During the last decades, the progress of technology has greatly contributed to the ecological survey of the animals. Bio-logging, which reveals the ecology of wild animals by attaching a small logger to the body, is a typical example. On the other hand, applying external stimuli to animals and observing their response is a useful approach to understand their behavioral principle. However, it is difficult to apply the approach to wild birds. To tackle the problem, in this paper, we propose a wireless chemical tele-pump system to enable to inject drugs by remote command. Furthermore, we develop a prototype and confirm its feasibility by experiments.

The purpose of this study is visualization of the search ability on different breeds and individuals for search and rescue dogs by using handler and dog behavior measurement suits. In this paper, to visualize the dog's search ability affected by multiple factors, the dog's movement speed, the amount of movement, the time spent finding victims, the number of handler commands, the relative distance to the handler, the distance from the handler is defined as six quantitative measures, and visualized search ability. In addition, to show the trade-off and tendency of the search ability in each dog, the overhead notation of search ability using diagrams is proposed.

Insects antennae react to an individual potential response depending on the odor. We constructed the system which can discriminate the receiving odor by using the difference of the response waveform. In this study, we constructed two subsystems, "measuring system for the insect’s antennal potential response" and "classification system based on waveform differences”, and combined these systems. In order to discriminate odors from waveforms, we used “Shapelet Transformation”, which time-series data patterns as feature values for classification. We performed odor discrimination experiments for Bombykol and Citral and we confirmed the system can be obtained a highly accurate discrimination result of 92.0%.

The conception rates on dairy cattle have been decreasing every year. To solve this problem, estrus detection in dairy cattle using sensors is required. The purpose of this study is to detect the estrus behavior of dairy cattle using an inertial measurement unit (IMU) in hilly and mountainous areas. The increase in acceleration was used in estrus detection based on the increase in activity during estrus in dairy cattle. The true positive rate and the true negative rate were 70 % and 8.4 % respectively. This result shows the potential of the proposed method for estrus detection in dairy cattle.

In recent years, agricultural working population in Japan has been declining, and the employment of agricultural workers is aging. Because agricultural work is a tough job and the yield is highly dependent on the weather, making income becomes unstable. To solve this problem, we have been developing an automatic raspberry harvesting system that aims to grow high value-added crop raspberries at plant plants and automate harvesting.

In this research, we develop the raspberry fruit recognition system using deep learning or machine learning system which is mounted on a harvesting robot that harvests raspberry fruit automatically.

In this paper, the water surface drone, which has a laser distance sensor and an infrared camera module connected to the Raspberry Pi, is discussed. The damage amount of lotus root caused by duck reached 20 million yen in 2006 in Tokushima prefecture. The present our drone system has two characteristics for the way of avoidance from obstacles and detection of ducks. The first is that the laser distance sensor is rotated by RC servo motor to find some stems of lotus. Another one is that the images taken by the infrared camera before and after the duck arrives are used to find the duck. Some experimental results for detect obstructs are verified and the improvement of our system are shown.

The automatic pumpkin harvester is expected to be developed in order to solve the problem of the aging of farmers in Japan from the 2010s. This paper describes a newly designed grasping link-mechanism of a robot hand that will be utilized for automatic pumpkin harvesting. In order to develop the robot hand, the dimensions and weight data of pumpkin were measured in a field. The dimensions of the linkages for the robot hand was designed according to the measured data. The force required for grapping a 5 kg pumpkin was calculated in the static condition. Moreover, a 3D model of the robot hand was designed based on the decided size of the linkages. In addition, the robot hand was attached to a robot arm that was mounted on a base vehicle.

The autonomous driving tractors were developed and utilized in the fields for farming in recent years to solve the problem of the shortage of farmers in Japan. The global positing system (GPS) is mainly used for the autonomous driving system for which can provide absolute positioning information in the outdoor fields. However, in the conditions of adjacent buildings or trees such as windbreaks the position error of the GPS localization receiver can be several meters, which is impossible to be used for precise autonomous driving while spraying operation etc. Therefore, a local sensor such as a camera is expected to be developed to solve the problem. In this paper, a color camera that was mounted at the front of a tractor was utilized to capture images in a corn field. In addition, the crops were detected using an artificial intelligence (AI) model - single short multibox detector (SSD). The crop line was detected using Hough transform.

