equipments, with IoT system. In this paper, we introduce the outline of the prototype system.

Periodontal disease develops around gingiva and periodontal tissues and is morbidity which affects functions of gingiva and periodontal tissues. It is major disease as well as dental caries and considered to be one of the causes of cardio vascular disease such as arteriosclerosis or hypertension. Therefore, prevention of periodontal disease is deeply related to prevention chronic cardiovascular disease. The best way of prevention of periodontal disease is to learn proper toothbrushing to scrape plaque because periodontal disease is caused by bacteria in plaque which accumulates between teeth and gingiva. To learn or assess toothbrushing, some assistive devices combining toothbrush with sensor system were suggested. Most of them are equipped with only motion sensor to analyze the motion of toothbrush. We assumed that contact forces are important not only movements of toothbrush. In this work, we developed a fundamental measuring system to analyze motion and contact force during toothbrushing.

This paper presents a study on problems and risks on the nursing assistant system by autonomous mobile robots using deep learning to perform automatic fall risk assessment for the patients, when applying to the hospital. The system detects patient's fall risks from their surroundings and condition, and alarms to the medical staffs. From the viewpoint of fall prevention, one of clinical safety, there are problems on hygiene issues relating to movement in hospitals, loss and misinformation of risk assessment, impact on staff work by automation and real-time, quality sustainability, and so on. We propose countermeasures, and show quantitative evaluation methods to confirm effectiveness.

“Too much sitting” is a severe problem in the world. It causes health isuues from low back pain to heavy disease. This research aims at developing- a supporting system for health when people are sitting. For this study, the vital data was collected from the wheelchair ridden patient to analyze and detect pain using a prediction model. A two classes database was created: force data with pain or force data without pain. The classification model is built using CNN. The result has an accuracy of 90 %. Our system can highly detect between sitting with pain and no pain. We compare predicted points and true pain points, these points were also close. Pain prediction after noon was more frequent. From these two results, this model could predict accurately pain. This proposed system was extended to an online feedback system to communicate pain information to the patient with muscular dystrophy.

Long-term monitoring is indispensable for early detection, but Holter electrocardiograph currently in use requires a restraint to the person to be measured. Therefore, a person who is not aware of having a disease, it is difficult to measure. Many wearable biometrics measuring instruments have been invented, but binding to the person to be measured is necessary, too. In order to eliminate the sense of measurement for the person to be measured, we considered that long-term measurement becomes possible by conducting electrocardiographic measurement in daily behavior. Therefore, in this study, we have investigated the basic characteristics of electrocardiogram electrode which enables noninvasive electrocardiography measurement for behavior.

Medical equipment used for newborn babies have some problems caused by attaching sensors. In this research, we suggest unconstrained vital sign measurement system with Quartz Crystal Resonator load sensor (QCR sensor) as a solution of them and in order to realize this system, we optimized sensor sensitivity.

We realize unconstrained measurement by load measurement. Then, we used QCR load sensor because wide-range load sensor is necessary to measure both weight and slight vibration caused by beating and respiration. However, current sensor is designed for adults. Sensor improvement is necessary for babies measurement and we tried it by sensor miniaturization. We succeeded sensor improvement and improved sensitivity is 4.8 kHz/N, 12.6 times higher than current one. We will measure babies with them in the future.

In the field of orthopedics, it is important to analyze the kinetics of the shoulder joint. Our research group can analyze kinetic analysis of shoulder joints with high accuracy using the software developed. Kinetic analysis calculates six degrees of freedom of the humerus and scapula. This result is also important, however in the shoulder it is important to analyze the movement of the humeral head relative to the glenoid of scapula. Therefore, based on six degrees of freedom data, the relative relationship of bones on 3D-CAD was reproduced and the relative relationship between the humeral head and the glenoid of scapula was analyzed. We focused on the shoulder kinematics about dynamic scapular plane full abduction. Patients with rotator cuff tears and healthy subjects were seven and ten, respectively. As a result, the difference between the normal shoulder and the rotator cuff tears in scapular plane full abduction was confirmed.

