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 on a compliance mechanism that can passively absorb position errors, assuming the introduction of screw drive work using a vertical joint arm.

In recent years, the need of automation in the garment manufacturing has increased, due to labor shortage and an increased demand. In garment manufacturing the handling and manipulation of the fabric is estimated as the 80 percent of the total production line time. Therefore, there is a large research interest in designing grippers for handling fabrics. Our goal is to develop a robot system to handle garment and perform post-production tasks such as placing it into ironing or screen printing machines. In this paper, we describe the design of an underactuated robot hand that can grasp garment, and test its effectiveness through a task of setting the garment on a rectangular plate emulating a screen printing machine.

For factory automation, robots are required to handle large objects using both arms and hands. We apply view-based teach and playback approach to a dual-arm robot. In the playback phase, the robot corrects the instructed motion according to the estimated object position and posture. For this purpose, we describe the motion correction in consideration of the hand positions. The robot is thus enabled to use the arms and hands cooperatively. In the experiments, we show that the robot is able to handle a container by using the arms and hands through the teach and playback approach.

This paper aims to build dexterity into a machine. Specifically, we will focus on fitting precision parts, which is considered difficult with the current technology, eliminate "assembly failure" such as biting of shafts and holes due to machine learning, and aim to reduce total assembly time. In the final evaluation, the robot incorporating the appropriate control law and machine learning system performs the reducer assembly work, and determines whether the work time can be shortened or the assembly failure can be eliminated compared to experienced personnel. By advancing this research, I think that one assembly robot will be able to flexibly handle various tasks in the future. My ultimate goal is to create an assembly robot that is versatile and can do all the work without failure. This will solve problems such as "high installation costs", "difficult to set for each product", and "assembly work that cannot be automated".

Grippers with soft structures can be adapted to the object shapes, making it possible to handle various items without complicated control. They can change their apparent stiffness, for example, by using MR fluids or pin-array structures. However, MR fluid grippers have difficulty propagating magnetic flux to MR fluids. Pin-array grippers also have a problem that they need many pins to transfer object shapes. This paper proposes a magnetic pin-array gripper mechanism that combines MR fluid and pin-array structure. Each pin-unit has a flexible pouch filled with MR fluid, and a magnet can be inserted into the pipe inside the pouch. We conducted some experiments on magnetic flux density, pouch hardness, and object grasping. The results showed that the proposed mechanism can change the stiffness by effectively propagating magnetic flux to the pouches’ MR fluid. We also confirmed that it could grasp objects by its shape adaptability and stiffness variability.

The lack of successors is becoming a problem in factory. As a solution to this problem, there is much research how to teach human to robot by demonstrations. However, it is difficult to make accurate measurements while maintaining the quality of human movement. Therefore, in this paper proposes a high-quality motion and force information acquisition method without restricting the movement of the teacher by using a workbench with a force sensor and motion capture. To confirm the effectiveness of this method, in this paper conducted experiments to examine the usability of the tool used to acquire the movement and experiments to reproduce the acquired data by a robot, comparing it with other methods.

Peg-in-hole task has been studied as a benchmark task for robotic assembly. It involves two main phases: search phase and insertion phase. This paper proposes a method that uses Reinforcement Learning(RL) to achive search phase in the peg-in-hole task. In this method, the stiffness matrix for admittance control is generated online. The method uses a visual sensor to determine the relative position of peg and hole, and selects an appropriate stiffness matrix model. By using visual sensor, this method has two advantages: it reduces the number of learning episodes and speeds up the search process. The two advantages of the proposed method were verified by peg-in-hole task using a 6-DOF manipulator.

Marine resource survey by an underwater vehicle is expected. To realize cheap self-positioning for the underwater vehicle, we have been developing on the acoustic positioning system which is low cost and easy to maintain. The acoustic positioning system has four hydrophones which consists of molding commercially available MEMS micro phones with pressure-resistant resin, an audio device, and a micro controller. To evaluate the system, positioning experiments was performed in the small tank. In the experiments, the system detected delay time among four hydrophones, however its positioning was not good accuracy. In the future, we will improve the accuracy of positioning by interpolating data and considering the effect of baseline.