The purpose of this study is to develop a lighter and smaller power assist device that can assist the user’s arm to reduce burden on workers. The power assist device consists of a spring and a linkage mechanism in order to use a passive gravity compensation mechanism. In this paper, we used a spring without power supply. The spring is used to adjust the posture of the device. The effectiveness of the proposed power assist device is verified by the result of the experiments.

Recently, mechanization and automation of farming are required. However, they have not been improved enough. To thin out young pears with robot arms, it is needed to detect and localize fruits’ positions and sizes. In this paper, the method of localizing them is proposed. To avoid being occluded by surrounding leaves and other fruits, data of fruits from some viewpoints are integrated to localize fruits’ positions and sizes by multi-view stereo. In this study, data from other viewpoints are matched only by geometrical features. In the experiental results, fruits could be localized within 18 millimeter.

This paper describes a method of scene recognition for mobile robots able to recognize plants that can be traversed, especially for agricultural purposes. Conventional mobile robots rely on scene recognition methods that consider only the presence of objects. Those methods, therefore, cannot recognize paths covered by flexible objects, such as plants, as traversable. In this research, we tackle this problem by integrating deep learning-based image semantic segmentation and a conventional 3D mapping method to generate a 3D semantic map that stores the object class information of each region. We demonstrate the effectiveness of the proposed methods by the experiments of robot navigation inside the lab with a path with plants covering it, which imitates the paths in greenhouses.

Planting works of Allium × wakegi Araki bulbs is too heavy for old farmers to practice in the field because they have to work with half vent posture. Thus, the mechanization of the bulb planting process has been desired. When the bulbs are planted into the soil, farmers vertically put bulbs into the soil, and half of the bulbs are below the soil surface eventually. From a realistic viewpoint, it is necessary to develop an affordable system that can realize such a complicated process. This paper reports a trial of the development of a planting machine for the Allium × wakegi Araki Bulbs.

An agricultural robot is highly required due to decreasing of birthrate and aging population. So far, fully automatic agricultural robots without any manual operations have been proposed. However, there are still many problems to be solved regarding economical and technical side of putting this robot during the harvesting season in practical use. To cope with the problem, in this paper, we propose a tomato harvesting robot system with automatic tomato recognition, and feedback functions by operator. The operator corrects the result of automatic recognition by a robot obtained from the database.

Recognizing branch connections can contribute to automatic management of branches and fruits. For example, branch connection information is helpful to determining pruning branches. A fruit can be identified by recognizing which branch the fruit is attached to. It is necessary to accurately recognize branch connections to manage fruits and branches. A cause why the branch commections are not recognized is the lack of information. Therefore, it is necessary to compensate for the lack of information. In this paper, We describe autonomous movement control of the drone to compensate for the lack of branch connections data.

Chestnuts harvesting with a stoop is hard work in chestnut fields on the mountainside. Moreover, the harvesting work at the slope is in danger of sliding down. We propose a small chestnut-harvesting mobile robot operated by a remote controller. This robot consists of two main parts that are a chestnuts collection part and a basket to store the chestnuts. The collection part with many thin rods takes chestnuts using its rotation on the ground. The chestnuts in the collection part are removed by a claw to put them in a basket. In the experiment, the robot successfully collected chestnuts on the grass.

The purpose of this research is to develop a robot which assist to harvest citrus fruits in an orchard in the Seto Inland Sea island. The robot carries the fruits between trees and monorail. During this work, the robot needs to transfer the fruits to load carrying bed on the monorail. In previous paper, several mechanisms were developed. However, the robot didn't have the fruits transferring function. Therefore, the transferring mechanism are developed.

In this mechanism, basket with fruits is transferred to the load carrying bed. We produced the robot and evaluated the performance of the robot. Especially, the transferring mechanism is described in detail.

Mowing with a handheld power mower on narrow slopes is sometime dangerous to agriculture workers. Autonomous mowing systems will lessen the worker’s burden of heavy mowing and reduce danger works. A general-purpose small mowing system have been developed on the laboratory that the author belongs in. It can mow weeds on the range specified by an external computer automatically. However, it cannot go straight on slopes and it is better to have independent control system without extra computers on narrow space. We verified the effectivity of yaw-angle feedback control with PID controller and roll-angle feedforward control system.

