Echography has the advantage of minimally invasive, and compactness. However, echography is demanding on the examiner and requires specialized knowledge and experience. To solve this problem, many studies have been reported. We have been developing an Echography-Assist Robots, EARs, the reproduces the probe scanning of a skilled examiner and performs cooperative diagnosis between a human and robot. Input the position and posture data of the probe trajectory acquired by the mo-cap system. However, it was impossible to reproduce the procedure. EARs behave unstable near the singularity depending on the joint angle of the robot, or suddenly stops. In this study is to investigate the safe range of motion of the robot arm. In this paper, we report an experiment on the stable range of motion of EARs by calculating the manipulability using the joint angles of the arm and the Jacobi matrix when tracing the human abdomen with a probe.

Echography is a diagnostic imaging method with the features of low cost, minimally invasive, real-time, compact, and simple. However, since doctors and sonographers diagnose many patients in a day, mental and physical fatigue is a problem. Therefore, we have developed a system to reduce the physical fatigue of examiners by assisting the operation of the ultrasound probe. In this paper, we constructed a bio-signal measurement system to evaluate the effectiveness of the robot assist. In addition, we analyzed the examiner’s mental fatigue using a wireless electroencephalograph to evaluate the effect of ultrasound diagnosis by the robot on the examiner. As a result, we confirmed that the fatigue increased during probe scanning in cooperative motion with the robot, and decreased during probe pushing.

Echograhy is a diagnostic imaging method with advantages such as low cost, minimally invasive, real-time, small size, and simplicity. However, it is well known that the ability of probe scanning by the examiner is highly dependent on the expertise and experience of the operator. Therefore, we have proposed echo-assisted robotics (EARs). In this paper, we have developed a system for reproducingsurgical procedures, which works by considering the patient ’s body shape and movements, and an educational system using this system. By attaching a depth camera to the robot ’s tip, the system can respond to differences in the patient ’s body shape and body movements. Furthermore, we combined this robot system with an echo education system to build an echo education robot system that can acquire images by data matching. In the future, we plan to improve the sensor implementation to build a system that avoids singularities and create an image control system that measures the torque of the probe tip to control the force and feeds back diagnostic images to the robot control.

One of the problems that HIFU treatment is not automated is the inability to focus ultrasound on the target cadaver for a long time due to organ movement caused by respiratory motion. In order to solve this problem, we proposed AEMA (Angle Estimation Method by Area) using YOLACT++. In other words, AEMA can follow the cross-section of the organ in which the target organ is reflected, and the robot can keep following the target organ, so that the ultrasound can always be focused on the target organ during treatment in the future. In the future, we aim to improve the estimation accuracy in short-axis images.

Some conventional researches have proposed models that express the relationship between the ankle joint angle and the ankle joint torque during passive movement using a time-varying elastic modulus . However, the model can not explain the hysteresis characteristics during high-speed motion because the gravity and inertia terms are regarded as zero. Therefore, the gravity and inertia terms are added to this model in this study. The validity of the proposed model was evaluated by showing that the values of viscoelastic torques estimated by using the model were almost the same under the experimental conditions where the inertia and gravity terms were different.

An artificial eye model which novice doctors can train a glaucoma surgery has not been enough developed yet. A realistic simulation system for medical education is required. However, although we have been got a sensory evaluation with several eyeball models by experienced doctors, material measurement of only an elastic modulus was lacked to understand the relationship between design of the eyeball model and doctor’s feelings. Therefore, to develop a bionic eyeball for glaucoma surgery practice, it is necessary to extract some mechanical characteristics during incising a sclera with a knife. In this study, we aim to clarify the characteristics through a simplified incision test with an artificial sclera equipped with a layered structure. As the result, we were able to quantify the doctor’s feeling by measuring the reaction force when incising sclera with different fiber diameters in the model.

In this study, we investigated the segmentation and registration methods to develop an intraoperative cancer localization navigation system to support complete resection of prostate cancer. segmentation using YOLACT++ was very accurate in the prostate. YOLACT++ was able to segment the prostate with very high accuracy, and it was also able to obtain sufficient accuracy for the outline of the tumor near the capsule of prostate. Therefore, we evaluated the effectiveness of YOLACT++ for cancer localization prediction. For registration, we compared the accuracy of affine transformation and projection transformation, and evaluated that affine transformation was effective. In the future, we will examine the accuracy of nonlinear registration methods such as the B-spline method and registration methods between different modalities.

