Keywords: Integrative Biophysical Engineering and Informatics, Biological and Physiological Engineering
Abstract: Modern life science increasingly demands tools that can observe living systems without perturbing them. While single-cell RNA-seq and related profiling assays have transformed our understanding of cellular properties, regulation, and function, these methods are inherently destructive, sacrificing the very cells they aim to study. This fundamental limitation prevents us from tracking how individual cells change over time in their native context. To address these challenges, we have been developing AI frameworks that bridge diverse imaging modalities and genomic-scale molecular profiling. Our Raman2RNA (R2R) platform combines Raman microscopy with machine learning to infer single-cell expression profiles entirely non-destructively in living cells, while our SCHAF (Single-Cell omics from Histology Analysis Framework) platform uses vision transformers and adversarial deep learning to generate spatially resolved single-cell transcriptomic data directly from standard H&E histology images. Together, R2R and SCHAF illustrate a broader principle that AI can learn to translate rich but molecularly uninterpretable image data into genomic-scale readouts, whether the input is a Raman spectrum from a live cell or a histological stain from an archived tissue section.In this talk, I will also present our latest advances in scaling R2R to tissues through a new imaging modality, total internal reflection Raman microscopy (TIRR), which captures Raman photons in a massively parallel manner without requiring high-power lasers and achieves dramatic speed improvements over conventional confocal Raman microscopes.

Keywords: Signal and/or Image Processing, Robotic and Automation Systems
Abstract: This study proposes a mobile robot based on SLAM that is equipped with a microphone array to locate sound sources, such as human voices, during disasters. The experimental results showed that the robot reached the sound source which was not visible from the starting point by utilizing sound localization and an estimated map of the environment.

Keywords: Signal and/or Image Processing, Robotic and Automation Systems, Safety, Environment and Eco-Systems
Abstract: Accurate detection of landfill gas (LFG) monitoring pipes in unstructured outdoor environments is a critical prerequisite for autonomous inspection systems. This paper presents a multi-sensor guiding framework for autonomous all-terrain vehicle (ATV) inspection of LFG monitoring pipes, targeting three-dimensional position and orientation estimation for precise ATV navigation and stopping. The core contribution is the vision-based semantic detection module used during the approach phase, where the pipe is visible in fisheye images but cannot yet be reliably reconstructed from sparse LiDAR returns. The module employs YOLOE with a Semantic Proxy Mapping strategy to identify LFG pipes from onboard fisheye camera images. Instead of training a pipe-specific YOLOv8 detector from limited annotated images, the proposed method uses reference pipe images as visual prompts and detects geometrically similar pole-like structures without task-specific retraining, thereby mitigating the overfitting risk observed in the prior YOLOv8-based method. Field experiments using an ATV platform demonstrate successful pipe detection in 766 of 834 pipe-containing frames (91.8%), with 5 false-positive frames and an inference time of 12.2 ms per frame on an NVIDIA RTX A5000 GPU.

Keywords: Signal and/or Image Processing, Sensors and Transducers, Identification and Estimation
Abstract: Traditional marine engineering primarily relies on conventional contact-based diagnostic techniques for engine monitoring; however, the transition to unmanned surface vehicles (USVs) equipped with submerged electrical thrusters necessitates a digital evolution. This study proposes the use of hydrophones as an alternative to conventional stethoscope-based monitoring for underwater propulsion systems. The primary challenge addressed is the accurate classification of induced faults within “semi-open” basin environments, where hydroacoustic interference from marine life and maritime activities introduces pronounced signal noise. To address this, a deep learning framework based on convolutional neural networks (CNNs) was implemented, utilizing spectrogram representations of hydroacoustic signals. By capturing the interplay between time, frequency, and spectral energy, the model preserves critical contextual information associated with thruster dynamics. To mitigate shortcut learning caused by environmental noise artifacts, comprehensive data augmentation and noise reduction techniques were applied, ensuring robust and meaningful feature extraction. Experimental results demonstrate strong robustness across four model configurations, with the optimized architecture achieving a mean classification accuracy of 99.69% across five validation folds. Evaluation metrics further confirm high true positive rates and low false positive rates, indicating reliable diagnostic performance.

Keywords: Signal and/or Image Processing, Sensors and Transducers, Analytical Measurement
Abstract: This study utilizes the ”Source Sensitive Filter (SSF)” to detect and visualize two-dimensional resonant wave fields for frequency-independent defect localization. Focusing on standing waves generated inside defects, numerical and acoustic experiments on cylindrical models confirmed that the SSF successfully characterizes these fields by detecting standing wave nodes. This approach enables intuitive defect identification without requiring specialized expertise.

Keywords: Signal and/or Image Processing
Abstract: In this study, with focusing on a fault diagnosis on minor damaged rolling bearings, we try to introduce signal processing methods for detecting occurrence of non-stationary damping-oscillation waveforms generated during an abnormal operation and for quantitatively estimating the deterioration level based on waveform shapes. First, for an occurrence detection, we consider the non-stationary feature of the oscillation waveforms. Based on the spectral correlation method which allows for a separate evaluation of the occurrence frequency and the oscillation frequency, we develop novel techniques to improve detection reliability by weighting thorough use of the band-selective filter bank processing. Furthermore, for deterioration estimation, to advance an estimation performance in a low S/N environments, we introduce a filtering technique by applying the spectral differences. We also propose a novel method for a quantitative deterioration estimation based on the transfer function model by separating the non-stationary damping-oscillation waveforms into the transfer function component and the periodic component through the cepstrum analysis.

Keywords: Signal and/or Image Processing, System Engineering, Intelligent Systems
Abstract: This study focuses on excessive armpit opening during weapon manipulation in foil fencing and investigates whether this local form error can be evaluated frame by frame using IMU-based skeleton data. Human skeleton data were captured with Mocopi during practice matches and were labeled according to the extent of armpit opening. To incorporate temporal context, we compared a single-frame baseline method with a temporal window-based method using statistical features extracted from skeleton data. The results showed that using selected joints around the shoulder was more effective than using all joint features for this task. While the baseline method showed significant performance drops in some matches, the temporal window-based method reduced such degradation and contributed to more stable recognition.

Keywords: Signal and/or Image Processing, Medical and Welfare Systems, Integrative Biophysical Engineering and Informatics
Abstract: This study proposes a subcarrier selection method for Wi-Fi channel state information-based heart rate variability estimation. Building on a conventional heart rate variability extraction framework using wavelet decomposition and variational mode decomposition, the proposed method revisits the subcarrier selection process by combining two complementary criteria: the heart-to-noise ratio in the frequency domain and autocorrelation function variance in the time domain. Only subcarriers satisfying both criteria are selected, and the extracted signals are further processed by principal component analysis, wavelet decomposition, and variational mode decomposition to estimate R--R intervals. Experimental results for five evaluation subjects showed that the proposed method achieved lower root mean square error and variance than the heart-to-noise-ratio-only and autocorrelation-function-only methods in many cases. The average root mean square error of the proposed method was 53.9 ms at 3 m and 59.5 ms at 4 m. These results demonstrate that combining frequency-domain and time-domain criteria is effective for robust subcarrier selection in Wi-Fi channel state information-based heart rate variability estimation.

Keywords: Signal and/or Image Processing, Temperature Measurement, Intelligent Systems
Abstract: This paper proposes a method for automatically detecting abnormal overheating in power receiving equipment in real time using thermal image data captured by a thermal imaging camera. The thermal image data is processed through blue-regions extraction, application of HOG features, and PCA, and then analyzed using a One-Class SVM of an unsupervised learning method to detect abnormal overheating. The effectiveness of the proposed method is verified through experimental results.

Keywords: Sensors and Transducers
Abstract: High-temperature strain measurement is critical for extreme environments, but existing ITO gauges suffer drift and poor stability. We fabricated ITO gauges via magnetron sputtering; 800  °C heat treatment relieves residual stress, yielding a drift rate of 0.0496  h-1 at 800  °C and gauge factors of −2.95 (at RT), −1.85 (at 800  °C), and −0.97 (at 1000  °C), improving high-temperature stability and strain-sensing performance. This study provides technical support for high-temperature strain measurement of engine blades.

Keywords: Sensors and Transducers, Modeling, System Identification and Estimation, Flow Measurement and Control
Abstract: This study proposes a method for designing acoustic streaming, which is a kind of airflow, with high spatial degrees of freedom by designing its driving force field using airborne ultrasound phased arrays. Various approaches to control airflow have already been proposed. For example, fan-based flow control has been used in fuel cell air-cooling systems that supply hydrogen while cooling the cell, and plasma actuators have been used to control airflow to prevent flow separation on aircraft wings. In this work, we propose designing streamline shape by generating the driving force of acoustic streaming (airflow) remotely in midair with ultrasound phased arrays. Specifically, instead of directly designing target streamlines, we design the driving force field that generates them. We visualize the driving force field numerically and verify through physical experiments that the resulting acoustic streaming follows the desired streamline shape. Since the acoustic intensity that drives the acoustic streaming can be derived analytically and is described by the linear term obtained from a perturbation expansion of the Navier-Stokes equations, the relation between input and output can be expressed with simple matrix operations. Furthermore, as the degrees of freedom of the sound field that produces the driving force field are on the order of the number of ultrasound transducers in the phased array, the proposed method enables highly flexible shape design of airflow.

Keywords: Sensors and Transducers, Remote Sensing, Identification and Estimation
Abstract: LiDAR-based 3D object detection is a fundamental component of environment perception in autonomous driving. In particular, highly accurate boundary prediction of surrounding objects is essential for safe trajectory planning. However, mainstream 3D object detection models estimate object boundaries by regressing from the center coordinates, which often leads to boundary misalignment. While predicting the nearest corner can resolve this, its application to practical single-stage anchor-based models has not been sufficiently explored. This paper proposes a novel corner regression framework integrated into a single-stage anchor-based model. To prevent training instability, we introduce uncertainty (variance) learning to dynamically attenuate the loss penalty, combined with soft-transition curriculum learning. Evaluated on a real-world dataset collected on Japanese public roads, our method maintains standard recognition accuracy (AP) while reducing the maximum regression error (P90) of the nearest corner point by 0.089m for large vehicles, thereby achieving strict boundary prediction suitable for safe trajectory planning.

Keywords: Opto-Electronic Measurement, Signal and/or Image Processing, System Integration
Abstract: Optically coaxial cameras with different wavelengths or frame rates are valuable for multispectral and high-speed imaging. Optical axis alignment of the cameras is often performed using homography-based planar transformations; however, this approach is limited by the calibration range and cannot achieve accurate alignment across a large depth range. We propose an interactive alignment method that visualizes optical axis misalignment using an infinity mirror structure with a mirror, a beamsplitter, and circularly arranged dot markers. Rotation alignment is performed using a vanishing point at infinity, estimated from reflected markers, while translation alignment is guided by the markers at a finite distance, enabling high-precision alignment. Experiments have demonstrated reduced vanishing point estimation error under mount adjustments, robustness to positional variations, and consistently high accuracy across trials. We have also confirmed effective alignment beyond the calibration range in a high-frame-rate imaging demonstration.