In this work, we propose a structured approach of selecting and designing a set of grippers for an assembly task. Compared to experience-based gripper design method, our approach speeds up the design process by automatically generating a set of initial design options on gripper type and parameters according to the CAD models of assembly components. We use mesh segmentation techniques to segment the assembly components and fit the segmented parts with shape primitives, according to the predefined correspondence between primitive shape and gripper type, suitable gripper types and parameters can be selected and extracted from the fitted shape primitives.

In this paper, we propose a cap tightening method combining redundant parallel-driven soft fingers and a 4DOF SCARA robot. First, the degree of freedom of redundant parallel-driven soft fingers required for cap tightening work was shown from the problems of combining a soft hand and a robot arm. Next, the operation procedure of the cap tightening operation was described. Furthermore, we proposed a system combining redundant parallel-driven flexible fingers and a 4DOF SCARA robot, and controlled the position and orientation of the cap with respect to the body by visual feedback control. The control performance is shown by experimental results.

It is needed that the robotic bin picking in a warehouse grasps and moves various weights and sizes objects rapidly and safely. We developed the grasp plan algorithm with safety ratio that simulates contact area between the robot hand and the object and realizes a robust grasping. According to experimental result, we verified that our supposed algorithm could distinguish between success and failure of an object transfer by robot prospectively.

In this paper, we present our research on using the HRP-2Kai humanoid robot to take a long, thin and flexible belt out of a bobbin. By proposing a novel non-prehensile manipulation strategy ”scraping” and using a real-time 3D shape estimateor for 2D flexilbe thin object. We show experimentally humanoid robots can complete this complex task in industrial production.

We have been proposed D-PaDY, a robot co-worker that improves work efficiency in the door assembling process of automobile production lines. D-PaDY is a system that reduces the workload by automatically delivering a door and supporting assembly work. Since the vehicle body flowing on the line has a position error from the reference position, it is necessary to accurately recognize the vehicle body position in order to position the door with respect to the vehicle body more precisely. This paper describes a method for estimating the position error of a vehicle body to automatically perform accurate door positioning for each vehicle body flowing on a line.

The occurrence of convex defects on the surface of the product during casting is unpreventable. The aim of this paper is to propose a robot manipulation algorithm with a feedback control system, for an automatic removal processing of convex defects instead of manually, which consists of tool tip reaction force and position feedback control. Removal experiments were carried out to identify the appropriate value of each of the parameters used in the proposed feedback control system. It is shown that better removal time and high-quality surface can be achieved by applying the proposed feedback control in the removal process of convex defects, comparing with the non-control method. Besides, sufficient machining performance obtained from the removal processing experiment of convex defect with different shape proves that the proposed method is effective on convex defects with unknown shape.

Learning robot actions using a simulator has many advantages as compared to one using a real robot. However, transferring the policy learned in simulation to the real robot is difficult because of the influence of the “reality gap”. In particular, the visual reality gap is a severe problem for the End-to-End controller, which uses images as a state. In this paper, we propose a real-to-sim image transfer combining domain randomization with latent dynamics. Our proposed method can predict future real-to-sim images, even if we could not obtain images. We validate the effectiveness of the proposed method by using real images in a manipulation task.

This paper proposes a judgement system of success / failure of lid fitting tasks based on active sensing. It is important to automate the success detection of various tasks. When people closed a lid, it can be judged not only by watching but also by actually applying force and touching it. This is considered to be confirmed on the basis of the force generated by the fitting action. In the proposed method, a rapid change in force during the fitting action is detected, and then the stiffness on the control side is changed during active sensing. In doing so, the judgement can be made. As a result of verification by experiment, it was confirmed that success or failure of the task can be determined by active sensing based on forces before and after the moment of lid fit.

In the screw-driving work by the work robot, a position error occurs when the accuracy of the robot or the position setting accuracy of the work is low. If the position error is larger than the tolerance, the bolt cannot insert and the screw-driving work cannot be smoothly.