The purpose of this study is to estimate the motion trajectory of an ultrasound probe using Visual SLAM technology，and to reproduce a doctor's probe scanning using a bed-type robotic ultrasound diagnosis system (RUDS)． We investigated a method to mimic the motion trajectory of the RUDS probe by using SLAM technology．The proposed method was able to perform the imitation motion，but the error became larger as the distance from the origin increased．

This research proposed the measurement system of quantitative indices for evaluation of surgical skills during simulation of inner limiting membrane (ILM) peeling, one of the most complicated inner eye surgeries, on Bionic-EyE. This system consists of the 6DOF IMU attached to the forceps to monitor extra hand tremor and the artificial eye module embedded with the quartz crystal resonator (QCR) force sensor to measure tiny contact force of the forceps applied to the retinal model on the eye bottom during simulation. To extract hand tremor element from the accelerometer, we adapted extended Kalman filter to estimate posture of the forceps and remove gravitational acceleration at sampling rate of 10 ms. Using this system during ILM peeling simulation by an ophthalmologist, we succeeded in obtaining features of forceps motions quantitatively due to high resolution of the QCR force sensor and the posture estimation method.

In recent years, a new type of coronavirus (Covid-19) has been spreading throughout the world, and suppression through vaccination is desired. As a solution to this problem, we proposed the concept of a ”vaccine injection robot system” that automatically performs vaccination and dispensing operations. However, in order to achieve this, it is necessary to overcome some technical problems, and in particular, this time we aimed to automate the detection of puncture point in vaccination work, which is one of them. This puncture point means the intersection of the line connecting the upper margins of the anterior and posterior axillary folds and the center of the acromion. Therefore, this paper proposes an approach for detecting the acromion position using a force sensor and contact parts, and reports the results of preliminary experiments.

We developed a forceps mechanism with embedded optical fiber sensor manufactured by a 3D printer to realise force sensing in robotic minimally invasive surgery. First, we improved our previous design of the forceps to integrate the sensor into the mechanism while manufacturing by a 3D printer. Then, we conducted temperature, position input and loading tests to verify that the proposed mechanism can properly measure the force.These evaluations showed a positive feasibility to be implemented in surgical robots in the future.

Training vitreoretinal surgeries such as peeling of inner limiting membrane require a surgical simulator with elaborate retinal model and contact detection method. Quantitative evaluation of forceps contacts to the retinal model because excessive forceps contact might cause loss of eyesight. This paper reports the detection method using the digital image correlation for forceps contact to the retinal model. An evaluation system for forceps contact detection was constructed. This system has an optical system incorporating 850 nm LED, a CCD camera, infrared pass filter under a retinal model. This system detects the forceps tip and forceps contact to the retinal model from the change of correlation coefficient between the initial and measured images. From the evaluation result, the forceps contact was detected at a load of less than 0.1 mN.

Echocardiography probe handling during a cardiac scan physically strains the cardiologist. Scanning multiple patients during one working day may lead to musculoskeletal disorders. In addition, during pandemic situations contact examination should be avoided. In this paper, we proposed the design of a probe manipulation robot to tackle these issues. We combined precise 3D printer motors for translation movement and soft actuators for tilting movement and soft contact with the patient body. Stiffness of our actuator was confirmed to correspond with skin stiffness. All the required movement for probe handling can be performed by our robot.

A hydrogel heart model with a temperature memory function and a strain sensor was fabricated for quantitative evaluation of cauterization depth using an ablation catheter. The ablation depth was related to heat temperature, heating time, and input load. The hydrogel heart model elaborating the heat properties of the human heart was fabricated, and input load was evaluated by incorporating a flexible strain sensor. The cauterization depth was measured by the color analysis of the cauterized model with temperature memory function. An evaluation result confirmed a proportional relationship between the cauterization depth and heat temperature, heating time, and input load in the fabricated model.

The purpose of this study is to examine whether it is possible to extract and interpret the main cross section from the ultrasound image by using the deep learning model in the abdominal ultrasound image. Using image classification and segmentation, programs that determine organ enlargement and atrophy have proved useful.