Japan is one of the rich countries from the point of view of ocean mineral resources. However, present mining ocean mineral resources are not unprofitable. We believe that the underwater metallurgical process is useful for the stable supply of resources. Underwater assembly is required to construct a plant that carries out metallurgical work. Underwater assembly is carried out by a Remotely Operated Vehicle (ROV) with a robot arm. The arm’s actuators should incline to the arm’s inertia, and the arm is thinner from the point of view of ROV’s posture control. Therefore, this paper proposes Hexa type parallel link as an ROV arm and executes validation underwater.

Most of the conventional hydrophones (underwater microphones) use PZT ceramic as the material, and there are problems such as variation in sensitivity, the necessity of high-magnitude amplification and high cost.

We developed and evaluated a low-cost, easy-to-use MEMS hydrophone that is waterproof and pressure-resistant while maintaining the bandwidth of a commercially available MEMS microphone.

The robot hand, which opens and closes like a jellyfish’s bell, can act as a pump that sucks water from the surrounding area and discharges it in one direction. In this study, we investigate the characteristics of the flow field produced by a simple submerged gripper and estimate the pump efficiency. First, the behavior of water masses taken in from the surroundings and the induced flow enhanced by vortices were observed by using a one-degree-of-freedom gripper that generates a strong two-dimensional flow and induced secondary flow. Although the efficiency obtained from the cross-sectional flow rate and the rated power of the motor is as low as 0.89%, this study shows that the gripper can work as a pump that sucks the surrounding water mass and discharges.

Since Underwater Vehicle-Manipulator Systems (UVMS) are expected to play an important role in futre ocean development, we have proposed position and force control methods for UVMS. In this paper, we show the usefulness of the proposed control methods by an experiment using a 3-link dual arm underwater robot that grip a fixed underwater object.

We have proposed a Resolved Acceleration Control (RAC) method for Underwater VehicleManipulator Systems which is a position control method. In this paper, we report experimental results of performance comparison between computed torque method, which is a control method for joint space and the RAC method.

Micro plastics are one of the recent important subjects of ocean environment, so that several studies are conducted at worldwide. In order to investigate micro plastics pollution near the bottom of the ocean and water column, the mechanism which is able to collect micro plastics without filter is proposed. The mechanism utilize swirling flow to separate plastics instead of filter. Micro plastics is collected by pressure difference of inside and outside of the swirling flow. The experiments of swirling flow occurrence and micro plastics behavior observation are conducted, and it is confirmed that the mechanism have possibility to collect micro plastics.

The purpose of this research is to develop an underwater robot for underwater exploration. We proposed a disk-shaped underwater robot (Cakram) with a traveling wave propulsion mechanism. The Cakram composed of disk fin with 8 active fin strips and control box. The Cakram can generate the traveling wave when there is a phase difference between the fin strips at a same frequency. The fin push the water by the traveling wave. The reaction force received from the water is used as a driving force. The propulsion performance of the Cakram is evaluated for various parameters such as frequency, phase difference and number of passive fin strips. Characteristics of the robot velocity and fin trajectory is observed. The results indicated that maximum velocity is 0.11 m/s by suppressing the deflection of the fin.

The current scallop fishing has some problems. The fishing can not harvest the natural scallops for raw consumption by the problems. To solve the problems, this study proposes a new method using a robot arm. The new method needs a gripper attaching a robot arm. From some demands that meet for scallops gripping, this research judge a suction gripper gripping with negative pressure is suitable to harvest the scallops. However, a suction gripper has disadvantages for gripping the scallops. Therefore, we develop a new gripper with a built-in spring to improve the disadvantages. Additionally, this study experiments with the developed gripper and evaluates the developed gripper.

The mainstream method of underwater position measurement is using sound waves (e.g. SONAR). However, this method generally has low accuracy and is very expensive. The purpose of this study is to develop a position measurement system that is more accurate than the conventional underwater position measurement systems. In this study, we proposed a position measurement system using an optical sensor as a new underwater position measurement system. The problem of the proposed position measurement system is the tracking performance of the tracking device. Therefore, we evaluate the tracking performance of the tracking device, which is a part of the proposed system, by simulation.

General underwater-welding has bad affects to welded metals, such as quick cooling and hydrogen mixing. Dry welding is used to avoid these effects in the water. But, dry welding spends much time and money. Therefore, we suggest new types of dry welding, using "Universal Vacuum Gripper (UVG)". In this study, we tried two experiences. First, we tried to make a dry area in the water using UVG. We put the air to the sealed area by UVG, after UVG vacuumed a board in the water. Second, researched what effects are appeared to UVG by welding inside it. We confirmed the UVG’s vacuum performance in the water, before and after welding metal board in UVG at the atmosphere. Comparing with them, we considered the new types of welding is available.

