This paper examines the behavior of a group of pedestrians in avoiding a roadside obstacle. We believe that invisible forces between pedestrians are responsible for the fact that multiple pedestrians can walk at a certain distance without colliding with each other.

In order for autonomous personal mobility(PM) and pedestrians to coexist in a pedestrian space, it is important for them to be able to avoid and pass each other smoothly. Studies suggest that it might be possible for PM and pedestrians to pass each other within a short time without compromising comfort by understanding how pedestrians react to the behavior of PM and use it in the autonomous locomotion. Therefore, in this study, the avoidance behavioral characteristics of pedestrians was investigated by experiments, and a path planning method that enables smooth passing was developed based on these characteristics. Result shows that by using the characteristics of pedestrians’ cognition against PM it was able to develop a smooth autonomous locomotion.

The objective of this research is to develop autonomous mobility adaptable to the pedestrian environment with indistinct road boundaries using a recommended traversable area detection system. In previous research, we proposed a learning method for weakly-supervised semantic segmentation, which was effective in improving the performance of detecting recommended traversable areas. In this paper, we developed an autonomous navigation system based on a driving recommendation map using a recommended traversable area detection system. By setting appropriate weights to consider recommended traversable areas, generating paths in the direction of right and left turns while suppressing path generation to non-recommended areas was achieved. Also, by adding a new class of driving trajectory and sidewalk area, the system can selectively enter sidewalks in scenes where physical features and edges cannot be taken. Finally, the usefulness of the developed system in a real environment with indistinct road boundaries was verified.

What are the characteristics of small animals that many people find to be dear, and what are the factors that make them feel and heal? We discovered that some motions, shapes, and tactile sensations had a great influence, and developed it as a pet robot "Moflin®".

This paper presents a robot communication system for promoting driver's intrinsic motivation to drive safely. We developed a driving simulator with a communication robot used as a navigator for scoring the level of safety driving and providing some advice. Moreover, we proposed a method of human-robot interaction to prevent a lack of self-efficacy to achieve safety driving, based on the theory of self-serving bias. Furthermore, we conducted an experiment and questionnaire investigation to verify the effect of the proposed approach on driver's impression of the robot and discuss the applicability for motivating safety driving. From the experimental results, we found that the effect on driver's self-efficacy varies with not only the scoring result but also the frequency of car use in daily life and driving experience.

This paper presents a multi-robot communication system for guiding customers and providing information about products and deals in a shop. The purpose of the multi-robot communication is to offer a sense of belonging between the robots and a customer. However, even if the robots are collaborating with each other, the robots can be ignored by customers when the robots try to communicate with only their utterance. In the experiment that we conducted in a store, the robots set in front of the store can get the attention and sense of togetherness of the customers by offering to shake hands with them. In this study, we conducted the questionnaire investigation and monitoring to verify the effect of the proposed approach on customer impressions of the guide robots.

The purpose of this study is to realize a medical interview training system in which an android robot is employed as a simulated patient to provide consistent training and to provide a quantitative evaluation to medical students. In this study, we first analyzed the medical interview by medical doctors. Then we implemented a simple algorithm to classify interviewer's utterance into question and non-question. In addition, we developed a detection algorithm of interviewer’s nod to evaluate interviewer’s non-verbal behavior. The result indicates that our conversation system can automatically conduct medical interview and evaluate interviewer’s nod.

In recent years, the demand for service robots has been increasing, and communication robots that can interact with humans in various situations such as support for the elderly and improvement of QOL are attracting attention. The research area where robots try to interact with humans according to some scenarios is called "robot theater". heater (a play) is performed for not only entertainment but also education, support for elderly people or people with disabilities, etc. due to its characteristics such as communication, so theater constructed by robots has significance not only in robot science but also in various academic fields. The theater which is performed unplanned or unscripted is called "improv", and it is used in various fields and purposes because the performers have to play while paying attention to other performers and situation. Therefore, incorporating the idea of improv into a robot theater lead to construct a robot theater that has interactivity such as communication between humans and stage theater. In this paper, we propose a robot theater system in which robots act impromptu by changing scenarios, taking into consideration the reaction and state of the audience according to the scene.

