Keywords: Mechanism Design, Mechatronics Systems, Intelligent Transportation Systems
Abstract: For stair climbing by a mobile robot, we discussed a 1-DOF mechanism that allows a rotating axis to be lifted with less torque.　A prototype quadrupedal robot equipped involute-curve-shaped mechanism which is obtained from that discussion was developed and a control method for stair climbing was proposed. Furthermore, it was confirmed by experiments that the robot autonomously climbs stairs.

Keywords: Mechatronics Systems, System Simulation, Control Theory and Technology
Abstract: Elastic actuators find a growing interest in many research areas ranging from rehabilitation robotics to industrial robotics. Elastic actuators are actuators with elements of compliance, such as springs connected in different configurations between the actuator and the load. In particular, Rigid Parallel Series Elastic Actuators (RPSEA) is an emerging design that, in addition to the advantages provided by parallel elastic actuators, provides the ability to actively control the storage and release of spring energy. Owing to the spring element in these actuators, they often face the problem of oscillations. In this study, we propose an energy-based control strategy for vibration suppression by using only one motor of the RPSEA. The analytical findings are substantiated with simulations of the mechanism subjected to an impulse disturbance.

Keywords: Mechatronics Systems, Systems for Field Applications, Control Theory and Technology
Abstract: This paper discusses the combination of model-based feedforward and feedback controller for the hybrid wheel-legged robot. Uncertainties in the modeling parameters such as friction coefficients, inertia matrices, and position of CoM can bring negative effects in the performance of model-based motion controllers. The ability of a control system to compensate for unforeseen circumstances is not only utilitarian but essential if one wishes to bring the robotic systems into the real world. To develop such a robust motion controller for the wheeled biped robot, we present the design and implementation of a nonlinear feedforward controller together with linearized feedback LQR to control the robot on a flat surface. This study investigates the controller performance: feedback, feedforward, and the combined in simulations and also with the real wheel-legged robot. Furthermore, the Extended Kalman Filter (EKF) is implemented to estimate the system states and reduce the sensor noise of the real robot.

Keywords: Mechatronics Systems
Abstract: Servospheres are robotized observation systems that produce unbounded fields for behavior observation of wandering animals. Each servosphere is a vision-based control system that mainly consists of a camera, a sphere, and motorized omniwheels. As an approach to achieving high observation performance of servospheres, this paper first proposes a fast control technique under installation of a high-speed camera. No special devices other than a high-speed camera are required. The proposed technique improves the time resolution without changes in the spatial resolution. This paper also investigates performance of a servosphere with changes in the camera frame rate and the control gains. It is demonstrated that the regulation errors are significantly improved by the proposed technique at the high frame rate with high gain feedback. High gain feedback at the normal frame rate sometimes produces undesired oscillations, and it does not decrease regulation errors. The results imply that the high-speed servosphere improves observation performance, while normal servospheres might lead to an erroneous conclusion on animal behavior.

Keywords: Mechatronics Systems, Hardware Platform
Abstract: Recently, there are many studies about hyper-human technology like high speed grasping. However, almost such systems focus on just grasping. However, to improve the feasibility to apply the system to real application, both of grasping and releasing functions are required. In this paper, we proposed prototype of high-speed gripper, which have the ability both of grasping and releasing. Through the preliminary experiment, the effectiveness of proposed mechanism is evaluated.

Keywords: Mechanism Design, Factory Automation, Automation Systems
Abstract: For realizing a versatile, flexible robotic assembly system, the development of a versatile robotic hand with a minimally necessary mechanism which can grasp all the target parts and perform jig-less assembly of them is desired. Considering a rotational chuck-type hand with three parallel stick fingers as one possible solution, this paper presents a method to design an optimal mechanism of the hand through grasp optimization under a required external wrench set, and a hand mechanism for the assembly of target parts is actually designed with the proposed method.

Keywords: Systems for Field Applications, Mechanism Design, Automation Systems
Abstract: This paper introduces a structure and harvesting motion for the suction cutting device of a tomato harvesting robot, and reports harvesting experiments conducted in a tomato greenhouse. The suction cutting device comprises a suction part and a cutting part. The suction part separates the target fruit from a tomato cluster and the cutting part cuts the peduncle of the target fruit. A photoresistor in the cutting part assesses whether or not the target fruit is harvestable, and the cutting motion is performed only when the fruit is assessed as harvestable. The harvesting experiments were conducted in a tomato greenhouse to evaluate the suction cutting device. In this experiments, 50 tomato clusters were randomly selected as the harvesting objects, and there were 203 tomato fruits (including immature fruits). Out of the 203 tomato fruits, 114 tomato fruits were mature and within the robot workspace. Out of the 114 tomato fruits, 105 tomato fruits were recognized as target fruits by the harvesting robot. Out of these 105 tomato fruits, 65 tomato fruits were assessed as harvestable and 55 were successfully harvested (the harvesting success rate was 85%). Based on the results of the harvesting experiments, this study clarified the issues of the suction cutting device, classified the fruits according to whether they were easy or difficult to harvest, and evaluated the fruit characteristics qualitatively.

Keywords: Multi-Modal Perception, Sensor Fusion, Vision Systems
Abstract: This paper addresses a robust human trajectory tracking method through the data integration of 2D laser scanner and 3D camera. Mapping the deep learning-based 3D camera human detection onto the point cloud of the depth information to build up the 3D bounding box-represented human and using the state-of-the-art 2D laser-based leg detection are the main data streams of the human tracking system. The human-oriented global nearest neighbour (HOGNN) data association, inspired from the Hall's proxemics, was developed to improve both the 3D camera-based and 2D laser-based human detection techniques. The dual Kalman filters are employed to track the human trajectory in parallel. The integration of the 3D camera-based and 2D laser-based human tracking is the key function of the system providing the real-time feedback for both the HOGNN to reduce false-positives of the camera-based and laser-based human detection and the Kalman filter to enhance the quality of the human trajectory tracking under uncertain environmental conditions. We implemented the sensor integration on ROS and validated it through real-world experiments.