Multi-DOF (Degree of Freedom) endoluminal forceps is a useful device for assisting endoscopic treatment. However, articulating endoluminal forceps with conventional mechanism have a large number of parts, thus have some problems such as size and cost. Multi-DOF endoluminal forceps using a compliant mechanism has a smaller number of parts, and is advantageous in terms of miniaturization and cost reduction. In this study, we developed a mechanical handheld device without a motor. In order to accurately manipulate the bending motion of forceps with a joystick, a force transmission model has been formulated. With reference to this force transmission model, a prototype was manufactured with a 3D printer. Finally, we succeeded in lifting the simulated intestinal mucous membrane by the prototyped forceps, confirming the effectiveness of the proposed method.

In ultrasound therapy, a clear ultrasound image is necessary to determine the exact irradiation position. However, there is a concern that the accuracy of irradiation may be degraded due to the black noise caused by the reflection of sound waves on hard tissues such as ribs and stones. In this study, we aim to automate ultrasound probe manipulation to support monitoring of ultrasound diagnosis. The acoustic shadow and the target organ in the ultrasound image are detected by deep learning, and the control model avoids overlapping imaging in real time based on the overlapping area information of the two. This makes it possible to monitor the treatment target without any acoustic shadows.

The purpose of this study was to evaluate the appropriateness of diagnostic images for automated ultrasound operations. Therefore, two experiments were conducted. The first is to detect the target organ using deep learning. The second is to identify missing parts in the diagnostic images. In the first experiment of organ detection, the IoU and Dice coefficient were 0.947 and 0.972, respectively, indicating high accuracy.In the second experiment to identify the missing parts of the image, the percentage of correct answers for the missing parts on the right side of m was 75.3%, while the percentage of correct answers for the missing parts on the left side was 99.1%.

We propose a wearable system that detects abnormal gait due to imbalance using a 3-axis accelerometer in order to promote awareness of initial symptoms of cerebrovascular disease. Cerebrovascular disease is one of the most serious diseases and many people suffer from the sequelae of severe illness. In order to prevent severe cerebrovascular disease, early detection of initial symptoms is necessary. Therefore, we used the 3-axis accelerometer built into the smartphone, and analyzed the gait data by AutoEncoder, a kind of machine learning, for the abnormality detection. In the experiment to obtain the gait data, the subject wore the simulated elderly experience material and walked 10 m with the smartphone fixed on the back of the waist. As a result, it was clarified that it was possible to detect abnormal gait with the AutoEncoder by focusing on the combined acceleration and considering the walking tendency of individual walkers.

In order to prevent product accidents in the home, it is essential to prevent not only accidents involving individual products but also those involving product pairs that consumers actually use in combination. The purpose of this study is to identify the product pairs that increase the risk of accidents due to product interaction. For this purpose, an interaction analysis of product pairs was conducted by text mining from the big data of household accidents. Using a logistic regression method, 19 pairs of products that have significant synergistic effects (p<0.05) were identified. In addition, we developed an accident visualization system for providing risk information to consumers by integrating an accident case database, a combination risk database, and a visualization engine.

A Safety confirmation system for power outages during Large-Scale disasters in isolated islands and depopulated Areas is need to consider the elderly. We proposed an input method that combines a small liquid crystal display and a push-button switch, and compared it with a touch panel method through experiments. The results showed that the operation execution time and the input error rate were approximately same, and the questionnaire results showed that the system was easy to understand because it had a click feeling.

In this paper, the change of dynamic stability (MoS) by two different calculation methods of BoS is compared and investigated. As a result, it was confirmed that the MoS tended to increase with the increase of the cane tip position when the end point of the BoS was the cane tip position where the cane touched the ground, which is the conventional method. On the other hand, when the end point of the BoS was the point where the moments of the three loads applied to the cane, the left foot and the right foot were balanced, the MoS tended to become smaller as the cane tip position was increased.As a result, it was confirmed that the MoS tends to decrease as the cane tip position is increased. This conclusion is consistent with the feeling of actually using the cane, so the proposed method is considered to be appropriate.

Unintentional injuries are one of the most common injuries among children below the age of 5 years. Although the concern about the need for injury prevention is raising among caregivers, preventable injuries are still the 3^rd leading cause of death among children aged 0-14 years. Hence, there is a need for an effective system that can empower caregivers, parents, or schoolteachers in taking the right and faster decisions to protect the lives of children. In this paper, we propose a situation-aware system for unintentional injury prevention with the integration of image processing using RGB-D camera, spatial knowledge graphs and an accident database. This paper describes an implementation of the system and the advantages that it can offer in understanding the possible dangerous situations in the indoor environments like homes or classrooms and provide alternative suggestions on what kind of preventive measures be taken to minimize such injuries.