Guidance of people is an important task for mobile robots that provide services. In this paper, we propose a method for a ground-projecting robot to guide a person based on the property that a person moves to a position where the projected image is easily visible. The proposed method rotates the projected image in guiding, guides the person's direction to another route direction and a guidance object, and realizes natural guidance. From the preliminary experimental results, it was shown that the direction of the subjects changed with the rotation of the projected image.

Conventional conversational system technologies use speech to estimate the speaker’s intentions. However, the problem is that it is not possible to estimate the other party’s intention from various nonverbal information as in human conversation. In this paper, to use a neural network to learn about the relationship between tone of voice and probability word. Using this neural network, we estimated the speaker’s intention by analyzing the verbal range from tone. We proposed a Cue bar task to check the performance of this conversation system. We defined an evaluation function based on the total amount of movement. The results showed that the median of the evaluation function was less than 0.3. In other words, although the accuracy was low, it was confirmed that the speaker’s intention could be estimated. As a problem, it was suggested that the task needs to take into account the information transfer between humans and robots.

In this paper, a panda-type emotion expression robot is developed, which can react to user’s interactions considering its own emotion status. The user's emotion is inferred from his/her facial expression, voice and pressure force when touching with the robot. With the emotion generation system, the emotion of the robot is estimated by combing the user’s emotion and its own emotion status. In this way, the robot can basically react to the user, adjusting its own emotion according to the users’ interactions, as well as showing its own personality. The effectiveness of the proposed method is proven by reacting experiments using a panda-type robot.

This study presents a social robot that expresses emotion by controlling contact reaction force and sound in response to the touch from users. For social robots interact and communicate with human users, it is crucial to investigate the relationship between the stimulus provided by the robot and the how users interpret them as emotional behaviors. The developed cube-type robot is able to vary perceived stiffness by controlling an electromagnetic brake. The experiment demonstrates the robot’s potential to affect emotional valence with simple audio stimuli.

Some communication cases are significant barriers for individuals with autism spectrum disorder (ASD) because these individuals lack good nonverbal communication skills. Therefore, some studies discuss communication skill training using humanoid robots. In this paper, we propose the automatic dialog system for ASD using android robot. The system utilizes MMDAgent which is a toolkit for building voice interaction systems to realize communication skill training for individuals with ASD.

This paper describes sonic communication using an electrostatic driver. It can send and receive audible and inaudible sounds. This device is intended to communicate between humans and robots, or between robots, using sound waves. To verify this device, we transmitted and received sound waves using two identical devices. The frequencies are 1kHz and 25kHz. as a result of this experiment, it was confirmed that sound wave communication in the audible range and non-audible range was possible between devices.

Patients with severe gait disorders requires long-distance, multi-step gait training with appropriate movement of the center of gravity. However, the conventional gait training device with a load reduction function doesn’t have a function of controlling the trunk movement. In this paper, we described the results and prospects of an attempt to control the trunk and adjust the load for each leg in gait training with a load reduction function using wires. In the future, we will carry out gait simulations that reflect joint reaction force and body posture and movement, and build a system that estimates the posture and applied force of the body in real time. Finally, we plan to produce a gait training device with a trunk control function and conduct experiments with the subjects.

This paper proposed a self-standing-up training support robot that supports the patient's movements instead of the therapist for hemiplegic stroke patients. The robot consists of two passive robots with ER fluid brakes. We proposed a waist guiding control method that allows the patient to stand up and sit down. Two robots attached to the patient’s waist guide the patient’s waist in the left-right, roll, and yaw directions as the therapist. Moreover, the effectiveness of the proposed control method was evaluated from the view point of the maximum constraint force required at the end-effector and the virtual spring characteristics of the end-effector.

In developing a therapeutic device, the previous device could only apply one frequency to either the lumbar region or the lower leg. In the future, it is expected that individual frequencies will be applied to both the lumbar region and lower legs. Therefore, there is a demand for a device that can output different frequencies (left and right independent frequencies) from the two vibrators. Therefore, we have developed a new compact device. There is a problem that the amplitude of the oscillator varies depending on the frequency characteristics of the vibrator and individual differences. Therefore, I used voice editing software to create a stereo sine wave signal for vibration stimulation and adjusted the volume and left / right volume ratio to reduce variations. It was conformed from the constant amplitude test that the maximum error rate of 14.6% before the adjustment reduced that of 1.7% after the adjustment.