Keywords: Opto-Electronic Measurement, Signal and/or Image Processing, System Integration
Abstract: Quantitative diagnosis in endoscopic examinations of tracheomalacia requires measurement of dynamic changes in the three-dimensional geometry of the tracheal inner wall, including respiratory variations and cardiogenic oscillations. However, in tracheal endoscopy with diameters below 10 mm, constraints on the form factor of measurement devices are severe, and sufficient three-dimensional dynamic measurement is difficult with endoscopes of only a few millimeters in diameter. Therefore, we propose an optical design that enables high-speed scanning of the cone beam angle derived from a multimode optical fiber via skew reflection, using a high-speed rotating mirror and an elliptical mirror, toward fast three-dimensional measurement with a thin endoscope. Experimentally, we confirmed sufficiently high-speed cone beam scanning compared with conventional methods, as well as the performance of three-dimensional shape measurement using a probe with a maximum diameter of 3.25 mm.

Keywords: Networked Sensor System, Discrete Event Systems, Network System Integration
Abstract: This paper studies distributed time synchronization of atomic clock networks with anchoring to UTC (NICT) via GNSS receivers. To reduce communication cost, we propose an impulse-type event-triggered control scheme for anchoring inputs, where control signals are applied only at triggering instants. We show that standard event-triggered formulations cannot be directly modified into an impulse-type scheme, and introduce a modified triggering strategy with an additional threshold parameter. Numerical simulations demonstrate that the proposed method significantly reduces transmission frequency while maintaining acceptable synchronization performance, revealing a trade-off between communication cost and performance.

Keywords: Temperature Measurement, Flow Measurement and Control, Networked Sensor System
Abstract: This paper presents a measurement-driven study of the coupled dynamics between sea surface temperature (SST) and atmospheric carbon dioxide (CO₂) concentration in a coastal environment using a low-cost, network-based sensing system. A distributed sensor network based on LPWA communication (LoRa/LoRaWAN) was developed to continuously acquire high–time-resolution environmental data, including SST and CO₂ concentration, at intervals of 30 seconds to several minutes. Continuous observations were conducted at a coastal site and compared with data from inland and urban locations. The results reveal previously unobserved short-term characteristics, including (i) larger diurnal SST variations in winter than in summer in shallow coastal waters, and (ii) a hierarchical structure in CO₂ variability, where diurnal patterns are governed by ecosystem and boundary-layer processes, while baseline levels depend on local geographical and emission conditions. Furthermore, the observations suggest a possible causal relationship in which increases in SST lead to enhanced CO₂ release from seawater due to reduced solubility and increased biological activity. These findings indicate that coastal air–sea interactions may play a buffering role in short-term CO₂ variability. The study demonstrates that low-cost sensor networks enable high-resolution environmental monitoring that can reveal fine-scale dynamics not captured by conventional observation systems. The proposed a

Keywords: Temperature Measurement
Abstract: Thermocouples are known to exhibit degradation in stability and durability as the temperature under measurement increases. For ultra-high temperature measurements, advanced thermocouple development is essential, and accurate knowledge of their Seebeck coefficient, from which the performance of the developed thermocouples can be characterized, is very fundamental. In this study, the Seebeck coefficient of a type B thermocouple wire was evaluated over the temperature range from 1000 °C to 1500 °C for verifying the validity of a Seebeck-coefficient measurement system, specifically developed for short-length thermocouple wires for high-temperature measurements.

Keywords: Analytical Measurement, Signal and/or Image Processing, Entertainment Systems
Abstract: This paper investigates how humans and facial expression recognition (FER) models interpret ambiguous facial expressions through a comparative analysis of Thai and Japanese participants. Participants evaluated facial expression images under two conditions: without context and with context, where they selected a perceived emotion and provided a confidence rating for each image. Multiple FER models were also applied to the same set of images for comparison. The analysis focuses on response distributions, agreement, entropy, and variability in interpretation across human participants and machine models. The results show that human participants tend to interpret ambiguous expressions through distributed responses across multiple related emotion categories, whereas FER models tend to assign more concentrated predictions, often simplifying ambiguous expressions into dominant categories such as Neutral. Differences in interpretation patterns were also observed between Thai and Japanese participants, particularly in the degree of response variability for ambiguous expressions. In addition, contextual information generally reduced ambiguity and increased agreement across participant groups. These findings suggest that ambiguous facial expressions are not perceived as fixed emotion categories, but rather as expressions open to multiple interpretations influenced by context and participant group.

Keywords: Flow Measurement and Control, Standard of Measurement
Abstract: A flow measurement facility was constructed to measure the flow rate of solid particles from water-mixture in real time. A Coriolis flowmeter, capable of simultaneously measuring mass flowrate and fluid density, was applied. Equipped with a solid particle feeding and agitation mechanism allowing adjustment of the zeolite particle feed rate, the mixed fluid was allowed to free-fall from a height of 7 m. Measurements from the flowmeter were compared with those from the mass method. For solid particle mass ratios up to 10%, the solid particle flow rate based on Coriolis flowmeter measurements showed approximately 70% of the mass method results. This discrepancy was attributed to measurement differences in the density difference between the mixture fluid and pure water.

Keywords: Remote Sensing, Ubiquitous Healthcare, Components and Devices
Abstract: Fall detection is an important safety technology for elderly people living alone. However, wearable sensors burden users, and camera-based systems raise privacy concerns. This study presents a contactless fall detection method using Wi-Fi Channel State Information (CSI). The method reproduces an image-based amplitude baseline and extends it by introducing phase-difference features that are more robust than raw phase. Amplitude maps are processed by EfficientNet, while phase-difference maps are processed by a lightweight convolutional branch and fused for four-class activity recognition (walk, fall, sit, and stand). Two datasets were collected: a laboratory dataset acquired with Intel 5300 NICs under a clear line-of-sight condition and a residential-room dataset acquired with low-cost ESP32 devices. In the laboratory environment, the proposed method improved group-split accuracy from 0.6328 to 0.7152 for four-class classification and from 81.92% to 89.09% for binary fall/non-fall detection. In the residential environment, the proposed method achieved 0.8113 accuracy at a 2.5-cm device height and 0.7222 at 80 cm under group-split evaluation, outperforming the amplitude-only baseline at both heights. The results show that phase-difference features improve generalization across trials and placement conditions, and that CSI normalization is essential when using ESP32 devices.

Keywords: Mechanical Measurement, Signal and/or Image Processing, Sensors and Transducers
Abstract: This study aimed to detect texture-related information with high sensitivity using a gel sensor. By analyzing response waveforms during the structural collapse of food, we confirmed the potential for texture identification. Results showed that the sensor can quantitatively capture characteristic stress waveforms that reflect food hardness and vibration. This indicates that stress waveforms during food compression can be quantitatively measured, highlighting the potential for texture discrimination based on these waveforms.

Keywords: Human Interfaces, Medical and Welfare Systems
Abstract: Low back pain is highly prevalent among caregivers due to repetitive patient handling tasks that impose substantial mechanical loads on the lumbar spine. Despite the introduction of assistive devices and no-lift care practices, the incidence of back pain remains high, particularly among novice caregivers. In this study, we aim to reduce the injury risk to caregivers by improving caregiving skills using feedback. To this end, we developed a two-dimensional lumbar spine load estimation method for caregiving motions based on musculoskeletal model and confirmed the basic performance with able-bodied persons. The musculoskeletal model grounded in statics was employed to estimate lumbar compression and shear forces at the L5/S1 joint during patient transfer tasks. The model incorporates body segment parameters, external loads, and intra-abdominal pressure to calculate internal forces that cannot be measured directly. Motion data were captured using an optical motion capture system, and caregiving motions were classified as proper or improper based on biomechanical principles. Dynamic Time Warping (DTW) was applied to time-series data of estimated lumbar loads to evaluate motion similarity and perform classification. The results of experiments involving five healthy male participants showed that classification based on estimated two-dimensional forces achieved higher accuracy compared to conventional posture-based indicators. The classification accuracy based on lumbar load reached

Keywords: Human Interfaces, Signal and/or Image Processing
Abstract: Hearable devices have become versatile sensing devices in recent years, and now capable of incorporating small cameras. In this research, we designed a new interaction method that leverages mouth and tongue movements via a hearable device with a small camera and an IMU, investigating gestures that can be utilized around the mouth. We collected images of the side face and head movement data through the device to classify mouth gestures. Based on previous elicitation research, we generated 23 candidate gestures. We conducted data analyses and a usability study, resulting in a final set of seven gestures, such as tapping the tongue and smiling. An offline evaluation showed that the mean accuracy of gesture recognition was approximately 30–50% among the participant-dependent models. The results also indicated that gestures that can be performed with the mouth closed are difficult to recognize solely based on images.

Keywords: Human Interfaces, Medical and Welfare Systems, Human-Machine Systems
Abstract: This paper considers to develop a lower-limb device designed to improve the usability for physical therapists in tele-rehabilitation using electrical stimulation. First, the configuration of the lower-limb device is indicated. The lower-limb device consists of a leg torso, motors, and various gears, and is capable of replicating the movements of the patient's knee joint. In addition, strain gauges are used as torque sensors to measure the force applied by physical therapists. Next, the gain of the lower-limb device controller is designed using a control system similar to that of conventional paddle-type devices. Finally, the subjects operate both the proposed lower-limb device and the conventional paddle-type device. We will then conduct a survey to evaluate which one is easier to use and verify its effectiveness.

Keywords: Human Interfaces, System Integration, Biological and Physiological Engineering
Abstract: An sEMG is a biological signal generated during muscle contraction. It can be measured non-invasively from the skin surface and is correlated with muscle force. Force can be estimated using only sEMG sensors attached to the surface. The relationship exhibits nonlinear and dynamic characteristics, and long short-term memory (LSTM) networks have been widely applied to estimate force from sEMG. However, the relationship between sEMG and force varies across experimental trials. In particular, LSTM have been widely applied to model this relationship in conventional studies. However, such models are essentially black-box models, making it difficult to interpret them from physical and mathematical perspectives. Therefore, this study proposes an element description method (EDM). This method automatically generates a block diagram and enables physical and mathematical interpretability of the model. Although the model structure can be defined, parameter calibration is required based on experimental data. Therefore, this paper proposes a force estimation method that combines the EDM with LSTM. The proposed method generates a base model for force estimation using the EDM. In addition, the LSTM is used as a supplementary component that learns the residual between the EDM-based estimation and the target force. As a result, the proposed method achieved force estimation performance comparable to or better than that of conventional methods.