We have successfully formed cell constructs of the cardiac muscle cells without using any scaffolds by taking advantage of the spheroid fusion characteristics. We focused attention that beat of the cell constructs force acting on the dedicated jig. This study aimed to develop a software that can measure the beat power and timing of the cell constructs based on the displacement of the jig. The beating appearance of the cell construct was shooting video from above. We programed it possible to track the displacement of the jig on the computer. The force of the beat of cell constructs was calculated from the physical properties and displacement of the jig. As a result, the software we developed were able to noninvasively measure the period and force of the beat.

In order to master hemostasis and needling technique in hemodialysis treatment, In this wearable simulation armor, a self-sealing blood vessel which can be repeatedly used and an extravascular needling detector are sealed in a polyurethane gel and attached to the arm of a simulated patient. Further, the pressure distribution can be evaluated by a force sensitive resister array arranged on the lower surface of the gel. The usefulness of this system was confirmed by using to hemodialysis needling training.

In order to realize a simulator that efficiently learns the sputum suction technique, quantitative information (suction pressure / catheter motion) on the technique of the phlegm suction in the clinic is continuously collected using the pressure / acceleration sensor attached to the aspirator side · We constructed a system to analyze and examined its usefulness.

Magnetorheological fluids (MRFs) are composite materials made of ferromagnetic particles, medium oils, and several types of additives. Their rheological properties change rapidly, stably, and repeatedly when magnetic fields are applied. In this study, we developed an optimized design method for the MRF device. Three different MRFs are modeled with Bingham fluid model in off-state and on-state. Design procedures were conducted for a prototype with MRF-140CG. Finally, we developed a prototype of the MRF-based torque controllable device to confirm the modeling accuracy.

There are only a few papers written on how to fill MR fluid in MR fluid clutch. We examined the flow of fluid when the number of air holes and the number of inlet / outlet were changed. Improvement of braking torque could be confirmed by filling method and internal structure.

A surgical simulator is considered as one of promising applications in virtual reality and haptic fields. In our previous research, an encountered-type haptic interface using MR fluid had been developed. The system can display arbitrary force by controlling the magnitude of yield stress of the MR fluid and an operator can feel resistance force by cutting the MR fluid with operational instruments directly. For optimal design of this device, it is necessary to predict the resistance force while cutting the MR fluid. Therefore, DEM simulator has been developed for the prediction. In this research, the cutting flow mode of the MR fluid is analyzed by visualization experiment and the validity of the developed DEM simulator is carried out by comparing the visualization experimental result and the numerical simulation result.

For factory automation, robots are required to substitute for workers. For this issue, researchers have paid attention to humanoid and dual-arm robots. In these researches, autonomous generation of motions for target objects is a challenge. In this paper, we focus on the reaching movement. For this challenge, we propose an end-to-end motion planner based on the convolutional neural network, CNN. In contrast to other related motion planners, we use only right and left images captured from the head camera of robot. The network training is executed by learning from demonstration. This allows the robot to map relationship between the images (input) and joint angles (output). Through the experiments, we show that the robot is enabled to generate the reaching movement toward objects located not only at the demonstrated positions, but also at other unknown position.

Groping motion with sensing surroundings is effective for robots to manipulate objects in unknown spaces. To realize this motion, tracing object surface by using sensor feedback is essential. Especially in unknown environments, non-contact sensing such as proximity sensing is suitable for both measuring and avoiding collision with the environment. In addition to the reactive tracing motion, memorizing the sensor information and making use of the memory in the later groping stages are desirable in terms of operating efficiency. In this research, we firstly propose an approach to generating an efficient trajectory of tracing motion based on the memory. We also propose an approach to executing reactive tracing motion by using proximity sensing. Secondly, we propose a method to grasp objects based on the memory and proximity sensor reaction. Lastly, we validate the effectiveness of our proposed methods through a real robot experiment.