For smooth operation, the robot must have a characteristic (compliance) that flexibly absorbs position errors. As a method of obtaining compliance, there is a method of applying a compliance mechanism. This method has high responsiveness and is effective for screw-driving work.

In this study, we report the results of designing, manufacturing, and experimenting a compliance mechanism that can passively absorb position errors, assuming introduction into screw-driving work using a vertical articulated arm.

We have developed the Parallel Wire-Type Teaching Device (PAWTED) for teaching the desired trajectory to a robot easily and quickly. Generally, an operator uses a teaching pendant to teach the trajectory of required task to an industry robot. However, since it is difficult to move the robot freely with the teaching pendant, the teaching task consumes a lot of time. In this study, we developed the PAWTED3 with a deburring probe for deburring tasks and evaluated the effectiveness of the PAWTED3 by conducting a comparative experiment with the PAWTED3 and the teaching pendant. In this experiment, the operator taught the 5 deburring trajectories to the six-axis industrial robot. As a result, it was confirmed that the teaching time with the PAWTED3 was reduced more than 60% compared to the teaching pendant. Moreover, the PAWTED3 was found to be more effective for difficult trajectories such as a circle.

Introduction of Automation of mechanical work using robots is progressing for reducing workload and enhancing productivity in industrial fields. Therefore, we conduct research for the purpose of dexterous assembly work by robots. In particular, we study the automatic assembly of jig parts used in production lines by robots. We not use a parallel gripper or dedicated hand, but use a 4-finger 12-DOF hand. It's equipped with an RGB-D camera on palm and a 6-axis force sensor on wrist part and each fingertip. The purpose of this study is to insert an electric screwdriver into the hexagon socket head screw with 5mm width across flats for fixing the jig. The position of the driver's tip was estimated using the AR marker and the RGB-D camera. The position of the tip and the screw were aligned and inserted.

The purpose of this study is to automate the work of providing cotton cloth, which is often used for underwear, to sewing machines. In previous study, as the effectiveness of the method of winding a cotton cloth with a brush was confirmed, we take over the method and introduce a novel hardware and manipulation procedure for picking up more stably. Moreover, the proposed hardware was mounted on the tip of an articulated manipulator, and then the practicality on the tasks of picking and releasing were evaluated.

This paper discusses automatic insertion of an endoscope into narrow holes of metal parts. The endoscope is attached to the hand of a robot arm. Using image processing of the image obtained from the endoscope and using fine-tuning of the position, the robot accurately inserts the endoscope without contacting the metal parts. In the detection of the holes, we use a two-stage mask treatment of the image in order to automatically obtain the center of gravity of the holes. Furthermore, we also consider the inclination of the roll angle of the endoscope.

A gantry robot for wiring of thin cables of the heater is developed. The robot is required the tension control of the wire as well as the position of the wiring tool attached at the hand. Linux CNC, which is conventionally applied for the position control, is expanded for the tension control synchronized for the location of the hand traced on the wiring pattern. Firstly, path planning for the wiring that avoids collisions between the tool attached at the hand and the pin jig installed in the wiring pattern is proposed. Next, tensions during the wiring by a human operator is measured for investigating the tension control scheme for the robot. In the last part of this paper, a mechanism for controlling the tension was designed and its operation was verified.

Underwater robots capable of performing seafloor exploration are highly useful to discover resources and unknown creatures. However, traditional underwater robots equipped with screw propellers may be harmful to surrounding environment and ecosystem. This motivates us to develop a robot that incorporates the shape and function of aquatic animal, in this study, a stingray. We have developed a stingray robot consisted of soft material and cartilage that mimics the structure of natural counterpart. We show that the use of cartilage can be possible significantly improve the swimming speed and propulsion efficiency.

In recent years, research and development of autonomous ships are actively carried out in the world. Research on autonomous ships are divided into three parts, remotely-controlled ships, a remote control center and a communication network. The remotely-controlled ships has an autonomous navigation system which realizes collision avoiding navigation. The autonomous ships are threatened by cyberattacks because the communication network is normally connected to the public network. We need to take not only tangible factors such as network isolation but also intangible factors into account to ensure cyber security. In this paper, we develop a software of remotely-controlled ships to deal with alternation of a hardware configuration.