We develop a ROV to capture sea urchins. As a next step, we move the ROV automatically. To achieve this function, we need to know the distance between ROV and sea urchins. In this paper, a binocular stereo vision ranging system based on OpenCV-Python is designed and used with Viola-Jones algorithm for binocular recognition. Zhang’s algorithm method is used to calibrate the binocular cameras to obtain internal and external parameters, which provide necessary data for the subsequent image processing and stereo matching. In this paper, OpenCV-Python is completely used to build the system construction to achieve the simplicity of the platform, which has better robustness and high accuracy. Stereo matching adopts SGBM algorithm, which is relatively fast and real-time in stereo matching, and can be better applied to ROV sea urchin detection.

There are places where it is difficult for people to enter, such as inside sewer pipes. The development of robots is expected to make sewer pipe surveys unmanned. There is a risk of failure in screw propulsion due to the presence of string-like suspended substances in the sewage. Therefore, we adopted fin propulsion that imitated aquatic organisms. We developed an aquatic robot with four fins, and conducted movement control and posture evaluation experiments.

Mineral resources such as rare earths are deposited 2-3 m below the seafloor, and it is necessary to collect samples of seafloor sediments for detailed studies to elucidate their origin and existence. Currently, vertical drilling is commonly used to collect samples, but it is inefficient because the seafloor sediments are widely deposited horizontally. Therefore, we are developing a robotic seafloor explorer. In this study, we worked on horizontal drilling as the next stage. First, a mechanical model of the robot's penetration into the ground during horizontal excavation was developed, and the expected problems of horizontal drilling were discussed based on the model. After that, horizontal excavation experiments were conducted, and the problems were clarified from the result.

Valuable seabed mineral resources such as rare earth elements are deposited 2–3 meters into the deep seafloor, and sampling and analyzing the seafloor is necessary to clarify how they are generated and distributed. We have developed a seafloor robotic explorer that can excavate horizontally and collect samples of rare earth elements, which may enable wide-area exploration as multiple robots could be simultaneously implemented to autonomously search beneath the seabed. The seafloor robotic explorer may excavate sediments such as clay and sand before reaching the rare earth layer. In previous studies, it has been discovered that the rotation speed, penetration force, and characteristics of the sediment affected the excavation resistance. Therefore, we assumed that controlling the rotation speed and penetration force would enable an efficient excavation according to the ground characteristics. Therefore, in this paper, we examined the effect of rotational speed on drilling efficiency using dry silica sand ground.

Measuring the accurate position of AUV(Autonomous Underwater Vehicle) is essential for seafloor survey by AUV. In order to deploy positioning system easier, we developed the buoy named BUTTORI, which is capable of autonomous position keeping and measuring the position of AUV by GPS and USBL(Ultra Short Base-Line). In this article, we describe the evaluation result of position keeping and positioning performance of BUTTORI.

It is necessary to consider the influence of inertial force when designing a link-type force presentation system, position input system and rehabilitation support system. In this study, we propose a method for evaluating the effect of inertia and a method for improving its characteristics. Since the direction of the force vector applied to the gripper is significantly different from the direction of the acceleration vector of the gripper in 2-link system, the operation becomes difficult. It is shown that one property holds for the influence of inertia in 2-link system. In order to improve the inertia characteristics of 2-link system, a 4-link system consisting of closed link system is properly introduced. Then the above characteristics will be greatly improved. In order to derive the equation of motion of the closed link system, an undetermined multiplier vector, which is a force vector of constraint, is introduced.

In recent years, more and more people are becoming sensitive to coughs due to the spread of COVID-19. The type and frequency of cough can be used to determine the condition of the disease. Because conventional cough classification methods classified into just cough / non cough sounds, we are now developing a new classification method for several types of coughs from natural sound resources. In this paper, we proposed an evaluation function with Mel-Frequency Cepstrum Coefficients and Gaussian function.

The goal of this study is to generate a face mask stencil that fit individuals’ faces. We developed applicational software to automatically generate face mask stencils. In this software, a pair of photos, a portrait and profile of an individual’s face, are used to select some feature points as basic information for the stencil. We also developed a way to make an original mask from the stencil. Two types of face masks, the original mask and a square mask under uniformed breathability and color were evaluated with the particle image velocimetry. According to the results, lower level of leakage of breath with the original mask was found.

