3D object detection and pose estimation are an essential task for the bin-picking robot. We try to apply to solve 3D pose estimation task by using the pointcloud deep learning, which techniques are developing rapidly in computer vision. In this paper, we updated our 3D bin scene generation program we had developed for generating pointcloud deep learning dataset. We also propose a straightforward method which applies PointNet to estimate the 3D pose of a single object in the scene.

Since foods have irregular shapes, their individual recognition is difficult. We have previously proposed a food detection method for piled pieces of fried chicken. However, the method fails to detect a piece under certain circumstances due to a clustering error when the method handles a 3-dimentional point cloud. For the clustering process, we newly propose to use a soft clustering method in which each point can belong to more than one clusters if needed. The ambiguity does not become a problem for manipulation tasks and it will enhances the robustness of the clustering process. In this paper, we measure the improvement of accuracy with the new method.

Coding illumination-independent color information is an important feature for robotic vision because the apparent color of objects is strongly affected by the illumination condition. In this study, we developed a sensor system with color constancy that consists of a CMOS image-sensor and a field-programmable gate array (FPGA). To achieve color constancy, we designed a digital circuit that offers color constancy based on the center / surround retinex model, and the circuit was implemented on the FPGA. The digital circuit contains the following functions: integration of multiple exposure images, logarithmic transformation, and difference-of-Gaussian spatial filtering. The system was tested by presenting Mondrian under varying color of illumination. As a result, the system successfully reduced the influence of the color of the illumination.

In this paper, we develop an omnidirectional human tracking system for remote wide-area video surveillance; it can simultaneously track humans and objects moving at distant places in high-resolution video images by controlling a high-speed mirror-drive action vision that has three rotary mirrors omnidirectionally with 500-fps visual feedback, which involves real-time execution of CNN-based recognition for human detection. Its effectiveness is demonstrated by showing several omnidirectional video-shooting results when tracking moving objects and humans at distances of several dozen meters in natural outdoor scenes.

A mouse’s scratching behavior is one of the reaction effected by a new drug, which is rapid movement of its hind leg around its head. Currently, the number of mouse scratching behavior is counted by human, conventional automatic detections used a high-speed camera. Improving the accuracy of the detection, the automation for counting the scratching behavior is necessary and the evaluation must be quantitative. We propose the detection of the mouse's scratching behavior based on the shape and the motion feature using a wide-used camera. We apply Dense Trajectories to the classifier to identify the scratching behavior. The classifier could detect the scratching around 90 % accuracy. Based on the length of scratching, we could decrease the number of false-positive data around 30 % using the median processing.

Walking persons at a distance of 300 m are authenticated in video-shooting through a f = 1600 mm lens on a super-telephoto tracking system, which can control a mirror-drive pan-tilt system with high-speed visual feedback. Experimental results show the effectiveness of supertelephoto tracking in remote authentication for very distant persons in real scenarios.

In our study, a piston movement was tracked by controlling the gaze direction of the two mirrors in order to observe the piston that moves at high speed during driving of the engine. Since the piston motion is considered to be a sine curve, a mirror control signal of a sine curve was created using the function generator. In this experiment, the gaze direction of the mirrors was controlled according to the trigger signal from the engine under the condition that the engine speed was known. As a result, we confirmed that the piston movement could be tracked with some degree of accuracy.

Yohaku, which means the perceptual empty space in Japanese, is an important concept in the field of oriental art. It has a decisive influence on the aesthetic value of an artwork. Visual functions of yohaku in paintings have been deeply discussed from the viewpoints of art, philosophy and psychology in previous studies. In our research, we aim to propose an autonomous robot photographer system which takes people-pleasing photos by searching for an optical viewpoint for photographing. To construct this system, one of the main technical issues is evaluating and quantifying the aesthetic value of photographs. In this paper, we will introduce an aesthetic composition evaluation method by considering yohaku in a photo. First we confirm the definition of yohaku in a photograph and propose an automatic detection method. Then a novel composition evaluation method will be proposed based on then visual balance between yohaku and photographic targets in a scene.

Cross-ratio method can estimate point-of-gaze on a screen without performing a hardware calibration, and it has a tolerance of the head movement. However, this method requires the display, which near-infrared light-emitting-diodes are attached on, to detect screen corners. Therefore, we developed a polarization camera system to detect screen reflection on cornea without near-infrared-light emission. Most of the liquid crystal display has an internal polarizing filter, and thus, the transmitted light can be controlled by the external polarizing filter. We performed the evaluation experiment of cross-ratio based gaze estimation using display corner detection from polarized images.