Keywords: Human Interfaces, Sensors and Transducers, Biological and Physiological Engineering
Abstract: In baseball, it is often said that pitchers require fine tactile sensitivity at their fingertips to control breaking balls. Such claims extend beyond pitchers to ball sports and sports in general; however, there is little empirical evidence verifying differences in tactile sensitivity depending on sports experience. In this study, we examined whether the presence or absence of baseball experience, as well as differences between dominant and non-dominant hands, affect the ability to discriminate transiently presented tactile stimuli, using moving stimuli generated by an Airborne Ultrasound Tactile Display. We hypothesized that more extensive sports experience would lead to higher tactile sensitivity; however, the experimental results showed that the experienced group had significantly higher discrimination thresholds than the inexperienced group, indicating a tendency toward lower sensitivity to transient stimuli. Meanwhile, no difference in sensitivity was found between the dominant and non-dominant hands.

Keywords: Biological and Physiological Engineering, Human Interfaces
Abstract: Functional electrical stimulation (FES) is a technique used to induce muscle contraction for restoring motor functions. In FES-based joint angle control, antagonistic muscles are stimulated to generate joint motion, and their co-contraction contributes to joint stiffness and motion stabilization. However, increasing the co-contraction level also increases stimulation input, which may lead to muscle fatigue and reduced efficiency. In this paper, an evaluation-function-based calibration method for co-contraction level design is proposed. An evaluation function integrating tracking performance and stimulation input is introduced, and joint angle control experiments are conducted under multiple co-contraction conditions. Experimental results demonstrate that moderate co-contraction improves tracking performance while excessive co-contraction increases stimulation input. These results suggest that the proposed method can be utilized for experimental calibration of co-contraction levels in FES-based joint angle control systems.

Keywords: Biological and Physiological Engineering, Human Interfaces, Mechatronics Systems
Abstract: Functional Electrical Stimulation (FES) is a technique that induces muscle contraction by applying electrical stimulation to muscles and is widely used in the field of rehabilitation. In recent years, its application has been extended to assist activities of daily living and skill transfer. However, studies on its tracking performance for high-speed motions remain insufficient. In particular, performance degradation becomes a critical issue when dealing with movements characterized by high angular frequencies. In this study, a proportional derivative (PD) control combined with a disturbance observer (DOB) is adopted as the conventional method, and a novel control approach incorporating pre-stimulation is proposed. In the experiments, co-contraction control of the wrist using FES is performed, and the tracking performance is evaluated against sinusoidal inputs with different angular frequencies. This study shows that the proposed method with pre-stimulation can effectively enhance tracking performance within a theoretically relevant angular frequency range. At the same time, they suggest that, in the high frequency domain, a comprehensive evaluation considering not only RMSE but also amplitude attenuation and phase lag is essential.

Keywords: Biological and Physiological Engineering, Human Interfaces, Mechatronics Systems
Abstract: Functional electrical stimulation (FES) is a promising approach for assisting upper-limb function without large external hardware. However, stimulation parameter tuning still depends strongly on trial and error because the relationship between stimulation input and generated torque is nonlinear. This study investigates a static model of steady-state wrist flexion torque under noninvasive FES using pulse width, current amplitude, and joint angle. Wrist torque was measured under isometric conditions at multiple wrist angles and current levels while pulse width was increased stepwise. Based on the idea that pulse width and current amplitude represent different stimulation axes, an interaction-based model was introduced. The proposed model explicitly includes pulse width, current amplitude, and their interaction term at each joint angle. The model was evaluated by leave-one-current-out validation and compared with a PW-based reference model and a neural-network model. The results show that the proposed model can reproduce the measured steady-state torque response and can generalize to unseen current conditions while preserving a simple and interpretable structure.

Keywords: Integrative Biophysical Engineering and Informatics, Human-Machine Systems, Nonlinear Control
Abstract: Human eye–hand coordination relies on internal forward models that predict future states and compensate for sensory delays. During line tracing, the gaze typically leads the hand through predictive saccades, yet the extent to which this predictive window reflects expertise or intrinsic individual traits remains unclear. In this study, I examined eye–hand coordination in professional calligraphers and non-experts performing a controlled line tracing task. The temporal coupling between saccade distance (SD) and pen speed (PS) revealed substantial interpersonal variability: SD–PS peak times ranged from approximately −50 to 400 ms, forming stable, participant-specific predictive windows consistent across trials. These windows closely matched each individual’s pen catch-up time, indicating that the oculomotor system stabilizes fixation in anticipation of the hand’s future velocity. Neither spatial indices nor the temporal index differed between calligraphers and non-calligraphers, and none correlated with tracing accuracy. These findings suggest that diverse predictive strategies achieve equivalent performance, consistent with the minimum intervention principle of optimal feedback control. Predictive timing in eye–hand coordination reflects a stable, idiosyncratic Predictive Protocol shaped by individual neuromotor constraints rather than expertise.

Keywords: Electrophysiology and Kinesiology, Biological and Physiological Engineering, Virtual Reality Systems
Abstract: Humans adjust their gait in anticipation of their surroundings while walking, but the details of these anticipatory movements that occur before obstacle avoidance have not been clarified. This study examines this anticipatory adjustment through motion analysis and quantitative assessment of gait phases prior to obstacle avoidance. Participants walked on a treadmill while wearing an MR headset and stepped over virtual obstacles that were presented at random timings during walking. Leg angles were calculated from the measured body-marker coordinates, and instantaneous phases of the gait were derived using the Hilbert transform with statistical correction. Comparison between steady-state walking and obstacle-avoidance walking revealed a consistent trend in phase difference beginning one gait cycle before obstacle avoidance, suggesting that gait rhythm is modulated during the predictive stage.

Keywords: Ubiquitous Healthcare, Biological and Physiological Engineering, Human Interfaces
Abstract: Recently, wearable sensors are attracting attention as a quantitative evaluation of gait kinematics in daily environments; however, applying them to multiple body parts may impose significant physical burdens on users. Although deep learning approaches can estimate posture from sparse sensor sets, sensor placement remains largely heuristic, leaving accuracy gains dependent solely on algorithmic refinements. To address this, this study proposes a wearable system with a sensor configuration optimized through geometric and dynamic analysis of a human multi-link model. This system obtains the minimum necessary information for estimation while minimizing the user's sensory load.

Keywords: Remote Sensing, Safety, Environment and Eco-Systems, Signal and/or Image Processing
Abstract: Road blockages caused by landslides and mudflows can obstruct the passage of emergency vehicles and delay rescue operations. To ensure the rapid restoration of blocked roads, it is essential to analyze the damage status of each road section and efficiently clear the roads. In this paper, we propose a method of detecting blocked roads using topographic data. First, we calculate the difference between pre- and post-disaster results from the Digital Elevation Model (DEM) generated from 3D point cloud data. Next, we integrate the results with road segment data and detect blocked roads within disaster-affected areas where sediment on the road affect vehicle traffic. Furthermore, by visualizing this information for each road segment, we can obtain detailed information. Our approach successfully extracts sediment-affected areas, demonstrating the effectiveness of integrating road and topographic data.

Keywords: Autonomous Decentralized Systems, Robotic and Automation Systems, Discrete Event Systems
Abstract: This paper presents a decentralized control system for vehicle platoons enabling cooperative lane changing. A control strategy based on mode transitions of driving states is developed, and cooperative lane changing using vehicle communication among CACC vehicles is implemented and experimentally validated using a real world platform. The experimental results demonstrate the effectiveness of the proposed approach in enhancing traffic flow efficiency.

Keywords: Transportation Systems, Simulation of Large Systems, Autonomous Decentralized Systems
Abstract: Vehicle platooning is a promising transportation technology that can improve fuel efficiency and traffic flow. For vehicles in a platoon to move together properly, they need wireless communication. However, if wireless messages between the vehicles are lost during a sudden braking, serious safety problems can happen.

In our earlier work, we introduced a rigorous simulation-based method to study these safety problems. This method looks at the minimum distance between vehicles over time while using control settings that keep the platoon string-stable. It calculates the exact distance at specific simulation times that are appropriately selected such that the distance is guaranteed to not change too much between those times. As a result, the simulation results closely reflect what happens in real-world traffic in continuous-time.

We now extend this method to evaluate fuel efficiency while providing guaranteed error bounds on the fuel savings achieved by each vehicle. Our approach incorporates both the platoon’s dynamic model and the aerodynamic drag experienced by each vehicle based on its position within the platoon. Through several case studies, we identify control settings that reveal a tradeoff between safety and average fuel savings.

Keywords: System Integration, Safety, Environment and Eco-Systems, Networked Sensor System
Abstract: In large-scale natural disasters such as earthquakes, victims may become trapped under collapsed buildings, and delays in locating them can significantly reduce their chances of survival. Therefore, a compact and practical system for estimating the positions of buried victims is required for rapid search-and-rescue operations. In this study, we develop and experimentally validate an LPWA-based IoT acoustic localization system for victims buried under rubble. The proposed system consists of two independently driven acoustic sources, a microphone-equipped sensing unit, and a 920 MHz LPWA communication module. In our previous prototype, the small speaker's output power was limited, resulting in a detectable range of approximately 8 m with a 3 W sound source. In the present study, we introduce a higher-power acoustic source with a maximum output of 50 W and evaluate its effectiveness in a simulated rubble environment. The experiments focus on acoustic signal detectability, direction, and angle estimation performance, and LPWA communication reliability under obstructed conditions. The results are expected to clarify the feasibility and limitations of the proposed system for disaster-response applications. In the future, the system will be further integrated with mobile robots, drones, and rescue vehicles to support efficient victim search operations in disaster-affected areas.

Keywords: Networked Sensor System, System Integration, Network System Integration
Abstract: Internet of things (IoT) sensors are increasingly being installed in various fields. On-demand networked sensing is well suited to battery-free outdoor environmental sensors powered by energy harvesting. Timing control of microwave wireless power transfer (WPT) enables on-demand activation of energy-harvesting sensors. Appropriate timing control requires the explicit feedback on the charge state of a built-in capacitor via wireless communication from the sensor to the WPT transmitter to optimize power transfer efficiency. However, the feedback information increases communication overhead and power consumption, and also limits flexibility in sensing system configuration. This study proposes an on-demand networked sensing system using energy-harvesting IoT sensors without explicit charge state feedback. The proposed system estimates power transfer efficiency based on the sensor communication interval, in which sensor data are transmitted to an edge server via an access point and only the information on data reception interval is shared with the WPT transmitter. First, we experimentally model the WPT characteristics and clarify the relationship between the communication interval and antenna beam direction. Then, the antenna beam direction is adjusted by using a basic hill-climbing algorithm based on the data reception interval to shorten the power transfer time. Simulations verify the feasibility of the proposed system.

Keywords: System Integration, Mechatronics Systems, Network System Integration
Abstract: Unmanned aerial vehicles (UAVs) with wireless networked detachable rotors (WNDRs) provide high flexibility and reconfigurability for various applications by eliminating physical wiring constraints. However, the system is highly susceptible to communication failures such as packet losses and bandwidth constraints because each rotor independently communicates with a ground-based controller via wireless networks. In particular, packet losses severely degrade the flight stability and control performance, often leading to collisions or crashes. This study demonstrates a highly reliable control system based on a hybrid packet-loss concealment method that compensates for probabilistic and burst losses. The proposed method dynamically switches between two packet-loss concealment methods, i.e., a hold-input method, which maintains the previous control value during probabilistic losses to ensure smooth operation, and a zero-input method, which sets attitude control values to zero during burst losses to prevent severe accidents. The switching function evaluates the packet-loss condition of each rotor and coordinates the utilization of the two packet-loss concealment methods to stabilize the attitude of the UAV with WNDRs. Simulations confirm that the proposed method maintains flight stability and control performance even under severe packet-loss conditions.

