We propose an optimization-based time-series inverse kinematics for robot motion generation. In this method, the design variables are the combination of time-variant configuration, which is time-series joint position, and time-invariant configuration, which is the grasping point or the robot location. The inverse kinematics problem is regarded as an optimization problem, and the sequential quadratic programming is applied by describing the target motion as a task function and deriving its gradient. The generated motion is smooth because of the regularization of adjacent joint displacement. We show various robot motions generated by the proposed time-series inverse kinematics.

Light touch contact (LTC) proposed by Jeka in 1997 can reduce unstability in human standing by the user lightly touching a curtain or similar with fingertips. However, the potential contribution of LTC to human posture control strategy has not yet been elucidated. This paper proposes a novel human standing control model designed to reproduce the influence of LTC on sensory reweighting which is a principle of control strategy for human standing posture. The approach allows us to represent time-series changes in sensory weights caused by LTC and express body sway in a standing state. Simulation using the proposed model indicated its capacity to reproduce human sway more accurately than a previous model, and experimental results suggested that sensory reweighting was influential in standing attitude control.

In this paper, we experimentally verify the relationship between the change in the stiffness of joints around the lumbar due to dynamic tightening control of the Active Corset and the reduction of the lumbar load. The bending angle of joints around the lumbar in lifting movement were measured, and the elasticity of the joints were calculated. As the result of the measurement on 21 subjects, tightening of Active Corset reduced the lumbar load with substituting the lumbar joint flexion by the hip joint flexion in all subjects. In addition, the we confirmed that the ratio of the lumbar elasticity to the hip joint elasticity increase with tightening around the pelvis on the four subjects. We conclude that the decrease of the lumbar load with tightening caused by increasing the ratio of the lumbar joint elasticity to the hip joint elasticity.

There are many pipe-systems such as a water pipe, a gas pipe and a ventilation duct in a plant, factory and house. The inspection of these pipe-systems is necessary to prevent accidents and malfunctions. However, many pipes are very narrow, and it is difficult for people to inspect directly. In this study, a brush wheel type robot which has the simple and compact drive mechanism and does not need large space to move, is proposed. When moving inside of a pipe, this robot uses the elastic force of a brush wheel to hold the robot in pipe. Additionally, the robot can move a vertical pipe and a curved pipe by replacing the brush wheels. This paper reports the robot's structure, the drive mechanism, the operation principle, the prototype evaluation and three type pipe running experiments.

Investigation of active volcanoes by robots is required to grasp their situation. Considering that volcanic environments are rough terrain, tracked robots are suitable for the investigation. When a tracked robot travels on a volcanic environment, it must climb over obstacles. The obstacles on a volcanic environment can be roughly divided into “fixed obstacles” which can be moved by a robot and “unfixed obstacles” which cannot be moved by a robot. Although a tracked robot climbing over unfixed obstacles such as unstable rocks has risks of sliding-down and tipping-over, there is little research about climbing over unfixed obstacles. On the other hand, grousers on track belts are effective for climbing on fixed obstacles, such as steps or stairs. However, it is unclear whether the grousers are also effective for climbing over unfixed obstacles or not. Therefore, the research purpose is to reveal the effect of grousers for climbing over unfixed obstacles. In this study, the climbing experiment using a cylindrical obstacle and grousers with several conditions of height and gap was conducted. As a result, it was found that grousers also affect to improve the climbing performance for unfixed obstacles. Especially, higher grouser and grouser with a gap which is more than the size the obstacle just fits indicates better performance. Also, the sliding-down condition based on statics was derived to predict the climbing performance of tracked robots. Comparing the condition and the experimental results, it is reasonable for low-height grousers. According to the above research, it becomes clear that the effect of grousers on climbing performance for unfixed obstacles on a two-dimensional plane.

This study introduces a concept of transfer learning to search tasks such as function approximation and optimization, and aims to achieve effective search using fewer number of samples. We propose a search procedure transfer (SPT) algorithm which extracts knowledge of efficient search procedure from well-known task and uses it to search similar unknown tasks. Experiments reveal that the closer the target task becomes to the source task, the higher the performance of the SPT algorithm becomes. In addition, experiments also demonstrate the SPT algorithm shows higher performance than the existing method using the Bayesian optimization algorithm (BOA) when the difference between a source task and a target task is small. Applying the idea of domain randomization, the SPT algorithm can utilize even human ambiguous heuristic knowledge.