Gaze information conveys what the people are interested in, and it is expected that the analysis of the point-of-gaze can be used for marketing research. However, the point-of-gaze is computed on a screen, and it is difficult to estimate the 3D point-of-gaze in the world-coordinate system for gaze data of multiple people. Therefore, we propose a method for estimating the 3D point-of-gaze using structure from motion. The semantic 3D map is reconstructed using a semantic segmentation and keyframes selected from scene images. Additionally, the 3D point-of-gaze can be also estimated on the 3D map without prior information.

The corneal imaging method is used for estimating the focused object, and it is expected that the applications are spread widely if localization is achieved. Therefore, we propose a localization method using illumination information reflected on the cornea in this paper. A map in which illumination information is registered is created in advance, and localization can be performed based on the appearance-based approach using phase only correlation and particle filter. We confirmed that the performance of the proposed method through the evaluation experiment.

Infant health monitoring leads to lighten the burden of child care. According to Hyogo medical association, the observation of multiple behaviors is needed for estimating infants health conditions. This study develops a sensor-embedded plushie device for frequent observation of four kinds of behaviors in temper tantrum. To make the device more familiar for infants, the device is based on an idea of“security blanket”which is favorite item for infants. Especially, small-sized and lightweight sensors are developed: a tensile force sensor and an omnidirectional bending sensor. Taking into account sanitation, the outer cover of device is detachable and parents can wash it. An experiment conducted with adult subjects examines the basic measurement performance of the device. Experimental results demonstrate the possibility to distinguish the behaviors in normal and temper tantrum conditions.

The World Robot Summit defined the ”Stocking and Disposing Tasks” as a part of its Future Convenience Store Challenge. So far, most participants who execute these tasks worked under an environment where the products to be manipulated are placed on a horizontal plane, such as on a shelf. However, in a real convenience store, this is not always the case. In this study, we focus on products that are hanged on a vertical plane by hooks. We use recognition to allocate the hole on the package, and find the position of the hook indirectly by AR marker. Finally, we showed the manipulation result through experiments by using a humanoid robot.

In this research, as a new approach to indoor childhood injury prevention in which parents' intensive supervision are not required, we propose a system to detect the dangerous situation of children in real time in a daily environment. The proposed system consists of an accident situation knowledge database, a dangerous situation evaluation system, a RGB-D camera and a child human model. This paper describes a concept of the system and experimental results for a fundamental function of the system.

This article proposes an external force estimator, which can detect the net external force applied from the environment in such a case that human touches the robot, discriminating it from the object-manipulating force. For this purpose, an inverse dynamics model based on a recurrent neural network is utilized to calculate the object-manipulating force from the object motion. By using the proposed external force estimator, for example, a compliant motion to the net external force can be achieved to absorb the contact force in collision between human and robot. The validity of the proposed method has been confirmed by experiments in which a robot manipulated an object with 6 degrees of freedom.

Carrier carts and carrier bags, often used to transport luggage, place a big burden on body when the user goes up stairs with it. A power-assisted carrier cart for both climbing stairs and moving on plane is proposed in this study. The cart controls the drive motor in accordance with the pulling force, so that the user can carry the luggage with a constant force regardless of the load capacity and the inclination of road surface. In this paper, a carrier cart that can be used in both cases of climbing stairs and moving on flat planes is designed. It is confirmed that there is a difference in pulling force between when there is assistance and when there is no assistance, and the effectiveness of the proposed method is also evaluated.

Human-robot handshake is a deeply explored topic in human-robot physical interaction. Lots of researches have focused on the pressure distribution during the handshake movements. But the grasping force and joint torque are usually intertwined, and the current approach measuring the contact force cannot separate the two. In this research we proposed a new device for the purpose of measuring the interaction force generated by joint movement and the grasping force generated by forearm muscles separately. And we used the proposed measuring device to investigate two different types of handshakes and verified its effectiveness in doing research on analyzing the interaction forces of handshakes.