Keywords: System Integration, System Engineering
Abstract: The shoebill (Balaeniceps rex) inhabits tropical wetlands in Central and East Africa, where high temperatures and high absolute humidity are maintained year-round. Reproducing these environmental conditions in zoological parks is expected to contribute to animal welfare and support breeding. This study investigates an IoT-based environmental monitoring and wetland reproduction system for shoebills at Chiba Zoological Park. First, IoT sensor nodes are installed inside the current shoebill enclosure to continuously measure temperature, absolute humidity, and vapor pressure deficit (VPD). Particular attention is paid to reproducing the absolute humidity conditions observed in natural shoebill habitats. The measured environmental data are compared with climatic conditions in African wetlands. Based on these comparisons, environmental reproduction experiments using mist spraying systems and artificial rainfall equipment are conducted to investigate suitable environmental control methods. In addition, camera-based behavioral observations are performed to analyze feeding behavior, habitat utilization, and responses to environmental changes. The feasibility of papyrus cultivation and partial wetland reproduction inside the enclosure is also examined. The proposed system aims to establish a cyber-physical environmental framework for supporting shoebill welfare, habitat reproduction, and future breeding promotion in zoological parks.

Keywords: SCADA and Network Systems, Networked Sensor System
Abstract: This research focuses on the FPGA implementation and evaluation of ElGamal cryptographic processing for encrypted control systems. Homomorphic encryption schemes, including the ElGamal cryptosystem, allow control signals and parameters to be processed while remaining encrypted, which contributes to improving the confidentiality of networked control systems. In secure control applications, increasing the key length is important for enhancing cryptographic security; however, in the ElGamal cryptosystem, this increases the computational cost of modular multiplication and modular exponentiation significantly. To address this issue, this study aims to develop an implementation that exploits the hardware parallelism and low-latency execution capabilities of FPGAs to maintain real-time performance while supporting larger key lengths. The proposed implementation targets the ElGamal-based processing flow required for encrypted control, including data conversion for encryption, encryption, homomorphic computation on ciphertexts, decryption, and reconstruction of control values. The implementation will be evaluated by changing the key length and analyzing latency, throughput, and resource utilization to clarify the effectiveness of FPGA-based implementation.

Keywords: Transportation Systems, System Integration, Network System Integration
Abstract: Cooperative adaptive cruise control (CACC) is a vehicle control technique that enables efficient platooning by sharing control information between vehicles through wireless communication. Although CACC can improve traffic throughput and fuel efficiency, cyberattacks may degrade string stability and safety of vehicle platoons. In particular, false data injection (FDI) attacks on inter-vehicle distance (IVD) sensors can falsify sensor data and cause unsafe vehicle behavior. Our previous study focused on redundant IVD information obtained from both front-IVD and rear-IVD sensors installed on each vehicle. A tamper detection observer (TDO) is used to estimate control responses based on a vehicle model and detect tampered data by comparing the values measured by the sensors with those estimated by the TDO. However, the detection performance is degraded by communication time delays between vehicles. This study proposes an integrated tamper detector and delay compensator (ITDDC), which introduces a communication disturbance observer (CDOB) into the TDO to address the issue caused by time delays. Simulation results using a ten-vehicle platoon demonstrate that the proposed method can mitigate the impact of FDI attacks while maintaining platoon stability.

Keywords: Transportation Systems, System Integration, Network System Integration
Abstract: Cooperative adaptive cruise control (CACC) achieves stable platooning of connected vehicles by using acceleration information of preceding vehicles, obtained via vehicle-to-vehicle (V2V) communication. However, communication quality such as time delays and packet losses in V2V communication leads to string instability of the vehicle platoon. When the communication quality is degraded, CACC can be made the transition to adaptive cruise control (ACC) without using V2V communication. However, binary switching between CACC and ACC modes causes discontinuous acceleration input and chattering in the two control modes, which may deteriorate the stability and control performance. This study proposes a hybrid CACC and ACC platooning method, which combines CACC and ACC by applying weights based on communication quality to the control inputs calculated in the two controllers. The proposed method evaluates the impact of communication degradation on platoon behavior as a disturbance, and weights the control inputs based on the disturbance to adapt to dynamic communication environment. Simulations confirm that the proposed method achieves stable platooning even under the conditions with time delays and packet losses.

Keywords: Network System Integration, Information Management Systems, System Integration
Abstract: Body-linked interaction is one of the promising interactive communication services, in which human motion data are transmitted to a remote user and its motion is reproduced through the body of the remote user with functional electrical stimulation (FES) devices. Supporting this interaction over the Internet, particularly in multi-access environments where heterogeneous traffic coexists, requires strict end-to-end delay guarantees that satisfy the quality-of-service requirements of FES applications. Furthermore, when the network cannot meet the specified delay deadline, a feedback mechanism that notifies the application layer of deadline violations becomes necessary. Our previous study examined control architectures and controller placement under delay constraints. However, network models designed to satisfy these constraints have not been clarified. This study proposes a network flow optimization model that jointly accounts for control architecture design and delay constraints. The networked control architecture can be classified as centralized and decentralized architectures. This study considers a deadline requested by the FES applications as a delay constraint, assuming distributed FES control based on the decentralized architecture. The effectiveness of the proposed model for interactive communication services using FES is confirmed by numerical results.

Keywords: Process Control, Safety, Environment and Eco-Systems, Multivariable Control
Abstract: This study proposes a self-optimizing control (SOC)-inspired controlled variable selection method for resource-efficient water–light management in small-scale smart planters. Many existing planter systems rely on independent threshold rules, such as irrigating when soil moisture falls below a predefined level. Conversely, greenhouse and plant-factory systems often employ advanced optimization-based environmental control, but their modeling and computational requirements can be excessive for low-cost microcontroller-based planters. To bridge this gap, we formulate planter operation as a controlled-variable design problem. A virtual controlled variable is constructed from multiple measurements and features, including soil moisture, estimated drying rate, ambient temperature, relative humidity, light intensity, and daily light integral. Following the concept of SOC, the weighting coefficients are designed offline to reduce the expected loss in a multi-objective cost function penalizing water use, LED energy consumption, actuator switching, soil-moisture zone violations, and light deficit. The online controller then maintains the virtual variable within a desired region using lightweight event-triggered irrigation and supplemental lighting. Preliminary simulations and experiments compare the proposed method with conventional threshold-based control, demonstrating its potential to improve resource efficiency under varying environmental conditions.

Keywords: Process Control, Safety, Environment and Eco-Systems, Multivariable Control
Abstract: Smart home operation involves multiple and sometimes conflicting objectives, including thermal comfort, visual comfort, energy consumption, and actuator switching, under disturbances such as outdoor temperature variation and solar radiation. Model predictive control (MPC) is a promising framework for this problem, but its practical deployment may require substantial effort for model development, commissioning, maintenance, and integration with heterogeneous devices. This study investigates self-optimizing control (SOC) as a control-structure design method that translates such a multi-objective operation problem into an implementable feedback control scheme. A tabletop smart home equipped with temperature, humidity, and illuminance sensors, and actuators including a heater, ventilation fan, LED lighting, and motorized blinds, is considered. An operational cost function is defined by combining penalties for thermal comfort violation, visual comfort violation, energy use, and actuator switching. Based on this objective, controlled variables are constructed from available measurements so that constant setpoint regulation yields a small loss from optimal operation. The proposed SOC-based structure is evaluated by numerical simulation and tabletop experiment.

Keywords: Robust Control, Autonomous Decentralized Systems, Discrete Event Systems
Abstract: Wireless sensor networks (WSNs) play important roles in cyber-physical systems as they are capable to gather information in a distributed way from wide geographical areas. Within WSNs, it is critical that all nodes share a common clock for consistency in the collected data. Our work investigates clock synchronization based on pulse-coupled oscillators (PCO) and, in particular, their resilient operation in the presence of malicious nodes. The malicious oscillators aim to disrupt the phase synchronisation of normal oscillators by transmitting pulses at arbitrary instants. A resilient update strategy is developed by combining pulse-based update rules with filtering mechanisms that mitigate the effects of malicious pulses, and we present graph topological conditions for guaranteeing their resiliency. Through numerical simulations, we show that compared to conventional methods, the proposed approach can reduce convergence speed among normal oscillators under adversarial behaviors.

Keywords: Robotic and Automation Systems, System Integration, System Engineering
Abstract: In libraries, publishing distribution warehouses, and bookstores, book storage and retrieval operations are performed routinely, and many still rely on manual labor. Labor shortages have become a major issue in these workplaces, creating a need to automate book storage and retrieval operations. Although systems such as book sorters and automated warehouses have been proposed, they often require large-scale dedicated facilities, leading to high initial costs. They may also require changes to existing bookshelves and operational workflows, reducing usability. In this study, we develop an automated book storage and retrieval robotic system using a mobile manipulator that can be introduced without major modifications to existing bookshelf environments. The proposed system uses the mobile robot to move to the target bookshelf as a common operation. For retrieval, it recognizes the target book from visual information, grasps it, and removes it from the shelf. For storage, it grasps the book to be stored, recognizes an available space on the bookshelf, and places the book there. By integrating these functions, the system aims to reduce worker burden and improve operational efficiency.

Keywords: Autonomous Decentralized Systems
Abstract: In recent years, cooperative transportation systems involving multiple robots have attracted significant attention in the industrial and logistics sectors. However, most currently deployed systems can only transport objects of pre-defined shapes, making it difficult to handle large or complex-shaped objects. To address this issue, this study proposes a cooperative transportation system that utilizes visible light communication (VLC) to share information between robots, enabling stable transport of objects with an eccentric center of gravity. Furthermore, its effectiveness is validated through simulations using ROS2 and Gazebo. In this presentation, we propose an algorithm to suppress load concentration during both the preparation and transportation phases, and report on the evaluation results of its usefulness based on the simulations.

Keywords: Network System Integration
Abstract: Cyber-physical systems (CPSs) are expected to play a central role in Industry 5.0 but are exposed to the risk of cyberattacks that lead to physical damage. One promising countermeasure is an encrypted control system, which performs computations directly on encrypted data using homomorphic encryption. However, adversaries can tamper with encrypted data via malleability attacks and launch replay attacks using recorded encrypted data. These attacks pose a critical threat to CPSs safety and defending against them remains challenging. In this study, we develop an encrypted control system using an updatable keyed-homomorphic encryption (KHE) scheme, which is secure against malleability attacks and replay attacks. In the updatable KHE scheme, private keys and encrypted data are updated at each step, and the encrypted data are verified under updated verification keys, preventing replay attacks and malleability attacks. Moreover, we design a tamper-detection signal that indicates whether encrypted data have been tampered with, enabling localization of attacked communications and tampered data. By applying the developed scheme to an actual control stage, we validate its effectiveness against replay and malleability attacks in real time.