Keywords: Hardware Platform, Vision Systems, Motion and Path Planning
Abstract: The system aims to develop an algorithm, to be implemented in an open-source platform, that facilitate the navigation, and early door/window detection process, for people with visual difficulties in closed environments. The proposed algorithm relies in a three stage image process consisting of the removal of all unwanted information through grayscale conversion and edge detection (image pre-processing), followed by the second stage formed by the search of lines or patterns on the detected edges, with the help of a first order Hough Transform, in order to locate a region where the majority of these lines intersect with each other, or vanishing point, which is the pivot point of the entire algorithm. Finally, the last stage consist on the estimation of the main regions of the aisle (walls, floor and ceiling) in order to perform an early detection of special regions in the walls, or candidates, that could lead to the detection of doors and windows. The main idea is that, after the image processing, the algorithm will return instructions to the subject, through a physical medium to make the corrections on his trajectory. It was found that the estimation of the vanishing point through the use of Hough transformation and the second-order edge detection method using open source software tools is a viable method for navigation in closed environments and even its possible to use the same image processing tools to estimate regions that could be easily consider as the main regions of the aisle. It’s been demonstrated that perform the early detection of candidates for doors is also possible using the same techniques.

Keywords: Systems for Service/Assistive Applications, Systems for Field Applications, Welfare systems
Abstract: Japan has been a super-aged society in the recent years; thus, there is a need for mobility support for elderly people with difficulty in moving and walking. We aim to provide a mobility support using a mobile robot considering movement from a facility to home through indoor and outdoor environments. A seamless positioning system is necessary for realizing such a community. This study proposes a seamless positioning system that can obtain position data of both indoor and outdoor environments. We develop a system combining ultrasonic beacons and a global positioning system receiver to obtain the seamless location position. We install the proposed system and verify the positioning system by a basic experiment.

Keywords: Multi-Robot Systems, Sensor Fusion
Abstract: Mobile robots are gradually entering our human spaces. Apart from being able to accurately localize themselves, in populated environments robots also need to detect humans and recognize their characteristics. In this work we focused on using 3D LiDARs for sensing, and in particular on the question whether they can be used to classify characteristics of people around the robot. Moreover, since using sensors from multiple robots is expected to give more accurate recognition, we com-pared several ways how to combine multiple 3D LiDARs. We evaluated these methods using simulator data as well as actual real-world data. For the real-world data we created a novel dataset taken with multiple robots equipped with 3D LiDARs in a shopping center, which also includes manually labeled characteristics of the detected pedestrians. The results show that combining several 3D LiDARs makes the recognition more accurate. However, we were not able to achieve satisfactory accuracy with real-world data.

Keywords: Decision Making Systems, System Simulation, Software, Middleware and Programming Environments
Abstract: In this paper, we evaluate the performance of the automated mission planning system which we developed for ocean observation of microsatellite RISESAT. RISESAT is equipped with multiple scientific instruments including Ocean Observation Camera (OOC). OOC is a wide field of view camera system and sweeps the ocean surface to evaluate the amount of Colored Dissolved Organic Matter (CDOM), which influences the carbon cycle in this planet, which is closely related to climate change. Constraint conditions exist mainly in terms of cloud coverage and the elevation angle of the satellite at a targeted point because OOC observation is conducted in the range of the visible light spectrum. Therefore, we developed an automatic mission planning system based on cloud forecasts for efficient ocean observation. In this system, we use Open Weather Map (OWM) to create cloud forecasts, and Two Line Element (TLE) and Simplified General Perturbations Satellite Orbit Model (SGP4) for orbit prediction of the satellite. Then, we propose an algorithm to output an optimised mission plan maximising the total profit. Furthermore, the performance of the system was evaluated by actual images taken by OOC and we confirmed that this system captured targeted regions properly.

Keywords: Control Theory and Technology, Mechatronics Systems, Formal Methods in System Integration
Abstract: This paper describes procedures for the design, implementation and in-orbit verification of an advanced Attitude and Orbit Control System (AOCS) that can be applied to micro-satellites. Functional requirements imposed on AOCS are becoming severer as the mission becomes more complicated and challenging even though onboard resources are limited in micro-satellites. In addition, many verification steps must be taken for the system design, implementation, and in-orbit demonstration. In this study, an advanced AOCS design that is optimized for the mission of the artificial meteor demonstration satellite "ALE-2" is proposed. This paper presents three AOCS features that are specialized to ALE-2 but applicable to other micro-satellites: sensor calibration technique, attitude and gyroscope bias estimation using extended Kalman filter, and orbit control by small propulsion system. These functions were implemented using a hardware-in-the-loop simulator environment, allowing for quick and efficient ground evaluation. In addition, in-orbit demonstration of the proposed AOCS was performed after the launch of ALE-2. Through these verification process, it was confirmed that the AOCS functions required for the mission of ALE-2 were properly implemented and worked in orbit.