Autonomous mobile robots are expected to work in a human-robot coexistence environment, and they require to have a proximal navigation framework considering approach and contact with humans. Such a framework requires a function to select a path based on estimating “the risk of approach” and comparing it with “the benefit of efficient movement by approach”. To move even in the proximal state, we develop a proximal navigation planning considering the distance between the human and robot (time to interference). The control framework estimates risks and benefits, and adjusts the risks as necessary, while performing proximal movement. The results of the experiment showed that the proposed navigation system could estimate the risk and benefit of approach, adjust it through inducement, and select a path that is natural and safe for the robot and human.

In this paper, we present a paradigm to improve the followability on a path-following task for the wrist by practicing repeatedly using vibrotactile cues. Vibrotactile cues were provided as a direction to modify the wrist’s path when the wrist’s position deviates from the desired path. The experiment was conducted to test the learning effect of the vibrotactile feedback on a participants’ wrist motion during three days. We found that the root mean squared error of the participants wrist’s trajectory decreased when provided with vibrotactile feedback. Additionally, most participants could reproduce the desired path in a more accurate way after removing the feedback immediately after training. This effect continued slightly over three days and the performance gradually improved after training.

In recent years, machine operation support become popular. Particularly, automatic driving technology has attracted worldwide attention. As a result, the situation where machines are operated by humans is decreasing. It causes lowering of human’s skills. Then, this research focused on “Subliminal calibration”. Subliminal calibration is a support method of training machine operation. This method is changes machine’s dynamics slightly so that human cannot notice. Subliminal calibration can improve performance without preventing proficiency. However, it has a problem that dynamics change is not constant. Therefore, in this research, we decide target dynamics and assist training for the dynamics. Our result shown that our proposed method is effective for assist training for target dynamics.

In this report, we present the study of combining human visual and haptic perception in human-robot interaction to realize collaborative manipulation. In our human-robot collaboration scheme, cognitive capabilities of human and accurate motion control capabilities of robot are well integrated. Human operator is designated for coarse global-motion under feedback interfaces utilizing human visual as well as haptic modalities. Fine local-motion in an active manner without involving human intention-aware is realized by a dynamic compensation robotic module with high-speed visual feedback. Primary experimental results verified effectiveness of the proposed method.

Modern robots are required to interact with humans and changing environments in the real-world as well as to perform different works in the factory. However, the environment changes unpredictably according to the time and actions taken by humans and robots, and humans make different decisions in various situations. To solve these problems with more adaptation and flexibility, this research focuses on reinforcement learning that learns appropriate behavior by trial and error with the environment. In addition, we hypothesize that robots could adapt to various personality of humans. We could express various strategies by changing parameters in reinforcement learning as functions related to the human personality, where the neuroparameters, actions and rewards in the learning strategy can be adapted and modified. We show that during even a simple cognitive learning task experiment, human personality affects the result, and human improved their learning result in a human-robot collaborative learning task.

Teamwork assistance and objectively evaluation are a purpose of this research. For that reason, we focus on differences in a solo task and a cooperative task. This method is“Concern for Others: CFO” which is an element of evaluating teamwork in real time. In most of the studies, the teamwork evaluation was analyzed by using questionnaire after an experiment. When teamwork skill requires assisting, those methods cannot apply in real time and objectively. Therefore, we had been suggested a method to real time analyzing objective teamwork support by the CFO. In this paper, in order to analyze teamwork factors in more detail, we analyzed teamwork skills of individual operators during a cooperative task in real time and clarified the relationship between operator’s teamwork skills and influence on teamwork performance. The results suggest the deviation of differentiated the ECFO influences cooperative task performance.

At the present time, robots manipulated by human are active in various fields. Also, there are improved minute mechanical operability technologies. For example,there are throttle valve control for driving and remote control robot hand, etc. There can be manipulated using various human senses. Among them, we use manipulating machine with bimanual. People often use dominant hand in life. Therefore, people aren’t good at using non-dominant hand as to compare dominant hand. I assumed, if our non-dominant hand use like dominant hand, operability become better. In this paper, focused on dominant hand model adapt non-dominant hand. Our result showed, operatabilty of non-dominant hand became closer to dominant hand.

