Nikon (Japan)

Robotic systems operating in real-world environments often suffer from concept shift, where the input-output relationship changes due to latent environmental factors that are not directly observable. Conventional adaptation methods update model parameters, which may cause catastrophic forgetting and incur high computational cost. This paper proposes a latent Trend ID-based framework for few-shot adaptation in non-stationary environments. Instead of modifying model weights, a low-dimensional environmental state, referred to as the Trend ID, is estimated via backpropagation while the model parameters remain fixed. To prevent overfitting caused by per-sample latent variables, we introduce temporal regularization and a state transition model that enforces smooth evolution of the latent space. Experiments on a quantitative food grasping task demonstrate that the learned Trend IDs are distributed across distinct regions of the latent space with temporally consistent trajectories, and that few-shot adaptation to unseen environments is achieved without modifying model parameters. The proposed framework provides a scalable and interpretable solution for robotics applications operating across diverse and evolving environments.

Feedback controller design for multivariable motion systems presents a challenge due to the numerous tuning parameters and complex coupling dynamics. The aim of this paper is to develop a data-driven design procedure for decentralized feedback controllers in multi-axis motion systems. The decentralized feedback controller in each axis is constructed in an add-on structure, and the parameters are designed by iterative convex optimization using frequency response data. The developed approach enables interpretability in the optimized feedback controllers due to their structured representation, and the coupling dynamics between each axis can be considered by the data-driven optimization. The performance improvement is validated through numerical simulation based on experimental data from a 3-DOF high-precision motion system.

This report presents the optical performance and environmental durability of a radiation-resistant polarization-maintaining Er/Yb co-doped fiber (EYDF) developed by Fujikura Ltd. for space communications. The EYDF features a double cladding structure with a PANDA (polarization-maintaining and absorption-reducing) design, employing germanium doping for radiation resistance and phosphorus co-doping to improve Yb-to-Er energy transfer efficiency. To verify the reliable operation of this EYDF as a high-power optical amplifier (HPOA) in space environments, Nikon Corp. assembled and evaluated an amplifier incorporating this EYDF. The HPOA achieved an output power exceeding 30 W with ratios of signal to amplified spontaneous emission exceeding 37 dB and demonstrated stable operation for over 400 h. The polarization extinction ratio consistently exceeded 17 dB during continuous 10 W operation under temperature cycling from 25 °C to 45 °C, demonstrating robustness against temperature variations. High-temperature durability tests caused no significant changes in optical characteristics. These results demonstrate that the radiation-resistant EYDF is suitable for space-based optical communication systems.

Volume Correlative Light and Electron Microscopy (vCLEM) is a powerful method for assessing the ultrastructure of molecularly defined subcellular domains. A central challenge in vCLEM has been the efficient navigation of Regions of Interest (ROIs) across multimodal and multiscale imaging datasets. We developed two key tools to overcome this challenge. First, we developed a multimodal image registration tool (SegReg) that utilizes segmentation of common objects across modalities and uses a Graphical Processing Unit (GPU) for registration of imaging datasets in two and three dimensions. Secondly, we developed a dedicated image viewer to visualize multimodal image registration in three dimensions (NavROI). Here, we demonstrate the integrated use of SegReg and NavROI to navigate large mouse tissue blocks with preserved fluorescent signals to allow selective targeting for TEM tomography of ROIs containing synapses and the cisternal organelle on the proximal region of the axon of a selected pyramidal neuron. By providing real time guidance to precise X‐Y trimming of selected ROIs, reliable estimates of cutting depth relative to ROIs and a clear visual navigation of multimodal and multiscale images, our integrated workflow significantly improves the efficiency and accessibility of vCLEM analysis.

Digital Scanner (DS), a DUV optical maskless exposure tool, is being developed which has a fine resolution of 130 nm or smaller of L/S half-pitch and patterning flexibility that each chip design can be customized on wafer. In this paper, application of DS for Die to Wafer (D2W) bonding is studied to achieve fine pitch of bonding pads. DS exposure on carrier wafer before bonding can move the pads’ position to the ideal position by changing exposure pattern adjusted for each die placement error. We estimate the minimum bonding pad pitch at die placement accuracy of 200 nm is about 2 μm without DS while it can be down to 0.85 μm with DS248, increasing interconnect density by 5.8 times. Conceptual exposure experiment of placement compensation was performed assuming 5 × 5 dies that each die includes 8333 × 8333 pads with 1.2 μm pitch. The exposure experiment shows DS capability of improvement of pad placement accuracy of 50 nm.

The use of high-strength concrete in construction is expected to increase, and the manufacturing technology employed to produce such concrete needs to adapt to this rise. High-strength concrete has a workability that can vary significantly depending on the mixing conditions, even with the same mix design and in the same environment. This research focuses on the change in workability due to dry-mixing (i.e., mixing of fine aggregate with cement in the early stage) in the mixing process of a high-strength concrete mix using a revolving-double paddle mixer. As a result, it was confirmed that the slump flow value difference caused by dry-mixing was about 25 cm. Furthermore, Cryo-SEM (which can observe the specimens in the frozen state) image analysis revealed that dry-mixing causes flocculation of cement particles. The amount of admixtures adsorbed was then analyzed. This study concluded that flocculation affects the degree and timing of admixture adsorption in the later stages of the mixing process, leading to differences in concrete workability.

Abstract The metal additive manufacturing (AM) process allows for the production of three-dimensional (3D) parts with unique geometries that could not be formed by conventional methods. However, our aim is to utilize the AM process to create never-before-seen new materials because we believe it holds potential not only for new processing but also for creating a new material. The TiB2-Fe composite powder developed as a new raw material of the AM process, which was used as an additive powder in additive manufactured (AMed) SUS316L for effectively improving the mechanical properties of AMed SUS316L. The TiB2-reinfoced SUS316L synthesized by a Powder Bed Fusion (PBF) method using the TiB2-Fe composite powder had an extremely high Vickers hardness and strength of about 400 HV and 1250 MPa, respectively, being close to twice those of AMed SUS316L.