We aim at communication between human and robot where the robot claps hands according to a person who sings without accompaniment. This paper describes a method of acquiring the tempo from unaccompanied singing. So far, we have proposed a method to estimate the tempo from the difference in volume and the change in pitch. In this paper, we propose a new method for shaping time-series data for analysis by combining these two methods. A signal that matches the tempo estimated using the new method is sent to the clapping robot, and we confirme whether the robot can clap hands.

Currently, Japan is facing a serious shortage of labor due to a decrease in the working-age population. As a result, the demand for service robots is increasing. In this paper, we have focused our research on receptionist-guide robots. The interface robots, ApriPoco and LRF, are used to detect a person and the robot responds. We combine the robot's response with a human movement trajectory prediction system, Gaussian process regression. The Gaussian process regression can model nonlinear relationships, and can plot the acquired trajectory smoothly. In addition, we developed a system that can set the content of speech according to the position and direction of a person. We conducted a verification experiment and showed the results.

In this study, we use a software called OpenPose estimate the posture of carrying. A depth camera is used as the imaging device. In this study, we use a photographic device called Realsens. From the front view, we examine three points: the position of the hand supporting the baby's buttocks, the angle of the arm, and the angle of the arm supporting the baby's back. From the side view, the angle of the arm supporting the baby's back and two points of the back muscles are examined. From the space vectors with the depth information corresponding to the coordinates of the feature points, we calculate the values. We will consider a system to visualize the evaluation values from the analysis results in an easy-to-understand manner for users.

At present, it is said that about 10% of people in Japan show symptoms of postpartum depression after childbirth, and it is a symptom that troubles many parents. One of the factors of postpartum depression is the posture of carrying an infant. If the infant is not carried in the correct posture, the body is overloaded and the risk of back pain and wrist tendinitis increases. In this study, we focus on the posture of carrying an infant, and aim to create a system that can easily detect bad posture and prevent postpartum depression and injuries by using posture estimation.

In the case of bedridden patients with weak ability to expel sputum by spontaneous breathing or coughing (including the case of using the ventilator due to the former factor), it is necessary to insert the suction catheter into the trachea through nasal/oral or tracheostomy to suction sputum. In the process, the patient should be able to find out the position of sputum accumulation by listening to the diagnosis, and find out the position of sputum accumulation by manipulating the suction catheter. However, because the patient cannot breathe while suctioning phlegm, suctioning phlegm needs to be done in a very short time. In order to support this kind of suctioning phlegm, there is a need for a continuous multi-point lung sound diagnosis system that can judge whether suctioning phlegm is necessary or not and estimate the phlegm reservoir position. In this paper, we propose a method of discriminating normal and abnormal lung sounds using the auscultation data, which is necessary for the system of constant multi-point lung sound auscultation.

In this paper, we proposed indoor human location estimation by the multiple sensor nodes for monitoring the elderly indoors naturally. Optical distance sensor nodes are equipped with Bluetooth Low Energy (BLE) communication function. These nodes gather the data that measure the distance to a person using distance sensors and transmit the data to the management system. Using this data, the management system estimates human location using machine learning as multi-class classification problem. The experiment showed the potential of these nodes and this method to be used as a location estimator.

In this study, we developed an automatic detection algorithm of vessel and skin regions in a transversal ultrasonography image on the arm. We also developed an algorithm to generate 3D model from detected areas for training and assisting nurses' puncture on vein. In the algorithm, the candidate regions of vessel were detected using U-net, which is a kind of deep learning method for segmentation, and then appropriate regions were selected based on vessel properties. The skin regions were also detected using U-net. The 3D polygon data was created from paired pixels in sequential images. The experiments based on single arm scan data demonstrated that our developed model have capablity to detect vessel and skin regions and feasibility to confirm blood vessel running under arm surface.

In this paper, we propose to develop a handrail with pressure sensors and load cells in order to estimate the posture of griping the handrail and the grip strength. Eleven healthy male participants gripped the developed handrail in two situations. As a result of classifying the posture of grasping the handrail using SVM, the postures of griping the handrail can be identified with high probability. Grip strength was estimated using several machine learning methods. The gradient boosting method showed the best accuracy and the possibility of detecting approximately 4-kg decrease of muscle strength at least.