Keywords: Autonomous Decentralized Systems, Guidance and Flight Control, Adaptive and Optimal Control
Abstract: With the growing demand for cooperative mobile robot operations, formation control of multi-agent systems is becoming important in wide areas. When the number of agents is large, uniform information exchanges among agents may delay the achievement of control objectives. This study develops a hierarchical approach to formation control of multi-agent systems. In addition, group-wise obstacle avoidance and inter-group collision avoidance are addressed. Each agent is modeled as a double integrator. Agents are divided into some groups. To evaluate formation control performance hierarchically, individual, intra-group, and inter-group performance outputs are defined. Formation control laws are, then, constructed as a linear quadratic optimal regulator. An algebraic property of Riccati equations reveals that its positive semidefinite solution and the associated optimal control law are decomposed according to the hierarchical structure of the performance output. This decomposition allows us to design the optimal gains for individual velocity control, intra-group formation keeping, and inter-group coordination in a recursive manner. Furthermore, the inter-group coordination incorporates the adjustment of the inter-group control input for group-wise obstacle avoidance and inter-group collision avoidance by means of control barrier functions. Numerical simulations verify that the proposed framework produces appropriate group-level avoidance behavior while maintaining overall formation.

Keywords: System Engineering, Safety, Environment and Eco-Systems, Network System Integration
Abstract: This study presents a field demonstration test of smart safety equipment designed to reduce human error and improve safety during railway field operations. The test was conducted on the Osong Test Line and focused on an integrated monitoring framework for workers, drivers, and control-room operators. Before entering the work site, workers pass through a gate-type condition monitoring system that records work history, alcohol status, and body temperature to support fitness-for-duty screening. During field work, a smart safety vest equipped with an edge device collects worker location and task information and transmits it to a central server. AR glasses and mobile CCTV units are used to monitor work zones, detect boundary violations, correct positioning errors, and identify hazardous objects or approaching rail vehicles. For train drivers, cab-mounted CCTV monitors both the driver and the forward track environment to detect inattentive behavior and abnormal objects on the track while representing the driver as an avatar for privacy protection. The collected data are integrated into a central server and visualized using a digital twin interface, enabling real-time safety supervision from the control room.

Keywords: Robotic and Automation Systems
Abstract: This study aims to develop a robot that imitates the movements of a panda. It is assumed that pandas achieve quasi-static walking by shifting their center of gravity through the use of waist motion. By referring to such usage of the skeletal structure, this study constructs, in simulation, a multi-degree-of-freedom quadruped robot whose configuration is inspired by panda locomotion, and aims to realize panda-like walking by using simulation software and a reinforcement learning algorithm. As a result, although fully quasi-static walking was not achieved, it was confirmed in simulation that the robot equipped with waist degrees of freedom was able to move forward while exhibiting waist motion.

Keywords: System Integration, Intelligent Systems, Mechanical Systems Control
Abstract: AI-Formula is a technical challenge designed to foster practical Physical AI skills through real-world autonomous mobility racing. While conventional time-attack-style race rules enable the evaluation of basic autonomous driving performance, they are limited in representing diverse traffic situations encountered in real-world environments. In this paper, we propose a race rule design that incorporates scene understanding capabilities to enhance real-world applicability in AI-Formula.

Keywords: Transportation Systems
Abstract: Road-segment-level traffic accident risk prediction is a core need for municipal road safety planning, yet conventional approaches rely on static road geometry, historical accident counts, and scalar environmental summaries such as POI counts, ignoring rich visual context. We integrate aerial imagery features, extracted using DINOv3 pretrained on 493M satellite images, with tabular features describing road geometry, infrastructure, and traffic conditions from vehicle telematics in a segment-level Poisson regression model. On 2024 accident data from five municipalities in Tochigi Prefecture, Japan, aerial imagery yielded the best Capture rate at the top 5% of segments (ranked by annual expected accident count) across all five accident types, with a lift of +26.8% to +410.0% over a persistence baseline and +12.3% to +28.3% over an identically-tuned model without imagery; likelihood-based metrics (Poisson deviance and pseudo-R2) show the same ordering.

Keywords: Intelligent Systems, Computational Intelligence, Transportation Systems
Abstract: Preventing traffic accidents requires not only the detection of immediately observable hazards but also the identification of latent risks that arise from complex interactions among traffic participants. While Large Language Models (LLMs) possess strong capabilities in natural language explanation, applying LLMs alone to traffic risk analysis makes it difficult to ground explanations in structured scene dynamics. This position paper investigates explainable traffic risk understanding through the integration of Graph Neural Networks (GNNs) and LLMs, where interaction-aware traffic scene representations are encoded by GNNs and human-interpretable explanations are generated by LLMs. In particular, we focus on traffic scene graphs represented by numerical state descriptions without textual attributes, and discuss a simple and practical GNN–LLM architecture, as well as an offline–online operational strategy based on vector database retrieval to satisfy real-time requirements. Rather than aiming to report finalized performance evaluations, this paper organizes design choices, challenges, and open issues for principled investigation of GNN–LLM integration toward safety support, with the goal of stimulating further discussion.

Keywords: Virtual Reality Systems, Human Interfaces, Simulation of Large Systems
Abstract: Toward the reduction of motorcycle accidents in ASEAN, we have been developing multi-agent traffic simulations that reproduce mixed traffic flows in which motorcycles and passenger cars are intensively intermingled. We have so far built a multi-agent traffic simulation for Bangkok, Thailand, incorporating the behavioral characteristics of local motorcycle riders. By driving this simulation in real time, we have also built an environment in which a human rider can enter the virtual space and operate the system as a riding simulator. In this paper, we report that this multi-agent motorcycle simulator enables an evaluation environment in which the activation and effectiveness of Advanced Rider Assistance Systems (ARAS) can be examined within the naturalistic traffic flow of urban Bangkok. We further discuss, as future prospects, the challenges and directions for extending the simulation model to Vietnam, where traffic density is higher, and to India, where traffic is highly unstructured.

Keywords: Flow Measurement and Control, Mechanical Measurement, Sensors and Transducers
Abstract: Supersonic jets are widely used in industrial processes to remove adhered particles and thin liquid films from solid surfaces. In such applications, the removal efficiency is strongly influenced by jet oscillation characteristics and the stagnation pressure at the impingement surface. Although the flow structures of free and impinging supersonic jets have been extensively studied, the relationship between jet oscillation and the structural response of the target surface remains insufficiently understood. This study experimentally investigates the oscillatory behavior of a supersonic impinging jet and the response of a thin flat plate subjected to the jet. Multiple strain gauges are mounted on the plate to measure its dynamic deformation. The distance between the nozzle exit and the plate is systematically varied to examine its influence on jet oscillation and the resulting structural response. Analysis of the strain signals using the Euclidean distance indicated characteristic oscillation patterns of the jet on the plate.

Keywords: Mechanical Measurement
Abstract: A continuous force calibration method was proposed by EURAMET to clarify the characteristics of the force measurement system of a material testing machine. In general terms, this method is a comparison calibration between the testing machine and a reference force transducer; however, the two instruments may have a clock difference that must be corrected. The correction method proposed by EURAMET contains errors due to nonlinearity and the dynamic response of the force transducers. The authors demonstrate an alternative method to evaluate the clock difference using repeated step-by-step loadings. This approach can improve the accuracy of the continuous force calibration method.

Keywords: Mechanical Measurement, Signal and/or Image Processing, Identification and Estimation
Abstract: This study quantitatively evaluated the process of blending skin spots with surrounding skin tone through concealer application with a cosmetic puff. Skilled human application was first measured in terms of contact force, puff orientation, and trajectory to clarify motion characteristics required for effective blending. Changes in the state of the concealer film caused by repeated contact were then analyzed using z-axis reaction-force responses. Finally, robotic application experiments and image-based evaluation were conducted to examine how contact force, the number of contacts, and added shear motion affect concealment performance.

Keywords: Mechanical Measurement, Signal and/or Image Processing
Abstract: This paper is concerned with an advanced infrasound observation system using micro-electro-mechanical systems (MEMS) atmosphere sensors. The infrasound observation is used for a natural disaster warning system and so on. Recently, an infrasound observation system using a MEMS atmosphere sensor has been developed for easy and inexpensive implementation. However, it is difficult to distinguish infrasound due to sensor noise in the MEMS atmosphere sensor. In the previous study, the zero-phase filter has been applied to reduce the noise in the atmosphere sensor data. However, it was clarified that the waveform of the infrasound over 1 Hz caused by a thunder and so on might also be declined by the filter. In this study, we propose the infrasound observation system combining a finite-impulse-response type zero-phase filter and a MEMS atmosphere sensor array which lays out multiple sensors. In this approach, the MEMS sensor array is simulated by generating infrasound data, which includes the thunder's waveform and noise. The relationships between the filtering performances, which are noise reduction and decline of the thunder's waveform, and the proposed system's parameters, which are the number of sensors and the cutoff frequency of the zero-phase filter are analyzed in the simulation study.

Keywords: Manufacturing Systems, Human-Machine Systems, Modeling, System Identification and Estimation
Abstract: This study contributes to the development of an advanced evaluation system for manual pouring works in the casting industry. Maintaining the quality of casting products and a safe working environment in manual pouring work requires a certain level of pouring skill. In recent years, the development of an evaluation system for the pouring work has been promoted to enhance traceability of the casting products. It is required to analyze the pouring work on demand immediately. Therefore, we propose a system that visualizes the flow rate of the pouring work using a cloud-based platform. In the proposed approach, digital twin of the pouring work is implemented on the cloud-based platform. The pouring flow rate can be visualized automatically by uploading weight data of the molten metal that flows out of the ladle and tilting angle data of the ladle to the platform. The efficacy of the proposed approach is verified by experiments on the manual pouring work in practical use, and the differences in pouring behavior between the skilled worker and the novice worker are successfully visualized. The proposed approach provides a foundation for objective evaluation and training support of pouring skills in foundry work.

Keywords: Identification and Estimation, Mechanical Measurement, Biological and Physiological Engineering
Abstract: This study proposes a per-cycle Standard Linear Solid (SLS) parameter estimator that combines the SLS viscoelastic model with Hertz contact theory. By fitting a discrete-time linear regression model to force–displacement data from each indentation cycle, the equilibrium elastic modulus E_0, viscoelastic modulus E_1, and relaxation time τ are identified cycle by cycle. Validation experiments on two silicone gel phantoms with approximately fivefold stiffness difference showed that the dynamically estimated E_0 agreed with reference values obtained from constant-velocity indentation to within 5%, confirming the validity of the estimator. Application to the human gastrocnemius muscle revealed a monotonic decrease in E_0 of 26.3% on average over 300 repeated indentation cycles. A window-size analysis using a rolling median filter further demonstrated that reliable initial estimation is achievable with as few as 10 cycles (10 s) of indentation.