A Fishing line artificial muscle actuator is one of polymer actuators which can contract by heating. It is quite lightweight, soft, and cheap and thus it has been expected to use in robotic systems working around human living space. However, there is a room for improvement about the response because its input is heat. In this paper, a natural convective oil cooling system is introduced to improve the response in cooling phase of the actuator. By using a 1-link rotational system, the transient responses of the angle control in three-types of silicone oil are investigated and compared with that in the air to show the effectiveness of the oil cooling system.

This study proposes a novel mechanism to change the stiffness of finger joint with electrostatic force. In daily life, people use a variety of tools and interact with other people and they need stiffness control to manipulate these tools properly. In this paper, a method to change the stiffness is maintaining the deformation of the finger joint using electrostatic adhesion. The electrostatic adhesion use the electrostatic force between electrodes and generate secondary frictional force. We use the frictional force for proposed method. A light-weighted electrode sheets is developed to add the stiffness control in the underactuated mechanism and evaluate the stiffness. In this paper we propose the developed mechanism design with polyimide electrodes.

Able to grasp objects of any shape and size, universal grippers that use jamming phenomenon to change their own stiffness have been developed. The authors have focused on the Hot-Ice phenomenon of supersaturated sodium acetate solution as a means of variable stiffness. However, since the crystallization occurs rapidly and the crystals cannot grow large, the whole solid lacks stiffness to grasp heavy objects successfully. In this research, the authors proposed a new method of enhancing the stiffness of crystalized phase of the solution by adding insoluble fine particles. They conducted experiments to show the effectiveness of filler addition on further stiffening.

We use ultrathin polymer film called nanosheet to develop wearable battery-less devices which don’t need adhesive to be attached to the skin. Considering NFC (Near Field Communication) is not suitable for the skin-adhesive devices because the capacitance between the skin and the antenna coil reduces the voltage, we adopt electric NFC to transmit data and power to communicate through the human body. We measured the relationship between the IC tag yield rate and the three factors: the area of the electrodes, the distance between the two electrodes, and the thickness of the nanosheet. All of the 6 devices worked and the electric NFC devices are 79% thinner than the conventional NFC devices. Thus, electric NFC enables nanosheet devices to work without batteries.

In this paper, we propose a mechanism to deform the flexible gel membrane and explain the prototype. This mechanism was devised to simulate the function of the crystalline lens of the human eyeball and to perform simulation. The crystalline lens is a convex lens-like flexible object, and its thickness can be increased and decreased by extending and contracting in the radial direction, and focus can be adjusted. We aim to pull the lens uniformly in the radial direction by this mechanism. Therefore, we designed a mechanism using a film of flexible shape memory gel and linear motion. By the linear motion mechanism, it is possible to increase or decrease the amount of deformation of the flexible object. Lens model trial with triple size of human eyeball. The result of simple experiment using prototype machine show that soft material can be extended by about 10% in the radial direction.

Since Urushi has high durability, safety to human body, and high electrical insulation, Urushi-Based Electronic circuit (Urushi circuit) using Urushi as circuit board material has been proposed. Multilayer mounting of Urushi circuit can be considered as a method to make the Urushi circuit smaller and more advanced. However, in the current method of making Urushi circuit, unevenness caused by the thickness of the wiring occurs on the circuit mounting surface when laminating Urushi circuit. It is a problem because it causes distortion in the upper layer circuit. Therefore, in this research, we investigated the surface smoothing method of the Urushi circuit in order to multilayer Urushi circuit. this paper proposed the method making groove to embed wiring on Urushi film by silk screen printing.

Generalization to multiple in-hand manipulation tasks are difficult to realize and not well-investigated. This research studied how to generalize with task information. The Allegro Hand which has 3-dimensional distributed tactile sensors, uSkin and 6-axis F/T sensors on each fingertip was used and the abundant tactile information results in 312 (288 from uSkin and 24 from 6-axis F/T sensors) measurement. Furthermore, convolutional neural network (CNN) is implemented to process the abundant tactile information which has touch states between an object and fingertips and long short-term memory (LSTM) is applied as processor of time-series information. Finally, task information is exploited for adapting to multi-in-hand manipulation tasks (pushing and twisting). As networks are combined, CNNLSTM could generate two stable in-hand manipulation with task information.