In Japan, where the birthrate is declining and the population is aging, the burden on caregivers is increasing. In addition, it is burdensome for the caregiver to be assisted in daily activities, which may further impair self-esteem. Focusing on assistance, we have developed an automatic excretion processing system. In the conventional automatic excretion processing system, the position of the receiver, which is the receiving part of excrement, shifts depending on the posture, and it is highly rigid, so it is restrained by the care recipient. Therefore, in this research, we proposed a pneumatic soft receiver that can perform flexible compression and increase the rigidity only when necessary. In this paper, we report two types of soft receiver prototypes and basic experiments using them.

The pulse oximeters currently in widespread use measure pulse rate and blood oxygen saturation by contacting the skin surface. Therefore, it can be a burden to users with weak skin. Therefore, we conducted a basic experiment aiming at non-contact measurement of blood oxygen saturation and realized non-contact measurement of pulse wave by measuring light using an XYZ camera. In addition, we investigated whether it was possible to measure oxygen saturation by measuring the tendency of oxygen saturation to change with breath holding. In this paper, we report the results of these two experiments.

In recent years, the use of robots to address the shortage of human resources in the fields of nursing and care has been attracting attention. However, most of the conventional guardian robots introduce only a single recognition element, and there are few examples of practical applications. In this study, we propose a system for understanding abnormal situations using two recognition elements, skeletal recognition and acoustic analysis. In this paper, we analyze actual abnormal sound samples in order to investigate the functions and elements of this system. In addition, we compare the detection of the subject and the abnormal posture using image recognition and skeletal recognition, and examine the usefulness of the system.

The purpose of this study is to realize safe and sustainable nursing care by partially automating endotracheal suctioning to free up nursing staff time. As an elemental technique, an automated system for postural drainage is developed to move the position of sputum in the trachea by changing body position.

There have been numerous studies on the subject of tele-operated robots. Due to the challenges brought by the COVID-19 pandemic, we would like to expand this field of study to remote robot experiment systems that allow users to conduct robot experiments safely from a remote location. We propose a system that provides a robot monitoring feature and an experiment interface. The robot monitoring feature uses an RGB-D camera to enable collision avoidance and workspace violation in case of dangerous behaviours. Users can upload their experiment scripts and run them through the web interface hosted by the system. Experimental results validate that users can conduct simple experiments safely through the proposed system.

Generally, autonomous mobile robots in outdoors have to prepare occupancy grid maps, set waypoints. In addition, users need to be familiar with robot development tools such as ROS. This paper aims to develop a system for any user to navigate autonomous mobile robots to any destinations in unknown outdoor environments. Specifically, waypoints as routes on the electronic map can be provided easily using developed mobile application. An general action model based on deep reinforcement learning allows the robots to travel without path planning on the grid maps. This means that the proposed system does not need advance preparations such as setting waypoints and building occupancy grid map. Some experiments with multiple robots and users in outdoor environment were performed and evaluations for developed mobile application were provided in the paper.

The purpose of this study is to develop a cooperative navigation system based on sharing positions among multiple mobile robots in the same environment. The proposed system uses Rowma (Robot Web Manager) that integrates any ROS-based robot systems for topic communication among robots. In the case that several ROS-based robots publish same topic names, they must be differentiated in sharing topics. The proposed system shares the topic incorporating Rowma UUID into the topic name to distinguish topics representing location information among multiple mobile robots. The cooperative navigation experiment with the proposed system was performed. In the experiment, each mobile robot can select a route to the destination according to shared positions of other robots.

In the past, we have described a method of information collection using a mobile robot in a closed space during a disaster. However, there is a problem that the mobile robot gets lost due to the disconnection of wireless communication during information collection. Therefore, to prevent mobile robot from getting lost in a closed space, a distress prevention method using low throughput communication has been proposed. However, it is difficult to transmit images using low throughput communication in TCP/IP compliant teleoperation. In this paper, we discuss an image transmission method that is robust to packet loss using low throughput communication.

The objective of this research is to improve the convenience of teleoperation of a small 4-wheeled robot. To this end, we adopt a teleoperation method using a steering wheel type controller and aim to improve the convenience of teleoperation by presenting new images to the operator. After creating a virtual space and displaying images from a omnidirectional camera on a Virtual Reality Head Mounted Display, a small 4-wheeled robot is teleoperated with a handle-type controller. In the experiment, a virtual mirror was set up on the virtual space calculated from the human's viewing angle and neck range of motion, and the image behind the moving direction was transmitted.