We have developed a finger extensor facilitation training device “iPARKO”, which imitates the manual treatment method by a therapist. Many hemiplegic stroke patients are unable to open a paralyzed hand. Therefore, they need to strengthen extensor muscle to open their hand by himself or herself. In this paper the effectiveness of iPARKO on healthy persons was verified by two comparative experiments. The extensor muscle activity using iPARKO was approximately 40 % greater than that using our previous device PARKO. Moreover, the extensor muscle activity using iPARKO was approximately 80 % of that at maximum voluntary contraction for opening a hand.

In home-visit rehabilitation, a quantitative and easy-to-use system to evaluate motor function of the elderly is in demand. The authors have developed a portable-type motor function evaluation system as an application on personal computers. The application measured skeleton data from user’s motion with a compact RGB-D sensor and the designated middleware. We decided design specifications of the application from the viewpoint of the physical therapy. The results of data processing were displayed on the computer screen. In this research, we implemented a function to display a graph that quantitatively visualizes the joint angle and to synchronize the original video and the horizontally inverted video to confirm difference of right and left.

Patients with stroke may have hemiplegia, which reduces their ability to perform activities of daily living and reduces their quality of life compared to before the onset. To improve motor function of them, it is important to develop an assistive device that support rehabilitation as physical therapists do in clinical environment. Here, we developed a chair-type rehabilitation device that emulates intervention of physical therapist. The devices estimate movement of the user through pressure sensor and stimulate buttocks of the user. We firstly clarified the relationship between center of pressure of the seat and pinching force by physical therapists. This is implemented on the assistive device and the measurement experiment was performed. The results showed that the assistive device could successfully stimulated the user’s buttocks. Future work will be implementation of other intervention method of physical therapists.

Training time is shorter because the number of physiotherapists (PT) is not enough compared to the number of patients. This will reduce the effectiveness of walk-training. Patients do walk-training at home rather than on a walker. In this study, we propose a walking motion measurement method using an RGB-D camera when a human walk with a walker, which is for the purpose of enhancing the rehabilitation effect. An RGB-D camera is attached to a care-walker. The RGBD camera measures the position of the feet to estimate walking movement. By using the information, several walking motion parameters can be estimated such as stride length, stride distance, walking angle, foot height, number of steps, walking rate, walking speed, and walking distance. The estimation algorithm was verified through experiments.

For motor neuron diseases whose main symptom is muscle weakness, neuro-rehabilitation technology is expected to provide efficient improvement in brain functions through spontaneous exercise. In this paper, we develop a system that simultaneously presents physical hand movement support by a force-assist glove powered by pneumatic artificial muscles (PGMs) triggered by EMG signals and visual feedback by displaying virtual fingers. The experiments were conducted under the condition of changing the presence or absence of actual finger motion support while the virtual hand motion is larger than actual hand motion, and the effect of the difference in the motion by the simultaneous feedback of vision and force on the feeling of motion synchronization is discussed.

In gait rehabilitation, assistive devices for paralyzed patients have been studied in various approaches. However, patients with complete paralysis of the lower limbs cannot safely walk or hold a standing position using conventional orthoses. In addition, most lower limb orthoses do not consider the balance between the upper limb and the whole body. Therefore, in this study, we aim to realize an assist mechanism that enables patients with complete paralysis of the lower limbs to walk and stand safely and describe the design and prototyping at the current stage of realization.

In this study, we developed a system that automatically optimizes two control parameters, the contraction start time and the contraction time of the pneumatic artificial muscle, by measuring the effect of the assist while walking assist is performed. Those parameters were set based on the heel contact, and the system was created with the objective function of maximizing the thigh swing angle during walking. As a result of preliminary experiments of walking assist with this system, the control parameters converged to the optimal or next optimal parameters for the user through optimization search.

Hemiplegia is a severe paralysis on half of the body caused by cerebrovascular disease. Damaged hand motor function takes long time to recover compared to other places. Therefore, self-rehabilitation at home is important after a patient has finished rehabilitation by a therapist at hospital and returned home. However, the evaluation of hand motor function is difficult since it involves not only a simple control of a finger but also dexterous control with separated finger motions. To this end, we have developed a small and light automated finger motor function evaluation device, which has five load cells on each finger and is designed for natural pinching and grasping postures. The device can measure the degree of normality of a finger function by a combination of finger pressures. By using it, a patient can know the condition of the paralysis and daily achievements of self-rehabilitation at home. We measured grasp and pinch actions. The measured finger pressure reflects difference between these two motions.