Keywords: Manufacturing Systems, Information Management Systems, Factory Automation
Abstract: Industrial fields in Japan have conducted energy management and energy-saving activities since the 1970s. Energy intensity has commonly been used as an indicator to evaluate these activities. In the early stages, equipment retrofits resulted in significant energy savings and improvements in energy intensity. However, as most equipment has already been retrofitted, the additional energy savings from further retrofits are becoming increasingly limited. Therefore, attention has shifted toward organizational energy management frameworks that systematically promote energy-saving activities. Because energy management activities cover diverse aspects of organizational efforts, energy intensity alone cannot adequately represent their effectiveness. To address this issue, the Japan Electronics and Information Technology Industries Association (JEITA) proposed an international standard for assessing energy management system and the standard was published in 2023 as ISO TS 50011. Six factories in Japan, China, Korea and ASEAN region showed interests in this standard and conducted self-assessment. This paper presents the use cases of assessing with ISO TS 50011 in industrial fields and discusses the assessment results. Through these use cases, potential method for future utilization and additional criteria to be incorporated into the checklist in the standard are considered.

Keywords: Manufacturing Systems, SCADA and Network Systems, Factory Automation
Abstract: This paper reports the coordination design of a smart industrial energy management system testbed. The testbed generalizes the concept of smart industrial energy management systems and is developed to reveal to industrial sectors the necessary conditions for coordinating such systems. This paper focuses on the conceptual design and individual system modeling of the Factory Energy Management Systems (FEMS) coordination framework, which automates energy optimization within the factory and automatically shares the factory’s dynamic energy data among factories and an aggregator through trusted communication. The testbed architecture is proposed as an alternative to the method of individual FEMS. The proposed method shows a FEMS coordination system with EMSs and Manufacturing Execution System (MES) that automates optimization for both production and energy efficiency. Energy and Carbon Footprint (CFP) values are managed as dynamic data. The testbed implementation, the demonstration scenario, and the unit-experiment results for each module are explained. They present several use cases for the testbed, which generalizes a smart industrial energy management system in a factory to inform the industry sector and contribute to industrial decarbonization.

Keywords: Manufacturing Systems, Factory Automation, Adaptive and Optimal Control
Abstract: Realizing an integrated Energy Management System (EMS) requires a common framework for handling information across multiple systems to achieve factory-wide optimization. This study proposes a method for centrally managing information associated with multiple EMSs using a Common Information Model (CIM), in which industrial energy equipment is modeled in a unified manner. A case study considering equipment performance degradation demonstrates that CIM enables the simultaneous and consistent use of information from multiple EMSs, resulting in globally optimized factory operation. Furthermore, the proposed approach is expected to facilitate system interoperability and improve scalability across different factories.

Keywords: Autonomous Decentralized Systems, Signal and/or Image Processing, Laboratory Automation
Abstract: Even in the manufacturing industry, to improve added value in the engineering chain are required. The chemical industry is also made up of the fusion of several fields, especially chemistry and machinery, such as R&D and manufacturing equipment. This paper shows chemical digital twins.

Keywords: Manufacturing Systems, SCADA and Network Systems, Information Management Systems
Abstract: This paper is primarily for users of Manufacturing Systems who plan to apply digital transformation (DX) measures to manufacturing issues. It is often observed that vendor-originated structures of system configurations create confusion and make the user’s DX planning decision difficult. Therefore, it is needed to establish a single commonly applicable framework for effectively assessing those solutions in the market. To meet the need, a practical framework is prepared for manufacturing system analysis, based on the ANSI/ISA-95 (IEC 62264) and the NAMUR Open Architecture (NOA) and Three Systems of IT (SoR, SoI, SoE). The framework offers a form of generally applicable manufacturing systems architecture with the two DX measures of Automation (Type A) and Monitoring & Optimization (Type B). The framework has been applied for a few cases with successful results, although it is still in development.

Keywords: SCADA and Network Systems, Factory Automation, Information Management Systems
Abstract: In the manufacturing and social infrastructure sectors, addressing social challenges such as labor shortages, aging equipment, energy conservation, and decarbonization has made the promotion of digital transformation (DX) an urgent priority. OT (Operational Technology) systems, which support DX at the field level, are experiencing a growing demand for real-time processing near the field, driven by the increasing volume of data generated by sensors and control devices. At the same time, these systems face stringent constraints including limited installation space, high-temperature and dusty environments, and long-term operation requirements. Furthermore, ensuring connectivity with legacy equipment and establishing secure linkage with cloud systems are prerequisite conditions for advancing DX in OT systems. This paper presents edge computing—in which components installed near the field perform data preprocessing and local decision-making—as an effective approach to these challenges, organizes the requirements for OT devices in comparison with IT systems, and introduces, as a social implementation example that satisfies these requirements, the compact embedded industrial computer CP40 model 300, which features a compact size, fanless design, high reliability, expandability, and security.

Keywords: Mechanical Measurement, Networked Sensor System
Abstract: Rapid feedback of thickness measurement results to control systems is essential for improving thickness accuracy and reducing quality variation in steel rolling processes. In particular, the measurement update cycle and data transmission delay directly affect control performance under high-speed rolling conditions. This paper presents a high-speed measurement and low-latency data transmission method for X-ray thickness gauges. The measurement cycle is accelerated from 5 ms to 1 ms, enabling higher time-resolution thickness acquisition. In addition, a digital communication network is adopted to replace conventional analog transmission, reducing transmission delay and improving synchronization when distributing measurement data to multiple control units. The proposed system is designed for application to actual rolling lines. Evaluation under high-speed rolling conditions confirmed a reduction in the delay from thickness measurement to its use in the control system, resulting in improved control responsiveness. These results indicate that the proposed method is effective for enhancing measurement and control performance in high-speed rolling control systems.

Keywords: Intelligent Systems, Intelligent Control, Standard of Measurement
Abstract: Physical AI, a technology in which AI directly interacts with real-world objects, is rapidly spreading across diverse industrial fields such as robotics, infrastructure, and mobility. However, because it involves real-world risks, ensuring safety, reliability, and accountability is crucial. For example, various countries have introduced regulations to ensure the safe use of AI, and ISO/IEC is developing standards for AI technology. To provide an overview of the social and technological landscape surrounding Physical AI and to facilitate multifaceted discussions, a Special Panel Session will be held at SICE FES. Experts from each field will participate in the session, and general attendees are also welcome to attend the discussions.

Keywords: Transportation Systems, Signal and/or Image Processing, Remote Sensing
Abstract: Traffic accidents caused by stopped objects on expressways remain a serious issue, and the early detection of such objects is essential for traffic safety. Conventionally, stopped objects are detected through manual monitoring or background modeling. However, manual monitoring requires significant human resources and background modeling is sensitive to environmental changes. In this paper, we propose a method of detecting stopped objects without relying on background models. The key of the proposed method is the region of interest, which is used for road region restriction and stopped objects detection. Experimental results demonstrate that the road region restriction section achieves high recall and that the stopped object detection section largely detects objects in multiple scenarios. As future work, it is necessary to improve the detection accuracy for objects on lane markings and small objects.

Keywords: Transportation Systems, Signal and/or Image Processing, Remote Sensing
Abstract: Traffic surveillance camera footage is widely used for the early detection of accidents and traffic congestion. Vehicle detection using such footage is one of the key technologies for ensuring traffic safety. However, on highways during winter, even with traffic surveillance cameras, images can become unclear owing to blizzards and snow accumulation, making it difficult to accurately detect moving vehicles. In this study, we propose a method of detecting moving vehicles in winter using traffic surveillance camera footage. This method involves removing snow particles and enhancing contrast, followed by estimating and utilizing the depth of an image. As a result, we confirmed that the proposed method enables stable moving vehicle detection under winter conditions compared with conventional methods.

Keywords: Transportation Systems, System Integration, Intelligent Systems
Abstract: This paper proposes a shore-based radar-assisted remote collision avoidance system for an Unmanned Surface Vehicle (USV). Conventional approaches rely on onboard sensors such as marine radar and AIS, which are often impractical for small USVs due to cost and physical constraints. The proposed system utilizes a low-cost shore-based radar to monitor surrounding vessels and performs collision avoidance using externally obtained information. Vessel positions and motion states are extracted from radar images, and collision risk is evaluated based on the Obstacle Zone by Target (OZT) concept. When a potential collision is detected, an avoidance path is generated using the Theta* algorithm and transmitted to the USV for trajectory tracking. Field experiments demonstrate that the USV successfully avoids collisions while complying with COLREGs and maintaining a safe distance. The results indicate the feasibility of infrastructure-assisted collision avoidance without relying on expensive onboard sensors.

Keywords: Transportation Systems, Mechanical Systems Control, Mechatronics Systems
Abstract: In autonomous wheelchair navigation—for which demand is rising due to the aging population—ensuring safe and smooth interactions with pedestrians remains a critical challenge. This study proposes a Model Predictive Control (MPC) scheme that accounts for pedestrians' anisotropic avoidance characteristics to achieve control that minimizes psychological burden. Based on the analysis of prior intersection experiments, we identified that the risk area perceived by pedestrians extends in a fan-shaped distribution ahead of their heading. This spatial characteristic is formulated as a potential field using a Von Mises distribution and integrated into the MPC cost function to generate trajectories that actively suppress intrusion into the pedestrian's psychological domain. Experimental results using an actual wheelchair, together with a comparative simulation against the conventional isotropic model, demonstrate that the proposed method prevents the unnecessary detours and sudden decelerations associated with isotropic potentials, successfully balancing mobility efficiency and psychological safety in smooth navigation.

Keywords: Transportation Systems, System Engineering, Social Systems
Abstract: In this study, we propose a novel framework for jointly determining dynamic parking pricing and parking allocation based on a Stackelberg game. By formulating the interaction between a parking manager and multiple drivers, the proposed method simultaneously determines parking prices and individual parking assignments while considering the preferences of both parties. Drivers aim to minimize parking costs and travel distance, whereas the parking manager seeks to maximize total profit and to balance occupancy rates across parking lots. Unlike conventional approaches that determine pricing and allocation in separate stages, the proposed method integrates these decisions into a single-stage optimization, thereby reducing computational cost and enabling pricing that directly reflects the allocation of drivers arriving at the current time step.

Keywords: Manufacturing Systems, Factory Automation, Sensors and Transducers
Abstract: This paper proposes a tool condition monitoring method based on motor load current data acquired from a computer numerical control (CNC) lathe. Individual machining cycles are extracted from streaming current data, and kernel density estimation (KDE) is applied to cycles acquired under normal conditions to construct a time-varying reference distribution. Deviations of test current waveforms from this distribution are quantified using a KDE-based score. Experiments involving 1000 consecutive machining cycles showed that the summary score generally increased over successive cycles and decreased markedly after scheduled tool replacement. The proposed method also assigned higher scores to deviations occurring during periods of low reference variability. These results demonstrate the potential of KDE-based scoring for monitoring changes in tool condition during CNC machining.