Keywords: Decision Making Systems, Control Theory and Technology
Abstract: Blind system order and parameters identification are the challenging problems in robotics and automation engineering. In this paper, a blind system order and parameters identification algorithm based on a cascaded genetic algorithm for linear time-invariant systems is proposed. It estimates both the order and parameters of the system. The main features of this work include the adaptive length of the chromosome for ascending system order identification, no limitation of the order of the system to approximate, robust to the system noise, and capable of providing low order approximation to the system. In addition, no overestimation of the system order will be occurred by the proposed method. The experimental result reveals that the proposed method approximates the order and parameters of the black box system correctly and robust to the system noise.

Keywords: Control Theory and Technology
Abstract: UAV applications are being developed for various applications and are being marketed to provide solutions for agricultural, industrial, and even entertainment applications. In order to create these applications, it is vital that the UAV has good stability, and most control system of these UAVs still makes use of PID. Tuning of PID can be done by trial and error or by using the Ziegler-Nichols method. Another approach is to use a Genetic Algorithm that will find the optimal PID values without having to go through each step of trial and error or the Ziegler-Nichols method. Using the Genetic Algorithm approach optimizes the PID coefficients to produce an acceptable response to the system.

Keywords: Control Theory and Technology
Abstract: Position control of UAV becomes important when it involves autonomous application such as agricultural applications. It allows the activity to be accurately and efficiently achieved as compared to manual control. In this paper, 2D position control of a quadcopter UAV is explored using fuzzy logic control. Fuzzy logic is studied as it is known for its human-like implementation as compared to other control systems. The fuzzy controllers used have three (3) membership functions each for the two (2) inputs, and one (1) output. It uses two (2) open-trapezoids, and one (1) triangle membership functions. Three (3) fuzzy controllers were used to control the z, y, and phi values of the 2D model of the quadcopter UAV.

Keywords: Motion and Path Planning, Control Theory and Technology, System Simulation
Abstract: Among the existing control schemes, Fuzzy Logic (FL) is advantageous in offering robustness to uncertainty and noise in sensor readings. However, the performance of FL-based navigation strategies is well-known to be bounded by local-minima risks in polygonal environments with non-convex configurations. Here, we propose a twin-stage Fuzzy Logic scheme for autonomous navigation of mobile robots. Our exhaustive computational and real-world experiments using Robotino, an omni-directional mobile robot architecture shows the potential to avoid local-minima and obstacle collisions, enabling the computationally efficient and adaptive FL-based behavioural modules for mobile robots.

Keywords: Mechatronics Systems, Control Theory and Technology, Mechanism Design
Abstract: This paper presents a control method to avoid the singularity problem in the newly developed active ball joint mechanism. This mechanism is based on gear meshing but allows the driving of three rotational degrees of freedom on a single spherical surface, and it has no theoretical limits in the range of motion. These capabilities were obtained by meshing two innovative gears: Cross Spherical Gear and Monopole Gear. Despite their spherical shape, the gears were able to mesh with each other due to a unique structure `Pole' . However, our previous works suggest that in the vicinity of the Cross Spherical Gear poles, there are singularity areas where the velocity of the Monopole Gear changes rapidly. This paper describes the modelling and formulation of the singularity area and a method to avoid the associated problems. Finally, the developed method was embedded in the prototype's control system, and the effect was verified by experimental results.

Keywords: Environment Monitoring and Management, Systems for Service/Assistive Applications, Software, Middleware and Programming Environments
Abstract: Mobile robots are playing an important role as monitoring and immediate response services supporting risk management systems for natural or artificial disasters. In this work, an integration of systems based on a Unmanned Aerial Vehicle collects relevant information and show the status of a wildfire through an interactive dashboard, which can be accessed by the risk and disaster management systems through any computer. In addition, the system can identify people and communities at risk by presenting their estimated geolocation on a map, as well as the spread of the fire. In this way, it will be possible to understand the magnitude of the disaster, and thus develop better rescue and monitoring strategies.

Keywords: Autonomous Vehicle Navigation, Machine Learning, Intelligent Transportation Systems
Abstract: In the case of a mobility service using an automated vehicle, it is necessary to determine the safety of the passengers inside and outside the vehicle before starting the vehicle, and automation is required. Therefore, we have developed a system that uses camera image processing to determine the safety of passengers in and out of a vehicle. This paper describes a system for determining the passenger's status using a machine learning based image analysis method.

Keywords: Vision Systems, Autonomous Vehicle Navigation, Machine Learning
Abstract: In this paper, we present a novel pedestrian detection system for sidewalk snow removing vehicles particularly for night driving scenarios. The information in front of the snowplow is obtained by clustering and classifying objects using LiDAR point clouds. A robust pedestrian detection and classification algorithm using the support vector machine(SVM) is proposed. We tested the system on an actual machine and the accuracy our method is verified by experiments.

Keywords: Sensor Fusion, Autonomous Vehicle Navigation, Vision Systems
Abstract: Inclement weather conditions such as snowy environments poses a lot of challenge for autonomous driving. Because of the dynamic changes in the environment, there will be difference between the prior map obtained from a LiDAR system and current sensor data. To overcome this problem, in this study, we present a semantic segmentation based method to recognize the snow cover from the camera images and project the results on the LiDAR point cloud to distinguish the differences. Our results shows improved localization accuracy in snow environment.