In home rehabilitation support device, appropriate biofeedback is an important issue since there is no support by a medical specialist. In our finger rehabilitation support device, the machine will show the degree of unwanted finger movement in real time. Many similar devices show the degree of finger pressures as a bar graph in a separate display. However, this method requires the patient to gaze at the display while being rehabilitated, which reduces the effectiveness of visual feedback. In this study, we propose a method of presenting involuntary movements by changing colors at the fingertips of the device so that the patient can recognize them while gazing at the hand during rehabilitation.

We have been developing an articulated mobile robot that mimics the peristaltic movement of an earthworm. This robot is suitable for inside the rubble and pipe inspection because of its high performance in running on uneven terrain and its mechanism for moving in a narrow space. In this study, we developed a control system using Arduino and designed a robot body section to improve the moving speed, aiming to have the scalability to add sensors and actuators easily.

In this paper, a water suction unit which removes residual water was designed and developed to obtain clear images of the inside of sewer pipes from the camera. When PEW-RO V, which was developed in a previous study, inspects sewer pipes, it cannot inspect inside of sewer pipes taken by its camera in residual water. Residual water removal mechanism consists of the water suction pump and the water suction unit. In the experiment, the location of the pump was determined and PEW-RO V with the water suction unit successfully drained the residual water to a level where the camera could acquire images.

To mix and convey solid-liquid mixtures continuously, a peristaltic continuous mixing conveyor has been developed previously. This device imitates the movement of intestines by inflating a rubber tube using air pressure. We estimated the mixing degree of the contents via machine learning using data from pneumatic sensors, however, the accuracy was low. Therefore, in this study, to improve the sensing system, we installed an accelerometer that could measure the behavior of the rubber tube without any contact with the contents. The sensor angle was obtained from the accelerometer values for each input. We obtained different sensor angles with the rigid bodies and the powder-liquid mixtures. The results suggested that type of the contents can be determined by using the accelerometer.

Currently, soft actuators are being actively researched. Genellary, soft actuators are more difficult to control than ordinary actuators because of their flexibility. Therefore, there is a need for a sensor that can be unified with the soft actuator and measure the change in shape of large deformations without interfering with the deformation of the soft actuator. In this study, we propose a strain sensor that can measure the displacement of a soft actuator based on a conductive paste with silver as a filler. This paper describes the investigation of the basic properties of conductive pastes using silver filler and the measurement equipment.

The mixing and transporting process of a solid–liquid mixture and a high-viscosity fluid is separate batch processes, leading to an increase in maintenance costs and disposal costs. To achieve continuous mixing and transporting process, the authors have developed a pneumatic peristaltic mixing conveyor based on the movement of transporting bolus in the intestinal tract. The proposed conveyer is possible to mix and transport with one device continuously, however, it has not been studied to mix and transport at the same time like the actual intestine does. In this paper, we proposed a congruence which is able to continuously switch between segmental and peristaltic movement in order to achieve a continuous mixing and transporting.

In this study, we focus on a peristaltic motion that is locomotive pattern of earthworms. To realize effective motion of the peristaltic mobile robot, motion patterns are generated based on the dynamical model and the particle swarm optimization algorithm, one of the meta-heuristic optimization. The optimized results are investigated by comparing to simple periodic patterns through numerical simulations and experiments. The motion control with switching of input patterns in response to changes in the friction environment is demonstrated to improve the travel distance.

We are developing an actuator-less standing support machine for the future, when nursing care systems are under pressure. In order to design this device with safety in mind, we performed static analysis in SOLIDWORKS to design the strength. By referring to the human size data in the ergonomics database, we estimated the maximum value of the load by the user, and set the conditions such as contact and constraint according to the use situation. We report on the results of this study.

