There is growing interest in using X-ray computed tomography (CT) for inspecting industrial products, owing to its potential for reverse engineering and model-based design. However, to fully utilize CT scan data of assembled components, individual part segmentation is essential. Conventional thresholding methods are often ineffective when there is no luminance contrast at the boundaries between parts, making it particularly challenging to segment adjacent parts com-posed of the same material. Manual segmentation—such as outlining a target object on a single slice—can lead to contour misalignment across neighboring slices, despite the generally similar contour shapes within localized regions. To address this, we propose a deep learning-based active contour algorithm that propagates a manually segmented contour from a single slice to adjacent slices, correcting positional errors and aligning contours across cross sections. Experimental results demonstrate that our method reduces the need for manual segmentation to approximately one slice for every 20.

This paper proposes an algorithm that interpolates joint angles using polynomial equations after changing the IK (inverse kinematics) solution type for a 6-DOF manipulator, enabling smooth motion through the vicinity of the singularity. When passing through the singularity and its vicinity, some joints rotate rapidly. To suppress such rapid rotations, the algorithm switches between different IK solution types in the singularity. However, in the vicinity of the singularity, switching between different discontinuous IK solutions can still cause rapid joint rotations. To address this, the proposed algorithm interpolates joint angles using polynomial equations, thereby reducing rapid motion. In particular, quintic polynomials are used to ensure smooth angular acceleration. Although polynomial interpolation introduces position and posture errors at the end-effector, these errors are compensated using a numerical approach based on the jacobian matrix. As a result, all joint rotations remain below the maximum rotational speed of the servomotors, allowing stable passage through the singularity and its vicinity. This algorithm was verified on a large 6-DOF industrial robot. As a practical application, singularity machining was demonstrated using a robotic machining system equipped with a spindle motor attached to the end-effector. By applying the proposed algorithm, machining over a wide area including singularity and its vicinity was successfully achieved.

This study aims to prevent accidents in parking lots caused by sudden pedestrian or vehicle movements emerging from blind spots, through the development of an object tracking algorithm that utilizes fixed-point LiDAR. Unlike conventional RGB camera-based systems, which are vulnerable to low-light conditions and raise privacy concerns, LiDAR provides reliable distance measurements regardless of ambient lighting and enables object detection and tracking without capturing personally identifiable information. To this end, we propose a handcrafted method that combines DBSCAN, Random Forest, and ByteTrack, with an emphasis on practical applicability through adaptability to diverse environments, improved detection accuracy, and high processing speed. We evaluated the method across six different scenes, demonstrating that it achieves both reliable detection performance and high processing speed. Notably, when the number of points per object was below 4000, the algorithm consistently achieved processing speeds exceeding 10 FPS. These results indicate that the proposed approach is a viable and effective solution for real-time object tracking using fixed-point LiDAR in a variety of real-world scenarios, including nighttime environments.

Some measurement devices are commercially available for evaluating the dynamic behavior of the feed drive system of NC machine tools. However, most of these measurement devices have limitations in terms of measurement operation and the number of displacements that can be measured at one time, which hinders effective evaluation. In this study, a new prototype measurement device has been developed that can simultaneously evaluate displacements in some directions, and that does not change measurement sensitivity depending on the measurement position, in response to three-dimensional feed axes motion. The prototype measurement device has two linear guides arranged in a V shape and a single linear encoder in the detection section. An evaluation experiment was conducted by circular tests, using the prototype measuring device and DBB. As a result, the prototype measurement device has been able to measure the backlash and quadrant glitches of the X, Y, and Z axes in circular tests on the XY and ZX planes.

In this study, we developed an automated classification method for cell nuclei to facilitate the analysis of three-dimensional myocardial tissue microscopy images, a process that typically demands substantial time and manual effort. The proposed method takes as input z-stack images acquired using confocal microscopy and classifies cell nuclei into three categories—cardiomyocyte nuclei, vascular endothelial cell nuclei, and other cell types—by computing features related to nuclear morphology and physiological characteristics. Two datasets were prepared for evaluation, and the validity of the classification results was assessed not only in terms of accuracy but also with respect to cell type composition and mitotic activity. The experimental results showed that the proposed method achieved high classification scores for both cardiomyocyte and vascular endothelial cell nuclei. Furthermore, the estimated cell type proportions and levels of mitotic activity were found to be largely consistent with previously reported findings in the literature. These results suggest that the proposed method is effective for cell nucleus classification.