The research on force sensing communication technology using haptic devices over the Internet has been actively carried out. So as to realize communication using the sense of force between remote areas, we have developed a system that can stably control bilateral in an environment with communication delays. In order to construct a position, posture, and force-sharing system using an omni-directional mobile robot and to perform bilateral control, we fabricated an omni-directional mobile robot. In the force-sharing system, the effectiveness of the wave filter and scattering matrix was evaluated by simulation with/without the wave filter and scattering matrix.

The evolution of robotics technologies has produced many kinds of robots, for example, cleaning robots, guidance robots, and transfer robots. In the future, these robots will coexist in the same public facilities. Without traffic rules for mobile robot systems, highly qualified technologies must be implemented on mobile robots. Consequently, the robot market will not be expanded. To resolve this problem, RRI (Robot Revolution & Industrial IoT Initiative)’s subcommittee has discussed robot system architecture independent on highly qualified technologies. In this paper, mobile robot system architecture featuring an operation management map and an arbitration function is proposed. This architecture enables multiple mobile robots to run smoothly with each other by relatively simple technologies.

This paper introduces our activity about interface standardization in collaborative manipulation. The cooperative manipulation system is a system that is widely used not only for industrial applications but also for service robots. However, each company is currently implementing its own system. To solve this situation, we discuss about the software interface standardization for collaborative manipulation system to improve the usability and maintainability of their modules. In this paper, we show the overview of the standardization.

As robot development using open-source software (OSS) becomes popular, the issues related to OSS for robot product development are becoming apparent. Since 2021, the authors have set up a committee to discuss these issues at the Robot Revolution and Industrial IoT Initiative (RRI). The committee is working on contributions such as grasping the OSS community and its development and standardization trends, OSS quality evaluation, functional and documentation improvement, which are essential for companies to use OSS. This paper introduces the activity status in 2021.

In recent years, robotics middleware such as OpenRTM-aist and ROS has been used for the development of robot systems, and efficient development has become possible by reusing modules. In order to realize the practical use of robot systems, not only development efficiency but also security measures during operation are important. This paper describes the overview of robotics middleware extensions of the secure communication and HTTP communication, and the system that can be constructed.

This paper describes a rapid development of a support robot system for a librarian service desk using a tabletop communication robot. It is important issues that the remote operation of the librarian service desk for the convenience of the visitors and the efficiency of the librarians. We developed a robot system that interacts with high accuracy and low latency in a short period of time using a cloud service for speech recognition and synthesis functions. We conducted participant experiments for tasks for the librarian service desk to evaluate the developed robot system. Experimental results showed the feasibility of the developed systems.

In this paper, we propose a projection mapping system for stage production of puppet theater. We solve missing area problems of the projection-mapped stage due to occluded projection by performer such as puppets. Our system consists of multiple projectors and a RGB-D camera, the projectors work complementary each other, avoiding to project on the performers and complement the missing are such as shadow.

Automation of toilet cleaning is demanded in convenience stores. So far, an elevating toilet cleaning system has been developed however the elevating motion was manually controlled. In this study, we automated the motion by providing the user’s status as input to the cleaning system, then assess the developed system. The results showed that the success rate of correct operation was 45.7%, and the ratio that the operation time was acceptable to the user when it worked correctly was 65.6%.

Service robots are expected to move in entire buildings for several usages. And a kind of arbitrator systems for traffic management between different variety workers are important to maintain safety in the building. This paper discusses about an information management structure for multi mobility management called indoor dynamic map and introduces a system architecture for mobile robot service. It also explains demonstrations about book delivery service in Katsura Library, Kyoto University. Experiments have been demonstrated since Feb. 2022.

This paper focusses on a semi-autonomous teleoperation system for reducing heavy burden of an operator. Our proposed path planning system is based on Growing Neural Gas (GNG) for learning a topological structure from a metric map built by SLAM for applying the path planning to an unknown environment. At first, this paper explains our proposed semi-autonomous teleoperation system using a touch interface and learning algorithm of the topological map. Next, this paper explains the path planning that utilizes the topological map built by GNG. Finally, for verifying the effectiveness of our proposed teleoperation system, this paper conducted on experimental results to compare the beginner and expert operators.