Hydroelectric power plants have long-distance waterways called headraces. However, there is a problem in this waterway that the efficiency of power generation is reduced due to the nesting of insects called caddisfly It has now been removed by humans. But this is inefficient and hard work. Therefore, I have considered a cleaning robot that can remove the nest of caddisfly. The robot has four thrusters to stick to the wall, two thrusters for propulsion, in addition, by having a cleaning brush, This robot can clean while running while sticking to the wall. As a result of the experiment, it was confirmed that the robot could move underwater while sticking to the wall.

Fish can perform instantaneous accelerations and sharp turns in an emergency. In this study, we focus on the highly mobile underwater locomotion peculiar to fish, and develop fish robot using snap-through buckling. In our previous studies, we developed a mechanism can generate continues snap-through buckling mechanism using DC motor and triangle cum. In this paper, we propose a continuous snap-through buckling mechanism combined with wire drive, and develop a fish robot using the mechanism. The performance of the proposed mechanism was evaluated by swimming experiments. From experimental results, we confirmed that the turning performance of the robot was improved by the proposal mechanism.

We aim to develop an underwater swimming robot which can move quickly like the fish. Our adopted approach is to use the snap-through buckling to generate powerful fin oscillation. The series of our works have revealed the validity of our approach. This paper focuses on the caudal fin joint which connects the body part to the fin. Movable caudal fin joint will oscillate the fin with large amplitude, which may result in the more quick locomotion. This paper compares the swimming results in the case of the fixed fin joint and those in the case of the movable one.

The authors are developing a thruster type small underwater robot descending vertically from the surface of water for the purpose of the mud sampling at the bottom of lakes. The robot requires a method of which the risk of leaking water is low to communicate with the mother ship. In this paper, we propose a communication method with the mother ship using an optical wireless device. The effects on the communication speed of the proposed method due to an installation error which occurs when attaching the optical fiber cable to the robot are investigated. As the result, it was confirmed that the proposed communication method can communicate at a sufficient speed for the required specifications.

There are many marine resources including mineral and energy resources in Japanese EEZ and those have been surveyed by Underwater vehicles. It is expected that time-varying data will be continuously obtained with low risk in environmental surveys of the seabed. Therefore, we proposed the long-term observation method using Cable-restricted Underwater vehicles.

We have proposed a leader-follower system for wide-area surveys using passive towed robots. In this system, we require a communication amoung robots to estimate and control the relative position of the robots. However, if the communication is interrupted, the motion control becomes impossible. In this article, we propose a novel formation control method that does not require communication between robots. We develop the catamaran with a heading angle detection mechanism to measure the direction of water flow, then validate the proposed method by experiment.

We are developing an autonomous surface vehicle (ASV) capable of simple DPS for investigation and light work in shallow water areas such as coastal areas and lake, ponds. The body of the ASV used a surfboard as a buoyant body to make it smaller and lighter, so that it can be transported to Japan by loading it in a one-box car. At present, the ASV has already been completed and has completed sea trials by remote maneuvering in the actual sea area. For automatic control, angle preservation is already implemented. This paper describes the results and issues of the waypoint control test for reaching the destination, which was performed for automatic navigation.

In this paper, we describe the development of a position holding thruster system for ASV (Autonomous Surface Vehicles). At present, JAMSTEC, JAEA, and TUMSAT are jointly working on the investigation of radioactivity in mud deposited in estuaries in Fukushima Prefecture, Japan. The main objective of our project is unmanned mud collection using an ASV. The side thruster system was developed to enable an ASV to perform mud collection, and it can be maintained the ASV position. Furthermore, the stern thruster was changed and used for as forward and backward thruster to maintain the ASV position automatically. From the results, it was possible to maintain an ASV position when the wind was weak. However, when the wind was strong, the thrust of the stern thruster was weaker than wind, and we confirmed that the an ASV could not be maintained the its position.