Keywords: Vision Systems, Intelligent Transportation Systems, Autonomous Vehicle Navigation
Abstract: We developed a new method for 3D-obstacles detection using a 3D map. Three-dimensional-obstacles detection is a key function of autonomous driving. It is easy to detect static obstacles because they exist in the 3D map. However, the 3D detection of dynamic obstacles that are not in the 3D map is difficult for a typical in-vehicle camera that cannot measure the distance. We aim to detect dynamic obstacles three-dimensionally, using an in-vehicle camera. And we deal with the new problem of accurate 3D reconstruction by using a monocular camera and a 3D map. To solve this problem, we focused on semantic segmentation for detection and depth completion to complement the depth map. We propose a multitask neural network (NN) that shares the encoder of semantic segmentation NN and depth completion NN, whose inputs are an image and a 3D map. The proposed multi-task NN detects dynamic obstacles 1.4 times more accurately than the singletask state-of-the-art method.

Keywords: Vision Systems, Intelligent Transportation Systems, Machine Learning
Abstract: We aim to improve unknown object detection. We also deal with problem of designing the optimal class for semantic segmentation using depth image. There was a problem that unknown classes of obstacles were mistaken for road in semantic segmentation using depth image. Therefore, we focus on the superiority of 3D object detection in a depth image. The depth image is good at separating between horizontal plane and 3D objects. For this reason, we develop a method for changing the number of training classes from baseline 12 classes to new 3 classes (void, plane, 3D object) for segmentation, which are optimal to detect unknown object by using depth images. As a result, the IoU of our method is 6.9 points higher than that of a baseline method.

Keywords: Formal Methods in System Integration, Autonomous Vehicle Navigation, Motion and Path Planning
Abstract: Safely driving on the road is a highly complicated information process for autonomous driving systems, which requires knowledge in the form of the hierarchical structure to bind them together. According to the past works on Cooperative Intelligent Transport Systems (C-ITS), the concept of Local Dynamic Map (LDM) is one of the critical technologies. It can coordinate environmental information with different time scales to update by mapping onto a consistent geometric structure in a hierarchy. A possible representation of the LDM concept can be associated with a relational database or semantic web technology such as Resource Description Framework (RDF) and Web Ontology Language (OWL). However, moving objects on the road can take infinitely many states. In this study, we propose an integration with Linear Temporal Logic (LTL) to represent events with potential risks to express them on LDM. In the computer experiment, risky situations were examined by using LDM update rules with LTL at turning and roundabout.

Keywords: Systems for Service/Assistive Applications, Vision Systems, Welfare systems
Abstract: The authors aim to clean a toilet room using a home mobile robot. In this work, the authors developed techniques to recognize a toilet paper tube and a toilet bowl based on point cloud data obtained by an RGB-D camera on the robot. The robot was able to pick the tube up and wipe the bowl with its arm.

Keywords: Autonomous Vehicle Navigation, Intelligent Transportation Systems, Environment Monitoring and Management
Abstract: Automatic map quality assessment is a very important process to bring the mapping modules into levels four and five of autonomous driving. In this paper, we propose a robust framework to check the map quality on behalf of human beings with indicating the possible ghost areas without using ground truth. The essence is to conduct the assessment process in the image domain instead of the point cloud plane. Therefore, the road is described by a set of nodes and each node represents a considerable road texture in Absolute Coordinate System using LIDAR reflectivity. This converts the vehicle trajectory into grayscale images with encoding stationary landmarks and road shapes. In addition, the global position errors are converted into relative position errors between the nodes and transformed into ghosting effects in the image domain. Accordingly, a mechanism to evaluate the map quality at the revisited areas is proposed based on sharpness, luminance and structure factors of the road surface. The framework has been tested in challenging environments including open-sky areas, the world’s second-longest tunnel and courses of dense trees and high buildings. The experimental results have verified the novelty and reliability of the proposed strategy to provide very trustful map quality assessment by relying on map images only. Moreover, the system is scalable to compare the maps and significantly indicates the outperformance in terms of accuracy and quality.

Keywords: Autonomous Vehicle Navigation, Sensor Fusion, Multi-Modal Perception
Abstract: In this paper, a visual-inertial localization system that reuses a prior map built by Lidar is proposed. Relying exclusively on a monocular camera and an IMU, the point and line features detected in the images are reconstructed and utilized for geometrically estimating the relative pose of the robot with respect to the prior 3D point cloud map. To leverage the alignment between the body frame and the map frame, a modified normal distribution transformation(NDT) algorithm is tightly coupled into the bundle adjustment(BA). We extract dual layered grid cell map from the raw Lidar-built map for both the point-to-distribution and line-to-distribution registration. By utilizing line features, the proposed method can achieve competitive performance in low textured environments. Evaluations on different real-world environments, including tests on both the benchmark dataset and the self-collected one are presented.

Keywords: Autonomous Vehicle Navigation, Systems for Field Applications, Machine Learning
Abstract: Autonomous road surface following on forest roads by mobile robots and forestry vehicles carrying large logs is required for work in the forestry industry. In such situations, the detection of a passable road surface has to respond to changes in road geometry, surface conditions, and lighting. In this paper, we propose an efficient road detection method using LiDAR-SLAM and U-Net architectures: LiDAR-SLAM can accurately estimate the shape of the road in response to environmental changes, while U-Net architectures can efficiently estimate the edge of the road in a forest road. In the experiment, we used IoU (Intersection over Union) to evaluate the accuracy of the road surface detection in passing through a forest road. As a result, the proposed method achieved an IoU value of more than 90.2%.

Keywords: Machine Learning, Autonomous Vehicle Navigation
Abstract: This paper presents DNN-based self-attitude estimation by learning landscape information. The network predicts the gravity vector in the camera frame. The input of the network is a camera image, the outputs are a mean vector and a covariance matrix of the gravity vector. It is trained and validated with a dataset of images and correspond gravity vectors. The dataset is collected in a simulator. Using a simulator breaks the limitation of amount of collecting data with ground truth. The validation showed the network can predict the gravity vector with only single shot image. It also showed the covariance matrix expresses the uncertainty of the prediction.