In this study, we have developed the dorsiflexion support unit (DSU) by using elastomer-embedded flexible joint (EEFJ). The EEFJ is a kind of compliant joint but has non-linear and adjustable mechanical properties. In our project, we have especially focused on the design method suitable for mentality of frail users. In this report, we proposed the basic structure of the DSU. We functionally and aesthetically designed the DSU and prototyped its first version by using three-dimensional printer and nylon resin. A pilot test of walking experiment with the DSU was conducted to check the mechanical strength of the unit and measurement method.

The purpose of this study is to develop and evaluate a communication robot that which controls behaviors of elderly people with dementia in care homes. The proposed robot has functions to detect standing up motion of elderly people and to encourage sitting down in a restroom. Experimental evaluations have been demonstrated at several care homes after operational test. As a result, it became clear that the proposed communication robot is effective for elderly people during toileting.

This study develops a training system for a multimodal comprehensive care methodology for dementia patients called Humanitude. Humanitude has attracted much attention as a gentle and effective care technique. It consists of four main techniques, including eye contact, verbal communication, touch, and standing, with more than 150 care elements. Learning to practice Humanitude thus requires considerable time. To provide an effective training system for Humanitude, we develop a HumanitudE AR Training System (HEARTS) realizing simultaneous sensing and interaction by combining a real training entity and augmented reality technology. This paper presents a new version of HEARTS named HEARTS 2 consisting of HoloLens 2, a soft doll, and a full-body wearable tactile sensor.

It is important task to construct high-quality nursing-care and rehabilitation environment for elderly people and people who suffered brain stroke in the aging society. In this research, we are developing nursing-care and rehabilitation robot system for the standing-up motion. Although several standing-up assist machines have been developed, all that systems are not able to support trainee ’s natural standing-up motion. Therefore, we introduce a standing-up support machine which can assist user ’s pelvis and knee dynamically. This system can help not only reducing burden of user ’s muscle but also natural standing-up motion like daily living. In this paper, we describe a new transfer-style device for standing-up motion training which users can step into from the side．

In order to assist stair climbing movement for ALS patients and patients with spinal cord injury, we are developing a stair climbing machine. Although most stair climbing machine adopt crawler mechanism, crawler system has some problems such as damaging floor and slip dropping. Therefore, we adopt linkage mechanism for stair climbing. This device also has advantage that users can keep their posture horizontally．In this paper, we describe this device’s mechanism which can transfer its center of gravity in a horizontal direction.

Affective touch enables to relieve stress and pain. If robots can perform affective touch, they would support the patients who cannot receive enough care from therapists. In developing robots for touch care, it is important to develop an end-effector that can evoke pleasant feelings as much as or more than a human hand. In this study, we hypothesize that the feeling is changed by the variety of the contact area between an end-effector and the human body. Then we develop a simple and reconfigurable end-effector consisting of a plate and flexible or rigid joint parts. Its simpleness contributes to its productivity and we prepared and compared three end-effectors having different contact areas using its reconfigurableness. In the experiments, we evaluated 1) the contact area of the three end-effectors quantitatively, and 2) the difference in a person’s valence and arousal when touched by the end-effectors subjectively. The results suggest that 1) the end-effectors with flexible joints have a larger contact area than the end-effector with rigid joints, and 2) the end-effector with a larger contact area give lower arousal values and higher valence than the end-effector without enough contact area.

This paper proposes a wheel-type walking assist orthosis for persons with hemiparesis. The user wears the orthosis under the bending hemiplegic knee. This system comprises three modules: a driving module, a sensor module, and a telescopic clutch. The user is stable and need not lift the orthosis when walking because the driving module has wheels constantly contacting the ground. The sensor module monitors the stride and the walking speed. To maintain balance, a telescopic crutch plays a supporting role when walking, whereas it plays a central role when standing up and sitting down, and when ascending and descending stairs. Experimental results show that a non-disabled person can travel not only indoors, but also outdoors.

This paper proposes the method of muscle tracking control using a force sensor. A lot of people have stiff shoulder or lower back pain. A massage robot is studied to eliminate this problem. A force sensor based on a finger pad of massage robot gets moment, shearing, and reaction force data. the tracking control using force control is proposed to estimate the state of the target muscle based on these data. A force sensor gets the reaction force data and performs object tracking and pressure tracking in one axis direction. Using this method will be able to massage the target muscle correctly.