Keywords: Manufacturing Systems, Discrete Event Systems
Abstract: The complexity in shop production processes was modeled using process route complexity, processing time variation, and load factor. By applying damage spreading and percolation analysis, we evaluated the system's dynamic responsiveness to disturbances. We found that production systems exhibit phase-like transitions into freely flowing, fluidly correlated, or congested regions depending on these parameters. Notably, the fluidly correlated state demonstrates sensitive dependence on initial conditions, suggesting that effective production management requires strategies ensuring dynamic controllability rather than relying solely on optimal static schedules.

Keywords: Information Management Systems, Intelligent Systems, Manufacturing Systems
Abstract: Integrating heterogeneous ontologies typically requires manual transformation engineering. We recast this task as Type Inhabitation: ontology classes are types, catalog entries are typed functions, and conversion pipelines are well-typed compositions. We introduce a cost-aware backward search that ranks candidate paths by execution cost and confidence. When no path exists, we detect and classify structural gaps using forward/backward reachability. Missing functions are synthesized by an LLM and filtered through a four-stage validation pipeline (syntax, type, semantic, and integration checks). In our LLM-assisted experiments over three local catalogs---carbon footprint, GHG scopes, and ChemSHERPA(R) battery reporting---path search remained sub-millisecond, ablation experiments produced direct, bridge, and multi-source gaps, and two GPT-5.4 gap-completion scenarios generated 5 valid first-round candidates out of 10 and 5 additional valid second-round candidates after validator feedback. These results suggest that type-theoretic synthesis plus constrained LLM completion is a practical approach to automated ontology integration, while also highlighting the need for stronger semantic validation beyond signature-level checks.

Keywords: Components and Devices, Advanced Computics
Abstract: This study investigates the FPGA implementation of an encrypted controller for an industrial linear stage. An efficient FPGA-oriented architecture for an encrypted controller is proposed and implemented on an AMD Artix-7 FPGA with a 64-bit key, and its performance is evaluated. Furthermore, the Karatsuba method is introduced to extend the key length to 143 bits. The results clarify the trade-offs among key length, processing latency, and resource utilization, demonstrating the feasibility of encrypted control for industrial applications.

Keywords: Micro and Nano Devices, Integrative Biophysical Engineering and Informatics, Biological and Physiological Engineering
Abstract: Dynamic characterization of biomolecular reactions requires platforms that can apply time-varying concentration stimuli while continuously measuring the resulting responses. In this study, we propose an integrated microfluidic chemostat that combines an L-2L ladder microfluidic D/A converter with a ring-shaped microfluidic chemostat. The device generates multi-level concentration signals, delivers them to a continuously mixed reaction space, and observes the resulting input-output behavior within the same platform. Experiments demonstrated stable generation of eight discrete concentration levels and observable multi-level step responses. These results indicate that the proposed architecture provides a practical basis for dynamic characterization of biomolecular reactions.

Keywords: Laboratory Automation, Adaptive and Optimal Control, Manufacturing Systems
Abstract: This study proposes a high-speed optimization method for robot cell production systems using quantum annealing to overcome the computational challenges of layout and path planning. We introduce a stepwise optimization framework for RCMS and the Adaptive Area Awareness (A3) Algorithm, which main tains low computational costs by selectively increasing resolution in high-risk interference zones. Simulation results for a 6-axis robot show that our method generates efficient, collision-aware paths and layouts with signif icantly reduced processing time. This approach offers a scalable solution for automating high-mix, low-volume manufacturing and improves overall design efficiency.

Keywords: Signal and/or Image Processing, Identification and Estimation, Computational Intelligence
Abstract: This paper proposes a design method for finite wordlength FIR Hilbert transformers using a genetic algorithm (GA). The conventional design of finite wordlength filters often suffers from performance degradation due to coefficient quantization. To address this issue, we introduce a hybrid approach that first derives a continuous-valued solution using the Lagrange multiplier method and then optimizes the finite wordlength coefficients using GA. Simulation results demonstrate that the proposed method significantly reduces the phase estimation error compared to the conventional sequential quantization method.

Keywords: Signal and/or Image Processing, Identification and Estimation, Sensors and Transducers
Abstract: This study investigates the feasibility of non-destructively measuring beef with a high-voltage dielectric spectrometer to assess its internal state during freezing. To clarify how meat composition influences dielectric response, impedance measurements are performed on three sample types: lean beef, marbled beef, and beef fat, while freezing at 18°C. Furthermore, by mapping changes in impedance over time, the evolution of the internal state as freezing progresses is visualized. As a result, a significant change in impedance is observed around 0.5 Hz, and the timing of this change closely matches the trend in water content in the sample. These results suggest that high-voltage dielectric spectral analysis enables non-destructive observation of structural changes in beef during freezing.

Keywords: Signal and/or Image Processing, Identification and Estimation, Robotic and Automation Systems
Abstract: 歩行者を追いかけ、障害物を避けるために自律的に行きます。 車椅子は継続的な追跡が必要で、それを防ぐために 閉塞によるターゲット損失および 誤認。したがって、本研究では 電動車椅子用の歩行者追跡システム 歩行者追跡をLiDARとカメラで利用しています。私たちは 閉塞中の歩行者の連続追跡を実現 似た身体を持つ歩行者がいる環境 特徴が存在し、標的は一時的に 迷子になった。Pure Pursuitとモデル予測を組み合わせるこ&#

Keywords: Signal and/or Image Processing, Identification and Estimation, Computational Intelligence
Abstract: This study introduces a predictive model for determining the position of a crane payload by utilizing the phase difference of sound waves. The predictor is developed using an adaptive filter based on the recursive least squares method. Additionally, the wire length is estimated through a neural network. The efficacy of the proposed method is validated through experimental results.

Keywords: Signal and/or Image Processing, Identification and Estimation, Analytical Measurement
Abstract: Existing studies mainly emphasize sensor development and signal acquisition, while lacking systematic comparisons of how different sensing modalities, such as force, piezoelectric signals, perform in vibration-based classification of complex contact states. This study developed a prosthetic finger with embedded tactile sensors for contact vibration measurement. By simultaneously acquiring signals from one external contact force sensor and three embedded sensors, vibration signals were analyzed through time- and frequency-domain features, and their discriminative capability was evaluated using the Fisher score. The results indicate that feature separability strongly depends on both the target label and sensor type. Effective classification requires task-specific feature selection and sensor choice. The proposed analysis provides useful guidance for designing vibration-based sensing systems for contact state recognition.

Keywords: Signal and/or Image Processing, Remote Sensing, Social Systems
Abstract: The purpose of this study is to develop new applications for oblique aerial imagery. We propose a method for detecting cityscape orientation lines from aerial images. It consists of five main steps;(a)the reduction of image noise;(b)the calculation of edge angle and magnitude;(c)the removal of unnecessary regions;(d)the removal of short edges and the emphasis of long edges;(e)the detection of cityscape orientation. We applied the proposed method to oblique aerial images and evaluated it. As a result, we were able to achieve stable detection of cityscape orientation. Finally, we consider the applicability of the proposed method.

Keywords: Signal and/or Image Processing, Biological and Physiological Engineering
Abstract: Quantifying "deliciousness" is essential for adding value to agricultural products, but its subjective nature makes it difficult to evaluate. While previous studies focused on simple tastes, physiological responses to complex foods like tomatoes remain under-explored. This study proposes a Kansei evaluation model focusing on individual variability in response to sugar content difference. We compared subjective and biological responses (PPG, EEG) to high-Brix and low-Brix tomatoes. To isolate pure taste-driven reactivity, a sample-holding method without swallowing was used. Physiological states were quantified using the heart rate change rate (∆HR%), which normalizes baseline differences among participants to capture pure stimulus reactivity. Results showed a clear dissociation: while subjective arousal was significantly higher for high-Brix tomatoes (p < .001), mean physiological responses showed no significant difference (p > .05; high-Brix: +1.87%, low-Brix: -2.10%). This discrepancy suggests that while subjective recognition is uniform, physiological responses are highly individualized. The findings highlight the need for a Kansei engineering approach that accounts for individual sensitivity rather than group averages. Future work will refine signal quality to bridge the gap between subjective recognition and biological signals.

Keywords: Signal and/or Image Processing, Mechanical Systems Control
Abstract: Ultrasound shear wave velocity (SWV) is known to be related to muscle tension and contraction level. Most previous studies have treated localized SWV values as representative of the entire muscle cross-section. However, for investigating the relationship between SWV and stress or tension, the spatial distribution within the cross-section is mechanically important. Therefore, this study aimed to quantify the SWV distribution across the muscle cross-section in the triceps surae muscles in vivo and to examine its spatial heterogeneity. Because SWV measurement must be performed along the longitudinal direction of the muscle fibers, a custom-made device was developed to control the ultrasound probe position. By acquiring sequential longitudinal ultrasound images and reconstructing them into a three-dimensional dataset, the cross-sectional SWV distribution was obtained. Measurements were performed on a single healthy subject under rest and low-level contraction conditions. The results showed that SWV distribution within the muscle cross-section was non-uniform in both conditions. At rest, higher SWV values were observed in the peripheral regions compared to the central region, whereas this pattern partially reversed during contraction, particularly in the gastrocnemius muscles. In contrast, the soleus showed smaller changes. These findings suggest that localized measurements may not adequately represent whole-muscle mechanical properties when estimating muscle tension.

Keywords: Identification and Estimation, Signal and/or Image Processing, Computational Intelligence
Abstract: In crane operations, accurate estimation of the sway angle is essential for the anti-sway control. Although camera-based methods offer high precision, they encounter practical challenges in deployment, such as visual occlusion. Alternatively, sound-based estimation utilizing two microphones is cost-effective; however, it is vulnerable to sound wave diffraction and environmental reflections, which can lead to reduced accuracy. This paper presents a framework for estimating sway angles that combines the practical benefits of sound-based estimation with camera-based ground truth during offline training. A neural network (NN) compensates sound-based estimation errors derived from sound phase differences. This method synthesizes operational data by randomizing the noise phase while preserving the DC component (average estimation offset) and spectral power of real-world sound-based disturbances. Experimental results on a full-scale tower crane demonstrate that the proposed NN-based method reduces estimation errors compared to the conventional low-pass filtering approach.

Keywords: Identification and Estimation, Intelligent Systems, Safety, Environment and Eco-Systems
Abstract: Driver fatigue is one of the main risk factors for road accidents. This study proposes a fatigue detection approach based on Long Short-Term Memory Autoencoder (LSTM-AE) applied to time-series driving data. The model is designed to learn the normal temporal pattern of each participant during the baseline period and identify deviations from that pattern through reconstruction errors by calculating anomaly scores using the mean squared error (MSE). Data from 10 participants analyzed showed that the anomaly scores increased progressively with increasing driving duration with varying degrees of variation. The result of progressive accumulation of driver fatigue over time through a robust temporal smoothing mechanism and displaying through an anomaly window also shows the same trend although there are variations for each participant. These results indicate that LSTM-AE is capable of learning normal driving patterns and detecting temporal deviations related to fatigue development. With its ability to capture sequential dependencies in time-series data and adapt to individual baseline characteristics, the proposed approach has the potential to be a promising framework for early detection of driver fatigue.

