Peripheral blood mononuclear cells (PBMCs) are critical for immune system testing, research, and cellular medicine, and are typically isolated from blood samples. However, the separation process can often be unstable and cumbersome for both patients and operators. Previous attempts to automate this process using large pipetting machines connected to centrifuges and other equipment have shown a significant drawback: if any part of the system malfunctions, the entire operation halts, resulting in prolonged downtime. This study aims to enable independent operation of each PBMC separation process. We focus on automating the decanting step, which has been identified as a significant source of instability in PBMC separation. We began by analyzing the motions of operators and developed a concept model to replicate these movements. We then established a method to evaluate the decanting operation, allowing the concept model to perform various tasks, which we subsequently evaluated. Finally, we conducted blood separation using the most effective operation and validated both our evaluation method and the concept model. Although the potential exists for further refinement of the evaluation and verification processes, the separation obtained in our study was comparable to that achieved by skilled operators.

The authors developed shoes capable of measuring vertical force and friction force on the sole at four points where weight is concentrated, and devices that can feed back the force to reproduce the sensation of the sole during standing posture. These shoes are equipped with triaxial force sensors embedded in the soles at four points, and a roller for reproducing friction force is attached to the tip of an existing linear motion mechanism for reproducing vertical force. The authors also examined the effects of each force on motion estimation.

In this study, we propose an autonomous navigation system for a snake-like robot with the aim of efficiently searching for victims during disasters. This system was developed using PPO (Proximal Policy Optimization), a type of deep reinforcement learning algorithm. First, we constructed an environment in the simulator with a destination and obstacles, and placed a 3D model of the snake-like robot within it. Then, by repeatedly running the snake-like robot from the starting point to the destination, we confirmed that it acquired the ability to autonomously reach the destination while avoiding obstacles.

In this study, the authors propose a lightweight, earthworm-inspired robot equipped with circumferential brushes for inspecting large-diameter pipes in elevated or confined spaces. The robot achieves propulsion by gripping the pipe wall using radially arranged brushes actuated by pneumatic artificial muscles. This paper presents a method for switching propulsion direction via directional brush deformation, along with a theoretical model.

We investigate reinforcement learning (RL) for command-conditioned locomotion of hexapod robots in uneven terrain using simulated environments. Due to their high degrees of freedom, hexapods are prone to local optima during learning, resulting in suboptimal policies. To address this, we apply a data augmentation method called Virtual Command Allocation (VCA), which replaces commands in collected experiences with randomly sampled ones from the command space. Simulation results across multiple terrain conditions show that VCA improves the stability and generalization of learned policies, enabling robust and adaptive locomotion. These findings demonstrate the effectiveness of VCA for learning command-conditioned control even in challenging and variable environments.

This study focuses on the development of a harvesting end-effector for agricultural robots designed for ``Synecoculture^TM,'' a sustainable farming method from Sony CSL. The primary objective is to create a small, versatile end-effector to efficiently harvest crops without damage, even in Synecoculture's dense growing environments. We developed a multi-point cushion with bellows-type modules made of a new TPU material. The design achieved successful gripping by enveloping crops without pneumatic systems. This allowed gripping tomatoes with diameters from 67[mm] to 50[mm] without requiring position control.

Robot teleoperation is essential for the practical application of robots in real-world environments. While existing research has focused on visual feedback using VR devices and intuitive leader-follower control, these methods are heavily dependent on the operator's skill and face challenges in high-speed teleoperation. Therefore, this study proposes a control system for high-speed teleoperation by extending conventional visual feedback methods and incorporating feed-forward control for velocity compensation.

In this paper, simulation analysis of the net shape and capture conditions (enclosing and winding) during space debris capture is performed. The relationship between debris size, net face size, and ``enclosing'' and ``winding'' behavior during debris capture is shown by analyzing the differences in capture behavior due to differences in net shape through simulations that simulate a microgravity environment. As a result, the number of points of contact between nets during stable debris enclosing and the design method of net size for debris size were clarified.