In this study, to verify the reduction effect of lumbar burdens under wearing the endoskeleton-type power assist suit developed in our previous study, the lifting motion of heavy objects was analyzed using the musculoskeletal model analysis software, AnyBody Modeling System. Then, the muscle activity of the erector spinae muscles, and the disc compression and shear forces between the L4/L5 lumbar vertebrae were evaluated and compared in different conditions: under un-wearing the power assist suit, wearing the passive-type power assist suit with two or one artificial muscles, and wearing the active-type power assist suit with two or one artificial muscles. As a result, it was confirmed that under moderate assist force, the assist effect can be obtained without adding the burdens to the lumbar vertebrae, even if the endoskeleton-type power assist suit was used.

This paper proposes a balance control method for humanoid robots. Our method consists of CoG position control and torso angle control. In our method, the soles of the robot do not have to touch the ground. Simulation results show that proposed controller is effective for balance recovery after external force is applied and moreover the controller works well even if the robot is in the air.

In this paper, we propose a bipedal trajectory planning method to realize fast and human-like walking. Based on existing studies on the center-of-mass trajectory of human walking, we design swing leg trajectory to reduce the lateral fluctuation of the center-of-mass trajectory of a humanoid robot. Generated trajectory was evaluated simulation and proved that proposed method can walk faster than previous methods. The simulation results also imply that proposed method is better than previous method at walking stability.

The human’s fast and accurate manipulation movement has important factors in the torso motion. We investigate the direction of using the torso on the manipulation in the human-living environment and analyze the relationship between the torso and the legs in the movement. In this research, during fast motion of raising and lowering an object, we hypothesize that they keep their body posture to prevent the disturbance of stability. By wearing a vest that could fix the torso in the subject, we reproduce the restrained state of the robot． We derive the joint angles of the torso and both hip joints using the joint positions obtained from the motion capture system. We measured the ZMPs of both feet using a 6-axis force and torque sensor installed on the sole of the shoes and evaluated the stability and compared the stability of moving the object quickly and slowly. From the evaluation of joint angles and ZMPs of both feet, when they moved the object faster, we found that they spread the crotch sideways and they keep the torso in the middle of the target positions.

A new control architecture for real-world-oriented robots is presented. Unlike the subsumption architecture, the upper-layer functions modulate the lower-layers rather than overriding their outputs, and utilize them to adapt the system to situations. A simlutaneous exploration and navigation (SEAN) for a biped robot is implemented based on the architecture and evaluated on a computer simulation.

For a robot with redundant sensors and actuators distributed throughout its body, it is difficult to construct a controller or a neural network using all of them due to computational cost and complexity. Therefore, it is effective to extract functionally related sensors and actuators, group them, and construct a controller or a network for each of these groups. In this study, the functional and spatial connections among sensors and actuators are embedded into a graph structure and a method for automatic grouping is developed. Taking a musculoskeletal humanoid with a large number of redundant muscles as an example, this method automatically divides all the muscles into regions such as the forearm, upper arm, scapula, neck, etc., which has been done by humans based on a geometric model.

In this paper, we propose an inverse kinematic posture generation using redundancy in the musculoskeletal humanoid shoulder complex, and actual learning of the self-body image for object manipulation, and confirm its effectiveness by realizing the steering wheel operation of a car.

This paper describes the mechanisms of the upper body of a biped humanoid robot (KBHR-2). The KBHR-2 is composed of 23 DOFs, such as 6-DOFs in each leg, 4-DOFs in each arm, 3-DOFs in the waist. The movable angles of the KBHR-2 are about the same as those of a human. The robot is mainly made of aluminum alloy. The aluminum alloy is anodized to prevent the current from flowing through it. Its height is 1.40 [m] and its weight is 46.0 [kg]. Through the basic experiment of the upper body, the effectiveness of the mechanism of the upper body is confirmed.

This paper describes an automatic navigation system for a bipedal humanoid robot. The automatic navigation system consists of a 3D map generation system, a real-time obstacle detection system, a path generation algorithm. In the 3D map generation system, the humanoid robot estimates its own position by using a landmark observation model, and a 3D map is generated based on 3D environmental information obtained from an RGB-D camera. The real-time obstacle detection system detects new obstacles by using the 3D map and the environmental information acquired from the RGB-D camera. The path generation algorithm calculates an optimal path to a destination by using a preliminary map. If the robot detects a new dynamic object while walking, the robot will adjust its local path. The effectiveness of the automatic navigation system is confirmed through a walking simulation using ROS (Robot Operating System).