Keywords: Machine Learning, Vision Systems, Autonomous Vehicle Navigation
Abstract: Convolutional Neural Networks (CNNs) have demonstrated impressive performance in complex machine learning tasks such as classification and regression problems. A reliable neural network structure plays a decisive role in CNN studies. Through comparing and analyzing the structure of neural networks, a model structure for better visualization performance has been discovered, and such a method supports the development of deep learning research. These studies are of particular importance in end-to-end systems for autonomous driving to imitate human driving, where the interpretability of the system is limited. Because of the uncertainty of the ground truth, for the determination of human steering in an image, it is difficult to accurately compare the visualization performance of different CNN models or different visualization methods. For practical applications, however, an objective and quantitative measure for assessing visualization performance is necessary. Therefore, a method to evaluate the visualization performance of CNN models using a driver model instead of human drivers is proposed, to generate a data set which can be used to determine the decisional point (ground truth) in the input image. Then, an exclusive method is also put forth, to quantitatively calculate the relationship between the decisional point (ground truth) and the visualization results produced by CNN models. In this paper, five CNN models as an autonomous steering controller are designed based on PilotNet, and the visualization abilities of each CNN models is compared by three evaluation indicators. By comparing the visualization performance of five different CNN models, it is shown that the proposed method can successfully assess the visualization level of the CNN model.

Keywords: Human Factors and Human-in-the-Loop
Abstract: Abstract— In order to increase additional value in the development of products and services, it is important to be able to explain the relationship between external stimuli given to end users and their sensibilities and preferences. Although, general methods in current tactile modeling is multivariate analysis to the characteristics of material samples obtained from commercial measuring instruments, the residuals of these regression problems include two kinds of variances entangled. First, the individual deviation caused by hand and touch method-based differences. Second the repetitive error of the answers caused by the difficulty of the questionnaire survey. Therefore, in order to consider the individual deviation due to the hand and touch method-based differences between panelists, we measured the physical stimuli between the leather material and the panelists actually touching the sample and added them to the explanatory variable. Then, using latent factors of structural equation modeling (SEM), the repetitive error of the questionnaire was attempted to be taken into account while modeling the hierarchical structure of low-order perceptions such as “soft/smooth” and higher-order sensitivities and preferences such as "luxury/comfortable". In the conventional SEM, in case of large error is set on an observed variable, the calculation becomes unstable due to the local solution in which the error variance of a specific node becomes almost zero. But in this study, we have developed a method to enable stabilized calculation while considering a large error using the Bayesian SEM.

Keywords: Human Interface, Human-Robot/System Interaction
Abstract: This paper describes a light-weight method for hand-waving gesture detection. Gesture recognition is actively researched as a user interface of robots. Conventional gesture recognition methods need to employ complicated pattern matching algorithms, such as DTW, HMM, and DNN, which require a powerful computing platform such as fast CPU or GPU that consumes much energy. We propose a gesture recognition/detection method specially designed for the recognition of hand-waving gesture. This method uses Kinect V2 as the sensor and detects the waving gesture using only a simple signal processing. The recognition experiment suggested that the proposed method gave sufficiently high accuracy, and the processing speed was much faster than real-time.

Keywords: Human Interface, Human Factors and Human-in-the-Loop, Virtual Reality and Interfaces
Abstract: An ability of visually-guided and anticipatory adjustments of locomotion corresponding to upcoming obstacles is important to avoid trip-induced fall. For establishing gait training based on visually-guided and anticipatory adjustments, techniques reproducing realistic training environment are essential. Although some previous works proposed virtual obstacles using mixed reality, the feasibility of virtual obstacles encouraging people to perform realistic obstacle negotiation on a treadmill, where gait training is usually conducted, is still unclear. In this study, we investigated toe heights when stepping over the obstacle in both cases of virtual and real obstacles during walking on the treadmill. Five participants stepped over two types of mixed reality boxes and real boxes, with box placements close and distant from them. The results generally indicate that the toe heights of the leading foot tended to be similar between mixed reality and real obstacles in cases where the obstacle was located distant from participants, a condition that enabled participants to predict when obstacles approached. However, the toe heights of the trailing foot tended to be lower when stepping over the MR obstacles than the real obstacles. We discuss the feasibility and shortcomings of the future use of MR HMDs as replacement for traditional gait training setup.

Keywords: Human Interface, Control Theory and Technology, Human-Robot/System Interaction
Abstract: The advancement of robotics has created situations where the robot assists the human operator. How to interact between the human operators and robots is essential for the assistant. They have different specialty fields, and they should play each role. In this paper, we focus on the "operator and observer control style" teleoperation user interface based on the interaction between the autonomous robot (operator) and the human (observer). For example, the human observer will view the workspace from a broad perspective, expedite the flow of work, send commands, or advise the robots, whether industry or on a farm. We developed a semi-autonomous teleoperation system for remote object searches that can assist in the search using an omnidirectional camera. The proposed system can prevent task failure due to erroneous recognition of the object by advice given even when grasping the object. To verify the proposed style, we chose a bring-me task, in which the robot searches for and fetches a specified object from several objects in a room in an indoor environment. We conducted a subjective experiment with four conditions (two camera conditions x two operational conditions) and evaluated the usability of the SUS and execution time of the system with fourteen participants. The result indicated that our teleoperation user interface is more usable than the directly controlled teleoperation style commonly used.