The market size of service robots in Japan is expanding gradually. Along with that, “co-botics”, the development of robot technology sharing work space with humans, is progress rapidly. It may be possible for humans and robots to do collaborative tasks like carrying objects in industrial or home environment. The proposal of this research is to develop an expressive intelligent system for robots to cooperate with human in different tasks. For that purpose, the human affects are recognized through the force data using Recurrent Neural Network, and output as PAD (Pleasure-Arousal-Dominance) values. Using these values, we calculate the appropriate expression of the robot and implement it on Pepper with carrying task using Inverse Kinematics by null space control. After that, we conducted an evaluation experiment of the system and confirmed that human emotions are heading in the positive direction.

In recent years, one reason for the resignation of young people is motivation reduction. In order to solve this problem, companies are conducting education and management to improve motivation which usually facilitating human relation etc. However, if the assist is wrongly done, it could have a negative effect on motivation. Therefore, in this research, we focused on analyzing the relationship between motivation and biological signals. We compared the biological signals of a subject with no motivation.

The authors have proposed a method for automatic extraction of human action elements using HMM. When extracting behavioral elements from observed behavior data of human beings, since there are trace variations even when the actual human behavior is the same behavior, we absorb these variations and extract action elements from observation time series data It is necessary. In this section, we examine the influence on behavioral elements extracted by integrating previously generated similar HMMs based on distance between HMM using KL divergence.

In this research, we propose the realtime 3D reconstruction of whole body motion by using multiple cameras. Realtime motion capture is useful for sports training and cooperating human and robots. Video motion capture is superior in that the burden on the subject is small and does not disturb the motion. Placing a computer for each camera, and it is possible to disperse the calculation cost and shorten the calculation time at each computer. By separate the reconstruct process and the communication between computers, it is possible to estimate pose with a little delay. With this method, the average error of the joint position is 59.1 mm in realtime motion capture.

Knee osteoarthritis(Knee-OA) is a disease caused by age-related muscle weakness, obesity, or sport injury. One of the characteristics of knee-OA is rotation dyskinesia of the knee joint due to the degeneration of the system around the knee. In the present study, we focused on the rotation of the tibia relative to the movement of the ankle joint during walking, and we developed a mechanical orthosis, which induces rotation of the lower leg in conjunction with the movement of the ankle joint. We verified the effectiveness of the developed orthosis by measuring the amount of lower leg rotation and bone behavior while three subjects wearing medial wedge for imitating knee-OA patient performed walking motion.

In this study, a rollator equipped with an active driving chest pad was newly developed so that the elderly suffering weak lower limb can also walk independently with the rollator. Mechanism simulation was made and the new rollator is applicable to the user with body height up to 180cm. A small and high power motor was selected according to the necessary power for driving the user over 100kg.

Many existing exoskeletons use a perfect revolute joint as their knee joints, which does not match the movement of real human knee. Motors are also heavy and hinder user’s motion. This paper proposes a knee support device with a polycentric knee joint, which more closely matches the real knee joint. Application of thin pneumatic muscle as actuator reduces the weight of the device and its back-drivability gives the device soft characteristics. The proposed device was able to generate a maximum of 3.5 Nm torque at knee joint at 0.5 MPa applied pressure. Future work includes increasing output torque to allow the device to support standing up motion.

We have been proposing a finger rehabilitation support device for hemiplegic patients that can be used at home. This device instructs the patient to voluntarily lift the fingers and provides assistance when the patient is impossible to lift. In previous studies, we have shown an automated evaluation method which monitors the level of involuntary finger movement. However, a detailed procedure of the finger rehabilitation has not been clarified. In this paper, we show a detailed procedure of the finger rehabilitation. Especially, we discuss the safety issues during assisting finger lift. Specifically, we consider the following issues: (1) To avoid unwanted contraction of finger muscles by stretch reflex, the speed of finger lift assist is limited. (2) To ensure not to exceed the maximum excursion of MP joint, the angle of finger lift bar is constrained.

There is a need for a mechanism that allows patients with hemiplegia (especially paralysis in upper limbs) to continue rehabilitation at home without assistance by experts. To this end, a rehabilitation device that promote separative movement of fingers has been proposed. The device needs to identify the degree of voluntary control at each finger. To know the degree, the activities of the muscles that control individual fingers should be measured. Conventionally, the force generated by the fingertip was used. However, the force at fingertip is an accumulation of forces generated by multiple muscles. Therefore, it is impossible to detect movement of flexor and extensor muscles at each finger. A paralysis patient at a certain stage exhibits simultaneous contraction of flexor and extensor muscles. In this case, the previous measurement method does not identify the patient at that stage. Electromyogram (EMG) is known as a method of knowing the amount of activity of individual muscles. However, it is difficult to properly attach the sensors to the patient’s body without help by experts. In this study, we propose a new way to measure the activities of flexor and extensor muscles of fingers which uses displacement MMG (Mechanomyogram). MMG is a method of measuring local muscle activity by deformation amount of the skin surface using a simple noncontact optical sensor. By this, it is expected to realize more appropriate quantification of the recovery of paralysis by a sensing device which is easily used at home.