In many cooking recipes existing on the Web, only general cooking actions are written, however, for a robot to actually cook, we need to instruct the robot in more detail. As a method of automatically planning such detailed actions, this paper proposes a method to estimate physical properties such as shape and size from food images by using CNN and estimate cooking actions based on these properties. For estimating the cooking action, not only the physical properties of the ingredients are considered, but also the cooking action performed before, the age group of the person who will eat the food and the cooking time. Finally, we describe the results of the performance of the shape/size classification model and the cooking action estimation model.

This paper presents a new interpretation technique of the recipe the motion generation for the cooking robot. In order to generate robot operating codes by recipe written in Japanese, it is necessary to perform recipe data conversion, motion codes conversion, and robot motions generation. Firstly, recipe datas are generated by converting from the recipe getting at internet to recipe data type. Secondly, motion codes are generated by sorting recipe data based on database. Finally, robot motions are generated by using Genetic algorithm so as to meet requirement. A series of cooking operations generated by using proposed technique showed the motion expected by human, and it is successfully conducted to generate executable robot motions in real world.

The use of reasonable robot hands could be practical for industries. An Allegro Hand which is inexpensive and easy to maintain but inferior in performance to some highly sophisticated hands (e.g., TWENDY-ONE Hand) was used in our conventional research. To compensate for inferiority performance, we focused on getting abundant tactile information and curved fingertips which is important for in-hand manipulation. Moreover, such tactile information needs to be processed to provide appropriate feedback for the robotic hand control. To achieve this, we proposed using uSkin, 3-axis distributed tactile sensor and using convolutional neural networks (CNNs). We prepared the model that was trained with measurements from 3-axis tactile sensors, 6-axis F/T sensors and joint angles. In the experiment, the model executed in-hand manipulation with the ball and the cylinder 10 times. As a result, the successful manipulation was achieved 8 times out of 10 for the ball and 7 times for the cylinder.

Robots have been introduced to assembly work in industry field. However, since high accuracy in needed for assembly, a great cost is required for system construction. In peg-in-hole task, we propose a method to improve success rate by rotating peg on its axis. In experiments, we clarify the effect of rotation of peg and get peg-in-hole work be robust to the position error.

In recent years the market size of robots for homes has increased. One of the tasks expected to be automated by robots at home is to clean up dishes. In order to introduce a system to an environment that handles various kinds of dish, development of a mechanism capable of grasping various types of dish and an image recognition technology capable of recognizing various types of dishes are required. In this paper, we propose shape estimation and pose estimation of dishes of unknown shape for dish storage robot system. The proposed method expresses the three dimensional shape of dish by mathematical model and expresses the shape of various kinds of dish by changing parameters. Estimation of three dimensional shape and pose of dish is realized by simultaneous estimation of shape. We estimate shape and pose by estimating parameters of mathematical model of dish and pose of dish.

Representation and modeling of flexible objects is relatively difficult because these objects are deformed by external forces. When robots handle these objects, it is crucial to predict the deformation caused by external forces they add during manipulation. Somatosensory information has essential information to predict the deformation, but a method for prediction of deformation by multi-mordal information is not established. In this paper, we conducted experiments to predict cloth deformation images not by directly modeling the objects, but by using a deep neural network model. The model consists of Long-Short-Term-Memory module, which has visual images and somatosensory information as input. We manipulated dual-arm manipulator, handled cloth and collected visual and somatosensory data. Compared to inputting only visual images, the model is able to output vivid and long term prediction by using both visual images and somatosensory data.

This paper presents a constrained motion planning to flip objects grasped with two grasp points of the soft finger type while avoiding rotational slip/drooping. We study the drooping motion in inclined planes under the effect of the object weight and the friction torque from fingertips. The model of the drooping is derived and utilized in the proposed planner to restrict the robot poses that can cause the object in the end effector to slip. The planner is verified through the simulation of flipping a stick, and the generated robot motions enable correct execution of the flipping task without undesirable slip.