Keywords: Human-Robot/System Interaction, Human Interface, Systems for Service/Assistive Applications
Abstract: This paper describes topic decision system to select topics that users are interested in but are not familiar with. In recent years, dialogue systems have been implemented in various devices such as smart phones and robots. Although many of them are categorized into task oriented dialogue systems, non task oriented dialogue systems are also important for long-term human-system interaction. Many dialogue systems are also question answering systems, that is, the user inputs a question and the dialogue system returns its answer. However, it is important that the dialogue system can start the conversation to users especially in non task oriented dialogue systems. Adequate topics are required to start the conversation in such cases. We therefore aim to construct a topic decision system. In this study, we define adequate topics as the topics that users are interested in but are not familiar with and propose a topic decision system using the keyword search function of Twitter.

Keywords: Hardware Platform, Human Interface, Human-Robot/System Interaction
Abstract: Abstract— In the power distribution work using the indirect hot-line work technique, it is necessary to frequently set a heavy tool with a weight of 15 [kg] on an electric wire. To reduce the burden put on the worker during these hot-line works, we developed a motor-less power assist device that can completely compensate for the weight of the tool in a wide range of 600 [mm] in the vertical and horizontal directions and 90 [deg] in the pitch direction. The proposed device is lightweight and compact, so that it can be mounted on the bucket of the aerial work vehicle. To show the effectiveness, we measured the myoelectric activity during the overhead wire work and the time required to carry out the work. It was verified that the load on the muscles was reduced by 90% or more, and the time was suppressed to the same level as without assist.

Keywords: Multi-Robot Systems
Abstract: This paper proposes a distributed control law to improve network connectivity in cases some robots tend to breakdown or power-off. Although high node-connectivity engages connectedness against robotic-nodes failure, keeping high connectivity restricts configuration space of the multi-robot system and disturbs some cooperative tasks. So we are required to robustify the network structure only when the frangible robots appear. Here we employ a perturbed algebraic connectivity which we proposed in previous studies, and compose a control law to meet the control objective in this study. Numerical simulations show validities and utilities of the control law.

Keywords: Motion and Path Planning, Mechatronics Systems
Abstract: This paper presents a method for generating curved path sequences in a humanoid robot with a deficient degree of freedom (DOF).Typically robots not having the yaw DOF in their limbs cannot turn in a curved path. This paper aims at achieving that by combining the robot’s friction based slip turn mechanism with the simple straight walking gait of the 3-D linear inverted pendulum model into a smooth curved motion. The amount of turn caused by the slip model is estimated and multiple experiments are conducted to verify the prediction. We noticed that the co-efficient of friction does not affect the amount of turn for symmetric gaits. A novel mathematical model which uses the radius of the curved path to plan the gait for generating a smooth circular trajectory, has been developed. A 17-DOF robot named TONY is used to test the hypothesis. The results are successful and the robot is able to walk in a curved path.

Keywords: Human-Robot Cooperation/Collaboration, Human-Robot/System Interaction
Abstract: This paper proposes a method for generating the reference Zero Moment Point (ZMP) for teleoperated bipedal robots walking on non-flat terrains. It consists the preview control and auxiliary ZMP method to realize ZMP-based walking. An auxiliary ZMP is used to generate the COG trajectory in realtime by reducing the time delay caused by the preview control. The problem with such ZMP-based walking is that a ZMP cannot be defined when the both feet are not on a same plane. Hence, in the proposed method, the virtual ZMP method is applied to calculate the ZMP. In order to calculate the virtual ZMP, the proposed method determines a virtual plane from the positional relation between the feet. This method does not use environmental information to determine the virtual plane. Therefore, this method is suitable to generate the reference ZMP of the teleoperated bipedal robot. The proposed method is validated through a realtime simulation environment involving haptic devices.

Keywords: Systems for Service/Assistive Applications
Abstract: In this study, a method is proposed to carry out food peeling task using our dual-arm robot system with a peeler. First, the problem of peeling a food using robot arms is discussed.Then, a new method for food peeling is introduced, which contains two parts: “One peeling motion” in which the robot peels part of a rind or skin through a pre-generated trajectory, and "Rotation motion" where the robot calculates the pose of food for the next "One peeling motion".The introduced peeling method was implemented into the dual-arm system and then put into test by peeling five types of food. The results prove that the proposed method can peel the skin of various types of food successfully.

Keywords: Control Theory and Technology, Motion and Path Planning, Autonomous Vehicle Navigation
Abstract: This work tackles a task to transport an object by a mobile manipulator to the designated location. Heavy objects that the robot cannot lift up and needs to push or drag are targeted. It is assumed that the mass property of the object and the frictional characteristics between the object and the floor are unknown. In addition, a caging strategy of loosely holding the object is employed in order to reduce the required motor torques. Thus, a challenge is how to deal with uncertainties accompanied with interactions between the robot, the object and the ground. A control system that stands on a combination of an instantaneous equilibration model of a planar frictional movement and the adaptive model predictive control (MPC) to modify the model parameters during the operation is proposed and implemented on a mobile manipulator. Efficacy of the proposed method was verified through experiments, in which even objects that are equipped with casters and thus have largely different characteristics from the supposed friction model were successfully moved to the desired destination.

Keywords: Surveillance Systems, Vision Systems
Abstract: Iris is an advantageous feature of biometric authentication that can be measured without contact and low possibility of damage and change. Conventional iris authentication systems, however, require approach to measurement devices or stationary of users due to necessity of the iris image with sufficient resolution, which reduces time efficiency of the authentication. In this paper, we propose a dynamic iris authentication system using high-speed gaze and focus control by high-speed image processing. We control high-speed rotational mirrors and liquid-based variable focus lens by triangulation with a wide-angle camera. We also control the liquid lens with additive modulation of sine wave to get the most focused iris image. We have experimentally evaluated the sufficient focus measure around eyes of static and dynamic users, and confirmed successful iris authentication.