Visual servoing is capable of positioning robots based on images captured by cameras. To calculate the command value for robots, hand-designed image features and extraction of image features are required. Positioning accuracy is significantly influenced by the selection of the image features. In this study, we focus on the ability of convolutional neural networks (CNN) to extract features from images and output the angular velocity to control a manipulator. We propose a visual servoing technique based on CNN enabling the precise positioning of a texture less object grasped by a parallel gripper. The positioning can be achieved even the grasping position is different from the position when the target image was captured. The positioning accuracy of the proposed method is verified based on the positioning of an object into an alignment tray using a six-DOF manipulator. We confirmed that the proposed visual servoing technique can position an object precisely.

In recent years, the aging population has become a serious solid problem in developed countries. Whereas the number of caregivers is decreasing gradually, and the care burden is increasing. In this study, a novel power add-on driving unit (PAU) is developed for the reduction of the caregiver's care burden. The PAU can be attached to or detached from the manual wheelchair easily. This PAU is very practical and has merits in cost, portability, compatibility and so on. Since the PAU is propelled by the caregiver, a novel control method is necessary for the ease of use. In this paper, a force sensorless observer is proposed, and the burden of wheelchair running on the slope is analyzed from kinematics. Finally, the effectiveness of our proposed burden reduction method is verified by the experiments.

Recent growth of a microscope and micromanipulators have been remarkable. With the growth, micro-manipulation in the micro world (less than dozens of micro-order) is significant in a variety of fields. However, 3D-estimation in microscopic view is difficult. We have developed a high-precision depth estimation method based on difference between blur widths of a particle and a tip of micromanipulator in our previous study. In our proposed method, one shot microscopic image can estimate precisely a depth distance without lens parameters. However, our method has a problem of a derivation of the depth equation. We propose a technique to derivate the equation using by Robust estimation and Mahalanobis' distance. To confirm stability and precision of the proposed technique, we show the validity through experimented results.

If robots can reproduce human-to-human physical contact which people feel happy, we expect an effective reduction of stress by robots. For reproducing physical contact, we aimed to achieve both the “active touch” function (shaking hands) and the “passive touch” reproducibility. We devised a mechanism and a production procedure to reproduce the “passive touch” such as fleshiness, pliability, and nail and made a prototype. We introduced an elastic material for maintaining force against external force to the wire-driven joint; we think this elasticity is necessary for “passive touch” reproducibility. By trying the actuation of the prototype and verifying the feeling with touch in the participant experiment, we confirmed the validity of the trial artificial hand.

This paper presents a localization approach that simultaneously estimates a robot's pose and class of sensor observations, where “class” categorizes the sensor observations as those obtained from known and unknown objects on a given geometric map. The proposed approach is implemented using Rao-Blackwellized particle filtering algorithm. The robot's pose can be robustly estimated utilizing sensor observations obtained from the only known objects by the simultaneous estimation. The proposed approach is efficient in terms of computational complexity because its complexity is same as that of the likelihood field model. Performance of the proposed approach was shown through experiments using a 2D LiDAR simulator.

In this study, by utilizing GNSS Doppler and low cost sensor, we estimate vehicle motion robustly and estimate highly accurate vehicle trajectory even in urban environment where satellite signal deteriorates. In addition, we propose a method that can improve the position accuracy by selecting the positioning result of GNSS using highly accurate vehicle locus. The proposed method intentionally does not estimate fluctuation of the clock error of the receiver, thereby making it possible to use the long distance (several 100 meters) trajectory and has the feature that position estimation accuracy is improved by the averaging effect.

This paper investigates a constrained control problem of a rigid body in three dimensions via Explicit Reference Governor (ERG), which suitably modifies the exogenous reference to satisfy the constraints based on invariant sets. In this paper, force/torque input constraints and pose (position and attitude) constraints are considered. The objective of this paper is to propose a control scheme to regulate the rigid body pose to the desired pose while enforcing constraints at all times. To achieve the goal, a pre-stabilizing pose regulation controller is first introduced while ignoring the constraints. The pre-stabilized system is then augmented with the ERG to handle the constraints. All the proposed methods are based only on the special Euclidean group SE(3) elements without any parameterizations. Numerical simulations are finally carried out to show the effectiveness of the proposed methods.

In this paper, we propose a method to classify the posture of a driver in a vehicle using time-of-flight (ToF) range sensors and machine learning. Conventionally, to estimate the posture of a driver, there are methods using RGB or depth cameras around the driving seat. However, previous estimation methods needed to handle large data from the cameras, and privacy can be an issue in such a camera based system. In this study, we propose a range sensor based driving posture classification method. We design a sensing system which consists of range sensing modules around the driving seat inside a vehicle, and the system identifies the posture of the driver by distributions of the sensor values. We installed the system into a vehicle and verified the feasibility of the proposed method.