It is difficult to align a moving body, such as a remote controlled robot or a construction machinery, with a target object of a remote operation using a first-person perspective only. Therefore, the accuracy and the efficiency of the remote operation are reduced, and it is an important problem of the remote operation. In this study, we propose two methods which use the first-person perspective and do not need environmental cameras. The first one is composed of emphasized optic flows and improves a feedforward operation of its operator. This method was evaluated preliminarily. The second one is composed of a depth index which is easily visible and improves a feedback operation.

Operation training using an actual excavator may be difficult in securing a place and safety. It is important to construct an environment for operation training of excavators. We developed a system to evaluate the skill of operating a hydraulic excavator using a remotely controlled (RC) excavator and augmented reality (AR) technology. We remodeled the RC excavator such that it can be operated in the same manner as an actual excavator and proceeded to measure the excavator’s state. To evaluate the skill of operating this system, we calculated several indices from data recorded during excavation work, and compared the indices obtained for expert and non-expert operators. Furthermore, we calculated the same indices from data recorded during excavation performed by an actual excavator to compare the indices of the RC excavator and those of the actual excavator.

When a natural disaster occurs, investigation and restoration are necessary to prevent damage spread. However, disaster scenarios always have a risk of secondary disasters, making it difficult to send in human for a response. In such situations, it is common to use unmanned construction machines. Moreover, recently, intelligent construction machines are required for efficient unmanned construction. Therefore, in the authors of this paper have tried to realize an intelligent construction system. This paper introduces the research and development conducted with regards to this intelligent construction system. This includes map generation by a drone, visualization for remote operation, and tumble avoidance.

The authors have developed a peristaltic motion type transport machine simulating the movement of the intestinal tract. As a drilling earth and sand transportation method, we have proposed sediment transport by a peristaltic motion type transport machine. Vertical transport experiments of samples with varying particle size and moisture content so far have been conducted to show that samples of each particle size can be transported. However, we don’t know how the conveying amount varies depending on the liquid limit and water content ratio of each sample. In this report, we discuss the boundary of lifting sample by discussing the transport amount from the ratio of moisture content and liquidity limit.

Recently long reach articulated manipulator is widely required for infrastructure inspection. Long reach manipulator can be realized by using thrusters for weight compensation. In this research, we propose a super long reach manipulator with weight compensation and joint drive by thrusters. Considering the static influence of weight and its longitudinal direction moment of the manipulator, the force required for each thruster was obtained analytically. The result of the analysis showed that the weight compensation force is constant, but the longitudinal direction moment caused by the weight depends on the arm posture. After that, we designed a system which controls joint angle with the simple PID controller and the feedforward term, and confirmed its validity by dynamic simulation of the 2 units model. Moreover, we developed 2 units prototype whose length was 4.3 meters and confirmed its basic motion.

Construction wastes discharged from the dismantling site is sorted for each type by the heavy machine or the hand of a person at the site and transported to the treatment plant. The social needs of the construction wastes automatic sorting system has been increasing due to soaring labor costs and labor shortage.In this paper, we describe the robot vision which recognizes the class of wastes.

Work efficiency of teleoperation of heavy machinery is less than a half of boarding operation because of difficulties in creating mental representation of work sites. Thus, we proposed a pre-offering view system for teleoperators to remember environmental information, which helps operators plan paths and working strategies, but some operators forgot their planning. Problems of previous researches including the navigation field are not considering changes of work states including moving and reaching and cognitive load. Therefore, in this paper, we develop a planning reminder system with low cognitive load which can adopt changes of work states. We display different AR depending on work states. Furthermore, we display AR only for current and next work states while operators are in low cognitive load to avoid increasing cognitive load to watch AR. The experimental results using a scale model indicates that the proposed system can decrease work time and cognitive load.