Keywords: Human Interface, Soft Robotics, Motion and Path Planning
Abstract: To provide better guidance for patients or the elderly during the self-rehabilitation, a haptic system based on the asymmetric vibration was proposed. Due to the Non-linearity of Perception asymmetric vibration periodic movement can provide a sense of pushing and pulling. In order to better support human motion, we evaluated the sensitivity of different parts of the human body to this haptic assist. And by combining with the gyro-sensor through the contrast experiment, we compared the accuracy of motion learning with or without the haptic assist. The results of three subjects showed that the accuracy of motion learning was improved compared to without haptic assist. Therefore, the system has great potential in the guidance of patients' self-rehabilitation.

Keywords: Medical Systems, Rehabilitation Systems
Abstract: Prolonged seated posture can cause thrombosis in the veins of the lower extremities, leading to deep vein thrombosis (DVT) and pulmonary thromboembolism (PTE). To overcome these problems, it has been proposed to stimulate blood flow through the lower limb muscles by repeated contraction and relaxation of the pump. In this study, we hypothesize that the timing of contraction in the lower limbs can be synchronized with the timing of heart relaxation and blood pumping from the veins to maximize blood flow, and we have developed a system that synchronizes the timing of stimulation with the timing of maximum blood flow velocity in the veins of the lower limbs based on the timing of the measured electrocardiogram.

Keywords: Virtual Reality and Interfaces
Abstract: Carbonated drinks are widely used as a luxury item because they provide a feeling of coolness and refreshment just by drinking them. However, once the package is opened, the carbonation is removed over time. The amplification of the carbonation sensation and the taste of carbonated water by changing electrical parameters such as current, frequency and duty ratio was evaluated. In electric taste system, the current command from the signal generator is input to the constant current control circuit, and the stimuli are presented to the user through the electrodes. A two-layered cup-shaped device is used to present the electrical stimuli. The voltage is applied to the beverage when the conductive tape with the anode wired and the staple core come into contact with each other. Drinking a beverage while touching the conductive tape to which the cathode is wired creates a closed circuit through the beverage, tongue and hand. The system is designed to provide electrical stimulation to the user's mouth. The experiment was conducted on eight subjects with the eleven parameter patterns. The questionnaire items first examined the subjects' preference and frequency of consumption of carbonated drinks. Next, they were asked to respond to the five basic taste changes of carbonated water. Last, they were asked to respond to the amplification of the carbonation sensation by electrical stimulation and to changes in the five basic tastes of carbonated water. As a result, carbonation sensation is enhanced as the current value is increased and the frequency is decreased. On the other hand, there was a significant variation among individuals. There was no clear trend readily apparent in the Duty ratio. It is necessary to increase the number of test subjects, improve the reliability of the data by scrutinizing the experimental environment and procedures, and increase the number of Parameter patterns in the experiments.

Keywords: Sensor Fusion
Abstract: The goal of this research is to develop a system that recognizes the surrounding environment by the ultrasonic sensor in high resolution. In this study, the multiple signal classification (MUSIC) method is used to calculate the reflected wave's direction of arrival. However, the MUSIC method has a limitation of the number of arriving waves. In this research, the phased array transmitter is used to solve this problem. The transmitter emits directed ultrasonic beams to objects. We propose a system consists of an ultrasonic phased array transmitter and microphone array with the MUSIC method. The system is designed in which the number of emitted beams is less than the limit of the number of arriving waves. Finally, the experiment shows the system can localize multiple objects simultaneously.

Keywords: Machine Learning, Vision Systems, Systems for Service/Assistive Applications
Abstract: We improved accuracy of instance segmentation and pose estimation. Pose estimation is required to pick and place objects correctly if we develop a robot that stocks products. Mask R-CNN, a conventional method for instance segmentation tends to lower its accuracy under complex environments. Now, we focused on that robots deal with only one object per one task. We developed a new method that selects one instance that was segmented successfully and then refines it intensively. Our method lowered pose estimation error 1.3×.

Keywords: Vision Systems, Sensor Fusion
Abstract: This paper proposes a compact range image sensor with a multi-slit laser projector and a small camera. Distance is measured using image blur of the multi-slit laser lights. By determining the relationship between blur and distance per pixel in advance, 5.6 fps online measurement of range images is achieved. The effectiveness of the proposed sensor is verified by experiments of range image measurement.

Keywords: Hardware Platform, Vision Systems, Mechanism Design
Abstract: There are level gauges that measure the level with respect to the direction of gravity, but there are no compact devices like level gauges that measure the perpendicular direction to a line or a plane. The author has developed a lenticular angle gauge (LEAG) to visualize the relative orientation using a lenticular lens, and a "perpendicular gauge" to identify the perpendicular direction with respect to a line or plane using a LEAG. This paper describes the principle and structure of the perpendicular gauge and the results of a performance verification test using a prototype. The author will also discuss the various uses and future prospects for the perpendicular gauge.

Keywords: Vision Systems, Autonomous Vehicle Navigation, Virtual Reality and Interfaces
Abstract: Lenticular angle gauge (LEAG) is a planar pattern that visualizes relative orientation by the position of the black line, which moves according to the viewing angle. LEAG has the potential to turn arbitrary planar patterns into high-accuracy markers. The author has developed a new LEAG in which the direction of movement of the black line is 90 degrees different from that of the previous LEAG. With this, it is no longer necessary to place two LEAGs orthogonally to create a high-accuracy fiducial marker. This paper describes the principle and behavior of the new LEAG, and the results of its functional test. A new design of high-accuracy marker under development is also presented.