This study proposes a method to mitigate property damage accidents during pallet approach (reach) operations of forklifts in the logistics, transportation, and manufacturing industries. Accurate monitoring and control of the reach cylinder is crucial for accident prevention; however, industrial sites often require add-on solutions due to restrictions on modifying existing equipment. Therefore, this study developed a novel approach using non-contact Time of Flight (ToF) sensors. While ToF sensors enable high-precision distance measurement through light round-trip time, they present an inherent challenge for velocity estimation: measurement noise is significantly amplified by numerical differentiation, substantially degrading estimation accuracy. To address this, this study derived a novel velocity estimation algorithm. The proposed approach then applies the resulting velocity data to a reaction force estimation observer, establishing a method for high-precision, real-time estimation of reach cylinder loads. Experimental validation on actual equipment demonstrated the effectiveness of our method and confirmed its applicability as a practical solution for preventing forklift accidents.

This paper reports the particle excitation flow control valve using flexural vibration mode. The authors have been developing a compact, light-weight air flow control valve, which can control air flow continuously. This valve controls airflow using particle excitation caused by piezoelectric vibration and the vibration acceleration determines the flow characteristics. In order to reduce the vibration acceleration required for flow control of this control valve mechanism, we proposed a driving principle that uses torsional vibration as a vibration perpendicular to air pressure. However, in the method, the orifice did not open completely, so the controllable flow rate remained low. Therefore, to expand the control flow, we propose a control valve that uses flexural vibration, which is orthogonal to air pressure. In this paper, we present a prototype vibrator that generates flexural vibration mode, and demonstrate the effectiveness of the proposed valve control method through experimentation.

Trajectory errors in five-axis controlled machining are generally reduced by interpolation methods such as linearization and tool center point (TCP) control. These methods insert interpolation points along the trajectory of the command point at the tool tip and calculate the tool axis vectors at each interpolation point by linear interpolation of two rotation axes, without referencing the machining surface. However, in swarf machining, where the side of a straight tool is in linear contact with the workpiece, the tool axis vectors must be determined with reference to the machining surface. Conventional interpolation methods do not guarantee the tool side's trajectory, resulting in insufficient machining accuracy. This study proposes a high-precision tool path interpolation method for five-axis controlled swarf machining on ruled surfaces using a straight tool. The proposed method interpolates both the command point and the tool axis vector by referencing the trajectories of two points at either end of the line where the tool contacts the workpiece, thus achieving a highly accurate tool path. To evaluate the effectiveness of the proposed method, machining simulation and actual machining experiments were conducted and compared with conventional linearization and TCP control methods. As a result, the conventional methods caused a maximum overcut of 34μm and undercut of 32μm in areas where the rotary axes commands changed significantly and were far from the command point located at the tool tip. In contrast, the proposed method achieved good machining accuracy within the specified tolerance limits.

This paper presents a hyper-heuristics framework for the Online Order Batching and Sequencing Problem (OOBSP) in automated logistics warehouses that receive orders in real time and are equipped with Goods-to-Person (GtP) and Person-to-Goods (PtG) picking systems, followed by order consolidation. Two novel heuristics are introduced to capture the dynamics of the order consolidation process that significantly affect order completion time. These heuristics are combined with five conventional ones in a unified hyper-heuristics framework, where the weight coefficients of candidate heuristics are automatically adjusted to minimize makespan and total shipment delay. To determine these coefficients, two schemes are investigated: a static hyper-heuristics, in which fixed weights are optimized offline, and an adaptive hyper-heuristics, in which weights are determined online by a policy trained via reinforcement learning. Discrete simulation experiments using realistic order streams show that adaptive hyper-heuristics reduce makespan by 2.49% and shipment delay by 82.93% compared to a First-Come, First-Served baseline. Furthermore, the analysis demonstrates the effectiveness of the two proposed heuristics and confirms that the adaptive scheme outperforms the static scheme by dynamically tuning weights according to work progress and line status.

Sapphire is difficult to process due to its high hardness. Generally, grooves are processed by grinding, cutting, laser processing or etching. However, grinding and cutting cause severe tool wear and increase processing costs. In addition, laser processing requires expensive equipment, and etching has problems such as low material removal per unit time and long processing time. Although sapphire is chemically stable, it is known that a chemical reaction occurs at the interface with SiO₂ under high pressure and high temperature. The result of the reaction, double oxide composed of Al₂O₃ and SiO₂ is produced. Therefore, in this paper, we devised a method for processing into a narrow groove of sapphire by processing the side surface of a glass disk into a convex shape and cutting this convex part with sapphire. As a result, narrow grooves were processed into the sapphire surface. The effects of processing conditions such as processing time, circumferential speed, feed rate, and rake angle on the processed height, processed depth, and processed width were investigated for the obtained narrow grooves. Processed surface became a mirror surface. Also, processed width became the width of the glass protrusion. The processed depth and groove length could be changed using the rake angle and processing time.