Hemiplegic patients caused by stroke perform rehabilitation training for activities of daily living during recovery period. In this study, a motor-driven new assist suit for shoulder joint which assists bending and extending motions of the shoulder joint was developed to assist horizontal reach movements for the rehabilitation training. In order to evaluate the developed assist suit for shoulder joint, healthy subjects simulated hemiplegic state by fixing their elbow and waist using the rubber belt. Then, evaluation experiments were carried out under the conditions of the subjects wearing and non-wearing the assist suit. Bending angle of the shoulder joint and surface electromyogram of the deltoid were measured, and by wearing the assist suit, assistance for the bending motion of the shoulder joint was confirmed. Therefore, it is considered that the developed assist suit is effective for those who are difficult to bend their shoulder joint.

It is known that walking dysfunction occurs the stroke patients with hemiplegia and it declines quality of life. Even though cadence of stroke patients is one of the most significant factor of walking ability, it is difficult to improve the cadence of gait without abnormal effortful movement. Thus, we developed the implicit cueing device to improve the cadence in stroke patients. In this paper, we used the weber-ratio to prevent the notice of the change of cadence, and calculated the ratio through experimental study with stroke patients. Further, we conducted the clinical trial in six stroke patients. As a result, our study suggested that proposed device could increase the cadence of stroke patients without notice of the changes.

As a new approach to treat osteoarthritis, we are developing an Intelligent Insole. Intelligent Insole can control the detection position of the foot center of pressure during walking and the softness of the insole accordingly. To check the state of the knee joint, estimate from the detection of the foot center of pressure. We thought of controlling the softness of the insole by using the MR fluid as the material of the insole and controlling the magnetic field strength. It is already possible to measure the foot center of pressure. It was confirmed that softness changes by applying a magnetic field to an object containing MR fluid in the insole.

This paper presents a new control scheme for prosthetic hand. The new control scheme accepts data from a camera and an inertial measurement unit (IMU) sensor as input, which makes prosthetic hands capable of visual as well as motion sensing. The camera captures the scene where the user is willing to grasp the object. From the scene the control system recognizes the target object, extracts its surface features and estimates its pose as well as position in the scene. Then the spatial relationship is constructed between the camera (prosthetic) and the target object. With the help of IMU sensors, the relationship can be tracked and kept wherever the prosthetic hand moves even the object is out of the view range of the camera. The information of spatial relationship helps to control the hand prosthesis in a fine manner. An experiment is conducted to verify the proposed method.

By introducing the muscle activity quantity evaluation by electromyogram, evaluation of muscle tone degree, muscle burden, muscle fatigue and so on, identification of muscle related to fatigue reduction, change of muscle activity amount and driving fatigue assessment It aims to investigate the relationship. As a result, it was found that the muscle whose muscle activity amount decreased at each support position was found, and it was also found that there was a relationship between muscle activity and fatigue value even at some supporting position.

Hanger-reflex is a phenomenon in which a sense of rotation and haptic force are caused by attaching a wire hanger to the head. In previous studies, we confirmed the occurrence of this phenomena in the head, lumbar region and wrist, but strong expression was not confirmed in the feet. In this paper, we investigated the conditions that efficiently and effectively cause the hanger reflection on the ankle.

In this laboratory, we have focused on mirror therapy that has been done for patients with stroke hemiplegia, and have developed and evaluated mirror therapy system using tactile information. This suggested that the tactile information might be effective in improving the paralysis function. Based on this result, we thought that tactile sense might contribute to improving sense of motor participation. In this study, we developed a mirror system that presents tactile sense for abduction and inversion of the wrist joint and evaluated how much the presence or absence of tactile presentation influences the induction of the motor participant feeling. As a result, it is suggested that the tactile information may contribute to improving the sense of motor participation.

