This paper proposes a pulling-driven soft hand for fabric manipulation. Automatic fabric manipulation is required in various industries such as garment industry, linen supply industry, automotive parts industry, and composite manufacturing, where various fabrics with different materials, shapes, and surface properties are used. Therefore, we proposed pulling-driven soft hands to pick up various fabrics. We find that picking up a fabric on a table requires soft fingertips contacting with the fabric, large friction between fingertips and the fabric, and fingertips moving along the table to maintain the contact with the fabric. We thus introduced a closing-approaching coupling mechanism to the pulling-driven soft hand. We fabricated a prototype of the proposed soft hand to demonstrate the grasping of a single fabric and picking multiple fabrics one by one from their stack. Then, we conducted a static analysis of fabric grasping to clarify the conditions to pick up a single fabric from a fabric stack. We also experimentally verified the conditions to validate the static analysis.

One of the significant challenges in the development of care robots is recognizing and adapting to individual differences in users' physiques and symptoms, thereby providing personalized assistance. Among Activities of Daily Living (ADLs), dressing assistance poses a particularly complex problem due to the physical interaction between clothing, the human body, and the robot. This complexity is further amplified in care settings, as the primary users are elderly or disabled individuals, whose physical characteristics vary widely. This study focuses on the physical traits of elderly individuals targeted by dressing assistance robots and proposes an adaptive method that accommodates variations in body size and kyphosis. The proposed approach utilizes Dynamic Movement Primitives (DMPs) to generate dressing trajectories that conform to the user's body, based on skeletal information estimated by a vision sensor. Experiments were conducted with eight healthy individuals of varying body types simulating kyphotic posture, one healthy elderly participant, and one elderly participant with posture-related impairments due to Parkinson's disease. The results demonstrate that, compared to non-adaptive methods, the proposed approach enables more adaptive and less burdensome dressing assistance.

In this study, we propose a method for estimating the user's turn-ending intention (turn-end estimation) and turn-taking intention (interruption estimation) in real time, enabling dialogue agents to achieve natural and smooth turn-taking with humans. The proposed method uses linguistic cues (speech content) to extract features using a large language model (LLM), followed by estimation using a classifier (SVM). The implemented estimation models achieved high accuracy (F1 score of approximately 90%) in Japanese dialogue. Furthermore, it was confirmed that the processing time was within tens of milliseconds, demonstrating real-time responsiveness.

Water leakage from cracks in concrete structures often causes corrosion-related damage to ceiling materials in overhead spaces. Early detection and repair through regular inspections are essential. However, there are many obstacles in such space, making visual assessment difficult. In this study, we propose a Wire-driven Earthworm Robot capable of locomotion in confined environments by mimicking peristaltic motion. Experimental results demonstrated that the robot achieved a forward velocity of 28.4[mm/s]. Furthermore, by applying differential tension to the wires, the robot was able to lift its front section and overcome obstacles approximately twice the height of its wheel radius.

In recent years, controlling multi-degree-of-freedom prosthetic hands through myoelectric recognition has enabled diverse hand movements. However, existing products mainly focus on reproducing degrees of freedom, overlooking the finger configuration essential for precision pinching, thus restricting grasping postures. This study aimed to enhance passive grasp stability by integrating flexible finger pads and joints in the thumb IP joint and the DIP joints of the index and middle fingers. This approach enabled precise grasping by allowing IP and DIP joint hyperextension and pad-based gripping. Improved grasping performance was validated through grasping experiments.

This study proposes a haptic interface mounted on a robotic arm that presents tactile stimulation through magnetic force control. An experimental system consisting of a planar coil and a neodymium magnet was constructed to validate the magnetic force control model. In addition, a fundamental experiment was conducted in which a magnet was attached to a human finger to explore a method for delivering tactile stimuli while avoiding collisions between the finger and the robotic arm.

This study presents a likelihood-weighted method for estimating the relative states of multiple robots based on the consistency of their mutual observations. Instead of explicitly detecting sensor faults, the proposed approach evaluates the likelihood of each observation under the assumption that the others are correct and assigns weights accordingly. This allows outlier measurements to be suppressed smoothly without discarding potentially useful data.

Recently, imitation learning, which reproduces human actions, has attracted attention. In this study, we aim to realize a robot capable of grasping flexible objects using imitation learning. Specifically, two types of plastic bottles are prepared, and at the moment the robot grasps a plastic bottle, it discriminates between the two and reproduces human force. In this paper, we propose a method to derive stiffness from the angle and torque of the robot's end-effector, and to simultaneously learn the angle, angular velocity, torque, and stiffness using LSTM (Long Short-Term Memory). This method enables the robot to take into account the stiffness of the object at the moment of contact and to reproduce delicate human motions with higher accuracy.