In this study, we realized to make a humanoid robot carry multiple stacked objects. When a humanoid robot carries stacked objects, the positional relationship between objects that are not grasped changes due to the impact during walking. This misalignment accumulates and eventually causes the objects to drop. In this study, we propose a method to prevent sudden changes in the position and posture of object by smoothing the hand trajectory, and modify the misalignment by tilting the entire object. Through experiments using a simulator and using an actual robot, we verified that the proposed method is effective for carrying stacked objects by a humanoid robot.

Humanoid robots are expected to be work in complex and dangerous disaster sites, because they are capable of performing more complex manipulation tasks than other disaster response robots. This study tackled motion generation in debris removal work by humanoid robots. In debris removal work, it is a problem that some contact states of target objects are not only unknown but also easy to change. In this study, we addressed the problem of uncertainty and variability by sequentially estimating unknown contact states of target objects. In addition, we monitor the grasping state in hands using multiple sensors, generate contact constraints and modify robot motion. This enables the robot to detect failure and to compensate for the uncertainty of the estimated contact state. In our experiments, we demonstrated that the combination of monitoring the grasping state of debris and regrasping it as recovery motion was effective to prevent debris from sliding down and execute a robust debris removal work.

This study presents dynamics computation and control of a hybrid link system that integrates rigid- and soft-bodies. It is a challenging problem to install a softness in a robot system, which is an important factor in human body. Softness achieved by human muscles and ligaments contributes to dynamic motion. Flexibility of a sports prosthetic leg allows a handicapped person to run. However, traditional algorithms of dynamics computation for a robot system or human skeletal model only consider a rigid-body multi-link system. Recent progress in soft robotics such as piecewise constant strain (PCS) model provides the way to compute dynamics of soft deformation with a low computational cost. We construct a hybrid link system integrating rigid-body and the PCS model. For controlling a humanoid robot with soft links, we implement a dynamics computation with a floating-base and derive the center-of-gravity Jacobian matrix of the hybrid link system. Moreover, we demonstrate a forward dynamics simulation of a humanoid robot with prosthetic legs.

The purpose of this study is to reduce an error in the weaving motion of a 3-DOF planar manipulator. For this purpose, it is necessary to identify the physical parameters of the robot to calculate the feed-forward torque for weaving. Since noise in the experimental data is a problem in this process, we identify parameters that have small influence on the motion error even though the existence of the parameter errors by considering sensitivity analysis and error covariance of the parameters. As a result, it is confirmed by simulation that the error of the weaving motion is reduced by the control using the feed-forward torque of the identified parameters compared to the parameters identified by the Least Mean Square.

A robot in a dynamic environment needs to avoid static obstacles and perform motion planning while considering dynamic obstacles. This paper presents a Configuration-time-space (C-time-space), in which time-series information is added to the configuration space (C-space). The C-time-space is represented by the mapping position and joint displacement of the robot at each time in the generalized coordinate space. Motion planning using the C-time-space can generate a trajectory that avoids dynamic obstacles. The experimental evaluation results indicate that compared to the conventional method, the trajectory distance of the robot arm can be shortened in environments where dynamic objects exist using the proposed method.

Accurate identification of minimum set of dynamics parameters is required for high-precision or high-speed motion control of robots. Although the Least Mean Square method is generally used for the identification of these parameters, the identified values are not a consistent estimator because of errors included in the experimental data. Instead of considering these errors, this study regards that the minimum set of dynamics parameters fluctuates stochastically. Based on this idea, we propose a stochastic identification method to obtain identified values that have small effect on the control even when there is such fluctuation. The effect of minimum set of dynamics parameters on the control can be considered by calculating the sensitivity to the state equation including the controller. The simulation and experimental results show that the proposed method provides identified values that have small effect on the control.

Currently, in the field of regenerative medicine, cell processing is being performed manually. Such process is labor intensive and expensive, and its efficiency still needs to be enhanced. Recently, automatic cell culture apparatuses that are equipped with a vertical articulated robot have been proposed. However, the automation of all cell processing tasks complicates the system. This study aimed to develop a simple and rational cell processing system through the combination of the tasks performed by a robot and those performed by a human. In a previous study, we improved the efficiency of discarding and injecting tasks using a robot arm in the media changing process. In the present study, we propose an algorithm to predict injection stop timing by machine learning.