The objective of this paper is to model and predict the behavior of backhoe when working cooperatively with autonomous dump trucks. Specifically, this paper focuses on predicting when the backhoe operator is ready to load sand or gravel onto the dump truck. We employ Non-Parametric Bayes Hidden Markov Model to model the behavior of a human backhoe operator using data from sensors attached to the backhoe. This paper proposes a method to model the behavior of backhoe operators. Additionally, state transition of the constructed model is studied to discover a key transition that indicates that the backhoe operator is ready for loading.

Restoration work following a sediment disaster or volcano disaster may be executed in extremely dangerous places. Construction in such dangerous places can be executed safely if it can be done using construction equipment that can be operated remotely while observing camera images. Construction using such remote control type construction machinery is called “unmanned construction”, and it has been carried out more than 200 sites in disaster prone Japan. Generally, in the case of remotely operating a construction machine, images of "on-vehicle camera" mounted to the driver's seat of the construction machine and images of "external camera" for grasping the situation around the construction machine are used. However, it is difficult to install an external camera at the construction site which is a dangerous places.

In this study, we conducted a comparative experiment between "Around Monitoring System" which does not require an external camera and a system using a conventional external camera.

When a volcano erupts, it is too dangerous for humans to observe the volcano from close to it. Therefore, a UAV-UGV robotic observation system was developed by the authors’ research group. In this system, UGV (unmanned ground vehicle) is transported to a landing point by UAV (unmanned aerial vehicle), and the UGV moves to observation points by remote operation. The effectiveness of the system was confirmed in actual field environments. However, during such remote operations, the UGV may turn over or get stuck because a suitable landing place and path for the UGV have not planned. To solve the problem, in this research, we propose a landing point planning and path planning for UGV using three-dimensional terrains to prevent UGV’s turning over or getting stuck during its mission. As a result of landing point planning with real environment data, the landing point was flat enough and no vegetation. Also, as the result of path planning, the path was in the travelable area, and the distance to the obstacle was enough for safe traversal. Therefore, by the proposed method, safe landing and moving of UGV is realized without overturning or getting stuck.

This article reports a robustness analysis of damping control systems of a suspended load for truck cranes taking into consideration plant perturbation such a change of the boom length. The controller analyzed is based on a model that consists of the second bending mode round the hook and the input delay of existing valve controllers. In the analysis, the closed loop poles are investigated in five boom configurations, which are near the center and the four corners (edges) of the moving range of a truck crane. The results showed that a controller designed dedicating to the center position cannot stabilized the suspended load sway at the one corner where the boom length is shortest. It has also been confirmed that the suspended load sway is well damped even when the boom configuration is in the corners by using dedicated controllers. Thus, to cover all the moving range, a kind of gain scheduling will be needed, and this is a future work.

The productivity of widely-used tele-operated system is lower than that of operating actual excavator. One of the causes is lack of machine behavior information. Therefore, there is a possibility that work efficiency may be improved by realizing feedback of machine behavior information using a motion simulator seat. In this study, the subjective vibration magnitude was calculated by frequency correction, according to human perception characteristics, about the acceleration on the seat surface for the specific behavior of the hydraulic excavator. As a result, for the end-of-stroke impact of the attachment, it was found that the vibration in the lateral direction of the excavator is smaller than in the other direction. This suggests the possibility that the reproduction of the vibration in the lateral direction is not necessary for the machine behavior feedback.

So far, we have proposed a semi-autonomous control method focusing on excavations in teleoperation environments. The main issue of designing semi-autonomous addresses how much and how to introduce human interpose to autonomy. The conventional method has one DOF of the position and orientation coupling of the bucket. However, because it causes stuck motion with a high workload of the soil, an appropriate coupling and number of DOF have to be selected for human interface. In this paper, we extend the conventional method to improve operability. The position/orientation and velocity coupling are introduced and they are embedded into autonomy with an attractor controller. Moreover, the proposed method is evaluated through simulations and experiments using a prototype of the excavation robot.