At present, the offshore development field is attracting attention as an energy supply target with the expansion of energy demand. This paper describes the results of research on an underwater floating mobile robot that can be active in the field of ocean development. The underwater mobile robot is an autonomous underwater vehicle (AUV) with a new structure consisting only of an outer shell and an internal drive controller. This aircraft has a new drive mechanism that uses the difference in angular kinetic energy to obtain thrust by rotating the outer shell. A prototype of an underwater floating mobile robot was fabricated, and the operation of the driving mechanism was verified.

In recent years, the demand for underwater robotic arms applicable for inspection in revetment and offshore facilities is increasing. However, thick and heavy hermetic waterproof structure, high reduction gear systems, rubber sealings with high friction are needed for conventional underwater robots, which lead to a large amount of cost and lose flexibility against the environments. In this paper, to achieve high back-drivability underwater, a 2-DoF planar robotic arm using low reduction gears, and brushless DC motors naked to water is proposed. Experiments verified that the developed robot has back drivability underwater.

Autonomous underwater vehicles (AUVs) observe a wide area by using mounted camera and LED lights and is used for bio-resource survey. However, AUVs is difficult to capture the sample useful for scientists. Previous AUVs can capture only specific benthos that its developer decided the target before resource survey, but scientists for fisheries resource can’t chose sampling target during observation by AUVs. To realize an efficient bio-resources survey, the authors developed AUV that can capture the target instructed by the operator during observation. Because the slurp gun installed on the AUV has no canister, the AUV catches only one sample in one dive. This paper explains developed AUV and slurp gun with the canisters, and reports on sea trials using them.

In recent years, with the deterioration of underwater structures, the underwater drones have been proposed to conduct inspections and work in the sea. In order to stably work in the flowing sea, the drone is desirable to fix itself to the wall surface by adhesion. In this study, we developed the sucker fish drone based on Universal Vacuum Gripper (in short UVG) which is a vacuum suction gripper attached to uneven surface and curved surface. We evaluate the performance of the UVG in underwater, for mounting it on underwater drones. In addition, we conducted the experiment to adhere on the wall like a sucker fish by the developed drone.

Generally, GPS used for position measurement has a large attenuation of radio waves in the sea, and it is difficult to obtain position information in water. In this research, we use the seafloor image of the onboard camera to estimate the position of the underwater robot. The amount of movement of the underwater robot is measured by matching the seafloor feature points from the seafloor image. After that, we verify self-localization of an underwater robot by combining the acceleration, angular velocity, and geomagnetic data obtained from the 9-axis sensor.

Using design engineering, we revised the way printers are used as information output devices. In pursuit of the design required as a human-centered design home & office application, we will introduce the best form for home & office. There are two features. The first is "inconspicuous". Therefore, it eliminates the printer's presence. The second is "do not put". By proposing a printer with these two features, we will expand the printer usage scene. We first developed a prototype and evaluated it. Finally, we conducted a questionnaire survey to confirm the effectiveness of the proposed method.

Although the demand for fermented food exists nowadays, its production is governed only by rules created from the experience of the manufacturer. As a result, inexperienced producers face multiple problems until these rules are learned, increasing their operational cost. Therefore, in this research, we aim to develop a support system that helps its users, by applying Internet of Things technology, to easily produce fermented food while maintaining its quality. In this paper, bran pickles are preferred as the fermented food, and we aim to build a system to monitor rice bran floors. Therefore, we try to identify a correlation between odor, pH, and temperature of the rice and taste of the pickles.

This paper proposes a new tabletop cooking robot equipped with microwave and a arm. Conventional large cooking robots require a lot of space for installation and cannot be easily changed after installation. This problem would be solved by reducing the size of the cooking robot, but the cooking performance will be lower than with large robots. Therefore, we define the requirements for this cooking robot, and then created a prototype robot with microwave and a small arm. We show the good ability of this prototype robot through the cooking experiments for a simple recipe. The future goal of this research is to develop a tabletop cooking robot that can be easily introduced into ordinary households.

The purpose of this research is to develop a gripper with a surface structure that exerts a high frictional force on tableware’s even when their surface is wet with water, oil, detergent, etc. (wet state). Several textures for the finger surfaces were prepared by providing patterns of grooves and protrusions, to suppress the friction reduction even in the wet state, and then evaluated to see which pattern is appropriate for the finger surface.