We proposed a suction device with a valve opened by contact and combined a commercially available suction pad. This device has the suction pad, silicon rubber valve element, and adapter. The valve eliminate has pin, 3 plate springs, and ring support. Since the pin was supported by the plate springs, the moving direction of pin was vertical direction when the force direction was vertical. All of them was formed by silicone rubber. The device adhered the plate when it was contacted on the plate. The pressure difference is under 0.1 kPa when the valve was closed. However, the peeling time was 1 s, because of the small valve opening and the large inner volume.

This paper proposes a caterpillar-inspired finger for a robot hand that produces the peristaltic motion consisting of bending and contraction to realize in-hand manipulation. One of the motions that the traditional grippers made of hard/rigid materials (e.g., parallel grippers) cannot do, but human-fingers can, is in-hand manipulation. We believe that enabling this motion with a soft gripper is one of the key elements to achieve dexterous manipulation with a simple structure. Also, it will contribute significantly to the design and control of robot hands in the future. In this research, to demonstrate the effectiveness of applying peristaltic motion to a robot hand’s finger, we conducted a dynamics simulation and confirmed that motion can carry an object in a certain direction.

In recent years, robots are expected to play an important role in factory automation. Robots working in factories adapt to various shapes by changing end-effectors. However, in the modern era of advanced high-mix low-volume production. it is desirable to reduce time and costs on replacing the end-effectors. Therefore, soft robot hand need to grip targets of various shapes quickly and accurately. However, many of parameters are inconsistent and insufficient in previous gripper experiments. In this study, we measure the grasping force of the Universal Vacuum Gripper by varying the parameters and evaluate the performance and verify the necessary parameters.

With the evolution of robot technology, robot hands capable of grasping objects of various shapes and sizes are required for the automation of production processes. In a previous study, a soft robot gripper based on the jamming transition phenomenon was proposed. However, it was difficult to grasp fragile objects and small diameter objects. In this study, we propose a new grasping method in a soft gripper using vacuum adhesion and jamming transition, where the pressing motion is substituted by vacuum adsorption. We investigated the relationship between the gripping performance and the increase in the contact area between the gripper and the grasped object.

This paper describes the force diode mechanism which realize the reverse input prevention. Based on the basic principle we considered, the mechanical prototype model has been built and basic experiments with this prototype have been done so that the fundamental function and effect of the basic idea of the force diode mechanism.

This paper presents a robotic system for bulk picking of food materials. The system consists of a robotic manipulator, a 3D camera, and a parallel shell gripper. A 3D template matching was used to recognize the individual food material from a bulk. The parallel shell gripper constructed by rigid shell and soft chamber is pneumatically driven. The thin structure of the shell gripper facilitates successful grasping of individual food material from a bulk. The control of the robotic system is based on robot operating system (ROS). Experiments results on several food materials validated that the proposed robotic system can successfully handle food material from a bulk.

This study proposes a food classification database for robotic handling. The proposed food classification database classifies a wide variety of food products according to their conditions and characteristics, and associates them with various soft robotic hands or grippers that are commercially available and under research and development. By building a database using Microsoft Access and preparing multiple forms such as input and search, it is easy to share the food classification database with other researchers or users. The goal of this study is to build a food classification database for soft robotic hands and collect data to investigate trends in foods that soft robotic hands can grasp. This can be determined from the percentage of classification within the associated food items.

The market size of prepared foods exceeds 10 trillion yen and is increasing year by year, and more than millions of lunch boxes are manufactured and sold daily. food manipulation is difficult to automate because the foodstuffs are often soft and the shapes of the foodstuffs vary widely, and the current situation is that this arrangement work is mostly done manually. In the work of serving lunch boxes and the like, it is indispensable to sort foods in a quantitative manner, and in the case of serving with a robot, it is considered that an efficient motion plan and gripper performance are required for serving in a fixed quantity. Therefore, in this research, we propose tactile sensor for grabbing foods.

In this paper, double-layered Universal Vacuum Gripper (in short, d-UVG) is proposed for grasping various objects. The d-UVG is a vacuum suction cup based on the jamming transition phenomenon with two functional layers. First layer is used for grasping smaller object than the gripper based on vacuum forming method, and also protecting the membrane's damage from vacuum force that is one of the problems in the conventional UVG. Second layer is used for air-sealing for adhering uneven surface. We examined the d-UVG on several test pieces for evaluating its gripping performances and verify d-UVG’s concept and effectivities.