The construction industry has many fatal accidents. In addition, The number of labors is decreasing due to the declining birthrate and an aging population. Particularly, it is serious problem for local small and medium enterprises (SME). Therefore, our research group has been researching the autonomous system of construction machine. In this research, we aim at researching a system to realize autonomous surface compaction work by vibration roller for SME. To enable autonomous navigation of such a vibration roller, it is important to obtain its position and pose. Our approach is to use external LIDARs located at the edges of the target environment. In this research we implemented the above idea, and conducted experiments using an actual vibration roller to evaluate the proposed method.

In order to maintain the gas tank or rescure the people in disaster site, the wall climbing assist suits have been developed.The principle technique of the suits is the adhesive technology, such as vacuum pad, gecko adhesion, and Universal Vacuum Gripper hereinafter UVG.UVG is one of vacuum pad for uneven surface based on Jamming phenomenon of granular material.The conventional suit based on UVG had two functions, adhesion by UVG and power assist for wall climbing, but its weight was about 17 kg.In this paper, the belt-size assist suit about 5 kg is developed. We evaluate the performance of the developed suit by climing actual wall and moving along ceiling.

Conventional remote operation of an excavator has low work efficiency compared with direct operation. This is because it is difficult for the operator to recognize the state of the excavator and its surrounding environment. In addition, remote operation of an excavator has restrictions such as limits on the amount of information that can be transmitted, delays in communication, and tough working environments that cause the sensor failure. Therefore, we have developed a semi-autonomous control system which consists autonomy and human interpose. In this paper, we focus on human operation and present a variable admittance control approach to realize accurate environment recognition and high manipulation capability in a master-slave system. In the proposed variable admittance law, the admittance parameters are changed based on the difference between position of the master and slave. We implement the proposed method on the prototype of an excavation robot and verify the operation with three types of environmental external forces.

Spraying is one of the basic work for painting, coating, and consolidation. The purpose of this study is coating thickness control on the spraying work. In this paper, we formulate the thickness of painted surface toward automation. Simulation result and its problems are also described.

It’s necessary to recover quickly after disasters occur. However, human can’t work in seriously damaged area because there is a possibility of secondary disaster. Therefore, we use remote controlled construction machines to restore the damaged area. But now, the operator of these machines needs to be helped by a person who operates camera. The camera is manually operated by the operator. So there is a problem that a personnel is assigned to the camera operator and a difference in work efficiency is caused depending on the skill level. In order to solve such problems, we are developing an automatic tracking camera control system for construction machine operators by using image processing. However the reproduction of the manual camera operation is still not sufficient. Therefore, we aim to improve the system. In this paper, to improve the system, we recognized the moving direction of the construction machine from the image by image processing. In addition, using this recognition method, we improved the system to change the image displayed according to the recognized moving direction.

In recent years, improvement of productivity of construction industry using information and communication technology (ICT) is promoted. Particularly, for progress management of construction, estimation of sediment volume of construction sites is highly demanded. However, in the case of getting depth data with a sensor attached to the backhoe, the occlusion and narrowness of effective range of the sensor are problems. In this paper, a methodology to solve the problem by acquiring the depth data from a plurality of viewpoints is proposed. Accurate sediment volume is estimated by registrations of the point clouds from multiple views. Accurate depth data is generated by removing noise as outlier.

In this study, we designed a framework to evaluate quantitatively the operation status of human-operated work machines represented by work efficiency. The framework requires to reveal the sensor data affecting the operating conditions, the system consists of (a) formulating between explanatory variables (sensor data) and the purpose variable (operational status) and (b) extracting explanatory variables contributing to the objective variable. As a result of the experiments, the proposed system could grasp and evaluate the operating situation in a pluralistic and quantitative manner.

In recent years, the soft robots equipping the viscoelastic materials have been reported. However, sensors and control methods suitable for soft robots have not been established. In this paper, deals with a water-driven robot, and we measure the information about contact objects such as terrain and obstacles by using a pressure sensor. A water pressure sensor was attached hear the chamber, the contact state of the chamber was detected, and the breakthrough of the obstacle was confirmed.