Keywords: Virtual Reality and Interfaces, Human Factors and Human-in-the-Loop, Entertainment Systems
Abstract: The research described in this paper presents a new scope of human manipulability sharing (SoHMS) using a head-mounted VR system and wireless EMG sensor. An application of the new SoHMS is developed in the Unity platform so that a high-quality computer mannequin can be presented in a 3D virtual space. The computer mannequin is controlled through the motion data of a player, which is measured using a KINECT v2 sensor in real time. A set of manipulability ellipsoids for the measured human body form is attached at the hand(s) and/or foot(feet) of the computer mannequin. The new SoHMS can provide two sports-like motor tasks with the upper limbs, which were designed based on the manipulability theory, so that a player can study how manipulability can be regulated at will.

Keywords: Software, Middleware and Programming Environments, Formal Methods in System Integration, Software Platform
Abstract: This paper describes the HAIROWorldPlugin which is a plugin for Choreonoid. We have developed the plugin functions to virtually configure decommissioning task environments and situations for remotely operated robots by referring decommissioning works have conducted in Fukushima Daiichi Nuclear Power Station of Tokyo Electric Power Company Holdings, Inc. (FDNPS). HAIROWorldPlugin is a package of our previous development results such as Fluid Dynamics Simulator, Visual Effect Simulator, Communication Traffic Simulator, Motion Recorder, Model File Explorer, Crawler Robot Builder, Terrain Builder, and Operation Command Manager. Each function is explained, and the examples of demonstrations are presented.

Keywords: Software, Middleware and Programming Environments, Integration Platform, Software Platform
Abstract: This paper describes a simulator-based system for testing skill to maneuver the robot remotely. Our motivation is to apply the robot simulator to the skill verification process of the remote operation of robots. As a first step in achieving this, we developed the functions which work in conjunction with Choreonoid for collecting the data during operation and displaying collected data after the trial on-demand. In this paper, we described concrete implementation for considering the task of passing through the narrow passage in the dark and the result of a test run by using the developed prototype system.

Keywords: Systems for Service/Assistive Applications, Human-Robot/System Interaction, System Simulation
Abstract: We present the application of soft actor-critic (SAC) learning algorithm to train a mobile robot to attend a target person at specific locations inside a Gazebo simulator. Since our previous study confirmed that the appropriate attending position while the target person is standing or walking is at his left or his right side, we design a novel U-shaped reward function behind the target person's position with respect to the robot's position. To make the robot can better portray the surroundings, we also propose a novel SAC architecture which employs 1D convolutional neural networks to extract features from laser scans automatically during the training process. Our preliminary experiment result shows that the robot is able to attend the target person at the designed location using our proposed reward function and SAC architecture.

Keywords: Biomimetics, Biologically-Inspired Robotic Systems, Machine Learning
Abstract: In this study, we modeled the insect behavior of chemical plume tracing (CPT) by learning from measured data as a stochastic process. We expected that our modeling method could reproduce an insect's CPT behavior better than the typical motion strategy, which is a time-invariant system. As training data for modeling CPT behavior, we focused on the input of the odor stimuli and output of the CPT behavior. We constructed a sensor-array measurement system to obtain stimuli and actions at the same time. The relationship between these input-output data was described by a stochastic process to construct the CPT model based on these measured data. We conducted simulation CPT experiments for evaluating the performance of such model. As a result, we found that own model had a high CPT performance and robustness toward an unknown environment.

Keywords: Decision Making Systems, Human-Robot/System Interaction, Environment Monitoring and Management
Abstract: Precise observation of the behaviour of subjects in ethological experiments is a highly resource demanding process. Trained observers carefully identify all behavior elements, called coding, either in real-time or by watching recorded video footage. Different observers can code slightly differently, have disagreements, and, even a highly skilled observer can make mistakes. In order to improve the efficiency and accuracy, and to reduce the resource needs of the coding process, many computerized solutions have been introduced already. These solutions can not only save significant time and effort by using automatic timestamps, efficient footage and data handling, etc. but provide sophisticated statistics and analysis features. Anyhow, most of the coding aiding systems in use cannot determine the exhibited behaviour elements by themselves, it still remains the task of the human observer. In this short paper, we propose the basic concepts of a method capable of automatically coding behaviour elements in a simulated environment, without the need for a skilled observer person. The method was designed to evaluate the simulated dog-human interaction experiments, without having any information on the inner states of the simulation, based solely on the spatial position data of the participants. A preliminary evaluation was performed, showing that the method is capable of correctly recognizing the behaviour elements exhibited by the agent.

Keywords: Machine Learning
Abstract: The learning phase of the traditional reinforcement learning (RL) methods can be started without any preliminary knowledge about the problem needed to be solved. The problem related knowledge-base is built based on the reinforcement signals of the environment during the trial and error style learning phase. If a portion of the a priori knowledge about the problem solution is available and if it could be injected into the initial knowledge of the RL system, then the learning performance (and the learning ability of agent) could be significantly improved. The main goal of this paper is to highlight the effect of the external expert knowledge inclusion into the Fuzzy Rule Interpolation based Q-learning method, by briefly introducing a way for expert knowledge injection into FRIQ-learning and a discussion based on a simulated runs of a practical benchmark example. The methods presented here can aid designing behaviour-based robot control systems, in such cases where the available expert knowledge is not enough by itself to construct a sufficiently working system.