Today, collaborative work using multiple robots are used motion strategies that designed according to the environment. However,there is a problem in dealing with unexpected situations. In this research, We propose a motion design method that introduces the individuality found in human social cognition by chaos control.In this paper, The combined Van der Pol equation was adopted as chaos control, and a face-to-face traffic simulation was constructed to examine the effectiveness of the number of collisions with other individuals and the number of collisions with walls in terms of the number of individuals. As a result,we confirmed the dispersion of individuality and confirmed the decrease in the number of collisions between individuals.

This paper reports some results of network connectivity control experiments using a multi-robot system which we developed. We detail the developed hardware and software, results of measurement experiments, and connectivity preservation control experiments.

Deep Neuroevolution is a method of optimizing a Deep Neural Network (DNN) using evolutionary computation. By applying this approach for designing the robot controller, the robotic swarm is able to generate collective behavior from raw image inputs. However, we need to explore how to set up the structure of DNN and how theses settings affect the performance. In this paper, we applied DCNN and DCRNN to the controller of a robotic swarm. DCNN is a DNN that can automatically extract image features. DCRNN is a combination of DCNN and Long Short-Terms Memory network. We discussed the performance of DCNN and DCRNN in a two-target area navigation task under the environments with and without an obstacle. The results showed that DCRNN recorded better performance than DCNN in two both environments.

This study tackles the task for swarm robotics where robots explore the environment to detect a target. When a robot detects a target, the robot must be connected with a base station via intermediate relay robots for wireless communication. In our previous results, we confirmed that Lévy walk outperformed the usual random walk for exploration strategy in real robot experiments using the small differential wheeled robots. This paper investigated the performance of a new Lévy walk generator, which is recently proposed in biology, for the target exploration problem on robotics through a series of real experiments using the crawler robots. The results suggest that the new Lévy walk, based on nonlinear dynamics, outperformed the one based on probability for search efficiency.

This paper focuses on Topology and Weight Evolving Artificial Neural Network (TWEANN) approaches for designing controllers of a robotic swarm. TWEANN approaches are expected to design not only synaptic weights but also the appropriate network topology without the intervention of the designers. Mutation-Based Evolving Artificial Neural Network(MBEANN) is a TWEANN algorithm that only use mutations to evolve neural networks. In this paper, we applied MBEANN to design the controller for a robotic swarm. For comparison with the MBEANN approach, we used the NeuroEvolution of Augmenting Topologies (NEAT), which is a widely used TWEANN algorithm.The performance and the topology of robot controllers are compared in collective foraging tasks. The results show that MBEANN could perform as well as NEAT with smaller network topologies.

Pneumatic drive systems are widely used in assistance devices and flexible drive robots owing to their characteristics of light weight, high output, and flexibility. However, those devices cannot perform instantaneous operation because those devices have low responsiveness. Then we focus on the physical properties of flammable dimethyl ether (DME), and we considered using combustion of DME to improve the responsiveness of pneumatic drive systems. We developed artificial muscle which is driven by burning DME inside the device. In this paper, we conducted the experiment to investigate the response time and displacement of artificial muscle when DME burned in the device.

Legged robots have an advantage for moving on rough terrain. Especially, quadruped robots have high posture stability. However, actuators of many legged robots are actuated by an electromagnetic motor and a speed reducer, and are vulnerable to impact. On the other hand, hydraulic actuators have a simple structure and a small number of parts, thus they are strong against impact and can generate large force. For this reason, hydraulic actuators are used in robot leg mechanisms in recent years. However, there are few development examples of leg mechanisms that achieve both high output and lightweight at the same time. In this study, we developed a hydraulic quadruped robot that has a lightweight leg mechanism and is a self-contained system. A walking experiment of the robot showed that it can walk independently.

In this paper, we propose a novel pneumatic-driven variable stiffness mechanism capable of shape-locking. Various methods had been proposed to change the stiffness. However, there is no simple methodology to realize both controllable elasticity and shape locking. In this research, we present a continuum robot using flat tubes pathed through along the body inside the slits. The slit has multiple needle-like structures, which makes the tube grip with large friction when pressurized. This makes it possible to retain the robot in an arbitrary shape. First, we show the principle of the shape-locking mechanism and fabricate a prototype of the proposed mechanism. A basic test is conducted to check the performance of the mechanism.