I propose a control theory of a prosthetic hand based on a current sensor during controlling actuator for tendon drive. The prosthetic hand is thus controlled to protect the actuator by a software limiter utilized the current sensor. This paper describes experimental results in order to verify the advantage of the proposed approach.

Currently, gait analysis is used for pathological diagnosis. In the gait analysis, the measurement of the lower limb is the main, and measurement of the arm swing is not performed much. On the other hand, the arm swing is simple and it is easy to do it regardless of the place as compared with the movement of the lower limb. Therefore, by analyzing the arm swing, I thought that it would be possible to investigate signs of disease more easily than before. The ultimate goal of this research is to develop a system that encourages people to discover signs of disease by swinging arms on the spot and to undergo precision examinations. In this paper, as a foothold, we compared arm swing during walking and arm swing at foot stepping close to walking motion, looking for a common trend. However, we could not find a common trend.

The active corset is a pelvis belt type wearable assist device. In this paper, we experimentally verify the influence of the active corset on the waist and hip joint stiffness by tightening the waist. The active corset reduced the burden on the waist of twenty one of thirty subjects. Moreover, in fifteen of them, it is confirmed that their waist joint stiffness increased.

Upper limb motions play an important role in Activities of Daily Living (ADLs). Studies on the upper limb motion estimation have been conducted and applied to various fields. In this research, we develop a distance sensor array which can measure the upper arm deformation. An estimation method is proposed to calculate the elbow joint angle and the load in the hand simultaneously. The elbow joint angles are estimated using Support Vector Regression (SVR). The load is estimated using Support Vector Machine (SVM). In the experiment, we confirm the estimation accuracy by comparing the acquired and estimated angle and load.

Indoor tracking is in demand in various field. For example, pedestrian navigation, activity/health monitoring, augmented reality can be sited. VIO(Visual-Inertial Odometry) method has been established as a localization method, but it is difficult to estimate in the low visibility location. Although this problem can be solved by localization by the inertial sensor alone, it is known that large errors will occur. Recent studies have shown that the error can be significantly alleviated by using deep neural networks. Human beings move variously not only to walking or running. However, there are few studies focusing on this, and estimation accuracy for movement including multiple kind of movements such as including both walking and running is low. In this research we aim to develop a robust localization method for changes in human movement using inertial sensor.

Recently, applications of the illusory kinesthesia, which is a misdetection of own motion, have been focused on. The illusory kinesthesia is an illusion phenomenon caused by giving mechanical vibration to muscle tendons. This phenomenon is expected to be utilized for habilitations, motion skill assist for sports and so on. However, the mechanism and kinetic characteristics of the phenomena is not clear yet. Therefore, this study suggest the control of the illusion using the evaluation of the illusion kinesthesia. this paper investigates the effect to muscle tension of the illusory kinesthesia for the motion skill assist. So i tested it about the influence of the illusion on extension and flection of the arm Some experimental results suggested that the illusory kinesthesia relates to human muscle.

The galvanic vestibular stimulation makes the body inclined, In this paper, the authors propose a new notification tool for human employing galvanic vestibular stimulation. The authors conducted an experiment to see whether information can be reported efficiently by using this tool.

This paper discusses the methodology for extracting and classifying a style and characteristic component from time-series data of a drivers’ body motion. The body motion while driving is measured using four wearable motion sensors for acquiring posture and movement of each body segment. To extract the style and characteristic component, we use the singular value decomposition (SVD) of the measured data and evaluate the contribution of each sensor module for describing the physical meaning of the extracted components. From these results, the characteristic component of drivers’ body motion can be extracted by using the right singular vectors obtained from SVD. In addition, using the 2nd and 3rd right singular vectors, it is possible to classify the characteristics of holding the posture of the steering wheel.

We aim to quantitatively evaluate stress tolerance based on EEG. Stress tolerance is defined as resistance capability and recover ability. We measure stress tolerance of stressing subjects with white puzzle task. The questionnaire answered was compared with the measurement result of stress tolerance. Experimental results show the correlation between the recovering ability and the questionnaire.

Kinetic visual acuity (KVA) is one of the essentials in sports vision. Although virtual reality is thought to be a plausible tool for training KVA, it is not clear that the vision training in VR is effective in actual environment beause the focal point length is different. In this paper, the training effect in VR is evaluated in the real environment in the same visual motion task. It is confirmed that the task performance is improved after the training in VR in the medium target speed.