Keywords: Identification and Estimation, Sensors and Transducers, Adaptive and Optimal Control
Abstract: Encoders are essential angle sensors for motor control and positioning. Among them, magnetic encoders are attractive because they are low-cost, compact, and robust in harsh environments, although their accuracy is often lower than that of optical encoders due to signal distortion caused by eccentricity, contamination, and temperature variation. These non-ideal factors generate periodic error components that degrade electrical angle estimation. This paper proposes a compensation method that combines a phase-locked loop (PLL) with a two-stage ADALINE. The proposed method estimates and suppresses dominant periodic components in the phase detector output, thereby improving electrical angle estimation accuracy without modifying the encoder hardware. Its effectiveness is evaluated offline in both the time and frequency domains.

Keywords: Identification and Estimation, Modeling, System Identification and Estimation, Mechanical Systems Control
Abstract: In industrial systems, periodic maintenance is performed to prevent unexpected machine failures; however, it cannot eliminate all faults. This study aims to detect anomalies that may lead to critical failures at an early stage. Conventional approaches based on system identification require expert knowledge to construct control models, and it is often difficult to define quantitative criteria for anomaly detection. To address these issues, we propose an anomaly detection method that combines machine learning based parameter estimation with a robustness metric, namely the robust performance margin, which represents tolerance against parameter variations. The proposed method enables relatively simple and quantitative anomaly detection without relying on detailed prior knowledge of the system. Experimental results demonstrate that the proposed approach can successfully detect anomalies, although estimation performance may degrade depending on the physical characteristics of the target system.

Keywords: Identification and Estimation, Networked Sensor System, Signal and/or Image Processing
Abstract: Communication efficiency is an important issue in distributed adaptive estimation. Partial diffusion strategies reduce communication cost by exchanging only a subset of parameter vectors among neighboring nodes. However, most existing methods are limited to white Gaussian input environments, and partial diffusion under correlated inputs has not been sufficiently explored. To address this limitation, this paper proposes an improved partial diffusion algorithm for correlated input scenarios. Based on mean-square deviation analysis, the proposed method incorporates an effective whitening scheme based on an affine projection algorithm in the adaptation step and a weighted information transmission rule in the combination step. In addition, the algorithm does not require prior knowledge of the noise variance. Simulation results show that the proposed method operates reliably under correlated inputs while significantly reducing communication cost compared with full diffusion schemes.

Keywords: Identification and Estimation, Electrophysiology and Kinesiology, Mechatronics Systems
Abstract: In this study, we developed a simple robotic platform for ankle impedance estimation during standing balance control. The platform is equipped with a DC servo motor, a torque transducer, and a high-resolution encoder. To enable its application to balance control tasks that include voluntary movements, a high-pass filter is applied to extract the angular response to the perturbation torque from the recorded data. The developed platform was validated through experiments on ankle impedance estimation during quiet standing. Furthermore, the estimation of time-varying ankle impedance during voluntary postural sway was successfully demonstrated.

Keywords: Identification and Estimation, Signal and/or Image Processing, Human-Machine Systems
Abstract: We propose an unconstrained vision-based method to estimate handbell tilt during performance using custom YOLO11 keypoints tailored to handbell geometry. Two instrument-specific keypoints top_center, bottom_center are defined to compute tilt from video frames and validated against synchronized IMU measurements. On a 460-frame test sequence, raw per-frame estimates achieve RMSE 7.92~deg with 0.99 correlation; Kalman filtering reduces RMSE to 7.10~deg. We further identify and quantify the impact of specular reflection on keypoint detection accuracy, showing that periodic large errors correlate with reflection-induced mislocalization of the top_center keypoint. This sensor-free approach (the IMU is used solely as an evaluation reference and is not required during performance) enables natural performance analysis without instrument attachment.

Keywords: Identification and Estimation, Signal and/or Image Processing, Medical and Welfare Systems
Abstract: This paper presents a projection-based geometric framework for clinically crown-to-root ratio CRR estimation that measures crown and root lengths along the anatomical tooth long axis rather than the image coordinate system. The proposed method estimates the CRR using clinically relevant landmarks: the crown reference point, alveolar bone crest, and root apex, supporting periodontal diagnosis and clinical decision-making in digital oral healthcare. Three measurement strategies were evaluated: vertical-axis, Euclidean-distance, and projection-based measurement along the tooth long axis. By aligning measurements with tooth orientation, the proposed approach reduces sensitivity to radiographic rotation and geometric variation. The framework additionally supports side-specific periodontal analysis and is designed for integration into web-based digital health and periodontal charting systems for automated and clinically interpretable CRR assessment in digital dentistry.

Keywords: Opto-Electronic Measurement, Signal and/or Image Processing, Sensors and Transducers
Abstract: Numerous historical documents in Japan remain physically unopenable due to water and insect damage, leaving their contents inaccessible. Terahertz (THz) waves offer a promising non-contact and non-destructive approach to reading such documents. While THz radiation is transmitted through paper, it is absorbed by the carbon ink (sumi), producing a measurable contrast in the absorption spectrum that distinguishes the inked characters from the blank paper. In practical measurements, however, ambient humidity poses a significant challenge. Because water strongly absorbs THz radiation, fluctuations in moisture content obscure the spectral contrast between ink and paper and degrade the reliability of character detection. To address this issue, we developed an experimental system capable of precisely controlling the humidity of the measurement environment.This system is designed to enable high-contrast and highly reproducible measurements toward the non-destructive reading of unopenable historical documents.

Keywords: Analytical Measurement, Laboratory Automation, Signal and/or Image Processing
Abstract: To understand the reaction mechanisms in all-solid-state batteries (ASSBs), automated measurement systems have attracted attention. However, long-term operando measurements are often complicated by positional drift caused by volume expansion / contraction and morphological changes during electrochemical reactions. To address this issue, we developed an automatic position correction system (APCS) based on deep learning for the analysis of cross-sessional optical images. In this study, we report the application of APCS to operando Raman measurements of an oxide-based ASSB. A commercial oxide-based ASSB, CeraCharge (TDK), was employed as a model cell. This ASSB utilizes Li3V2(PO4)3 (LVP) as both the positive and negative electrode materials and Li1.3Al0.3Ti1.7(PO4)3 (LATP) as the solid electrolyte. The observation sample was prepared by Ar-ion milling and subsequently sealed in an operando cell under an Ar atmosphere. First, operando optical microscopy was performed during cyclic voltammetry using a digital microscope (VHX-X1, Keyence). Subsequently, operando Raman spectroscopy was conducted with APCS-enabled automatic position correction. Finally, the analyzed ASSB was characterized by SEM-EDX (JCM-6000, JEOL). The developed APCS successfully tracked positional drift caused by electrode discoloration and volume expansion / contraction. Those Raman spectra revealed transient changes in the electrolyte during the charge–discharge. Additionally, we will present APCS analysis metho

Keywords: Sensors and Transducers, Discrete Event Systems, Standard of Measurement
Abstract: This study aims to ensure the validity of calibration processes in measurement systems by combining calibration log analysis with formal method. In previous work, event logs generated by measuring instruments were analyzed using formal method to determine whether their operation complied with legal requirements formalized as logical specifications. Building on this approach, the present study extends the method to the validation of calibration procedures. Specifically, calibration logs are evaluated by verifying whether recorded events satisfy predefined logical specifications that represent correct calibration practices. A case study on pressure calibration is presented to demonstrate the applicability of the proposed approach. The results suggest that this method provides a foundation for formally and systematically assessing the validity of calibration processes based on explicit specifications.

Keywords: Flow Measurement and Control, Standard of Measurement, Ultra-High Precision Measurement
Abstract: As one method for establishing gas flow standards for specialty gases used in the semiconductor industry, the rate-of-rise method has been proposed. This method involves filling a tank of known volume with gas and observing the filling process to derive accurate flow rate values. The constant-volume tank system, which is one method used in general gas flow standards, is a static PVTt method; in contrast, the rate-of-rise method is considered a type of dynamic PVTt method. However, sufficient knowledge has not yet been acquired regarding factors such as equipment specifications and applicable flow rate ranges for dynamic PVTt methods. Therefore, in this study, we conducted experiments using a critical flow nozzle to investigate the feasibility of the dynamic PVTt method, utilizing the 11 m3 constant-volume tank system currently in use for gas flow standards in Japan. We also theoretically examined the factors required for stable operation and reported our findings.

Keywords: Signal and/or Image Processing, Sensors and Transducers
Abstract: Non-destructive estimation of chemical constituents is an important challenge in agricultural sensing. Although hyperspectral imaging provides rich spectral information, its practical deployment is limited by high cost and data redundancy.This study investigates task-oriented optical filter design for estimating glutamic acid content in tomatoes directly from spectral reflectance data. Instead of selecting spectral bands manually, optical filters are learned jointly with a neural network regressor through end-to-end optimization under non-negativity and smoothness constraints.

Experimental results demonstrated that accurate glutamic acid estimation can be achieved using only a small number of learned filters. Interestingly, the learned filter responses exhibited a strong correlation, suggesting that the network constructed a low-dimensional observation space specialized for the target task. Further analysis revealed that the difference component between filter responses contained substantially more information about glutamic acid concentration than their common component, indicating that the learned filters behaved as a differential spectral sensing system. These findings suggest that task-oriented optical filter learning can automatically discover informative spectral observations and provide an effective framework for designing compact sensing systems for agricultural quality assessment.

Keywords: Ultra-High Precision Measurement, Standard of Measurement, Signal and/or Image Processing
Abstract: A weak frequency signal measurement scheme based on optical local oscillator enhancement is proposed which uses homogeneous laser and increases the sensitivity of the receiver by 10dB compared to the Intensity Modulation/Direct Detection (IMDD). Stable frequency signals can serve as time references, enabling quantum communication and ensuring the precise operation of global navigation satellite systems (GNSS). Transmitting frequency signals over optical fibers holds great potential in existing multi-node networks. The optical carrier carrying a frequency signal will become extremely weak after long-distance transmission, and excessive noise will be introduced directly after passing through the photodetector (PD) for beat frequency. Traditional fiber optic radio frequency (RF) transmission systems use Intensity Modulation/Direct Detection (IM/DD). However, as the transmission distance increases, the system accumulates more link-induced noise, which leads to a degradation in detection sensitivity at the receiver. Erbium-doped fiber amplifiers (EDFA) are typically deployed along the link to boost optical power; nevertheless, this approach results in relatively high cost.Therefore, identifying a transmission approach that preserves high signal sensitivity while minimizing cost is of critical importance. In this paper, a demodulation technique based on optical local oscillator enhancement is proposed, which uses envelope detection (ED) at the rece

Keywords: Medical and Welfare Systems, Robotic and Automation Systems, Electrophysiology and Kinesiology
Abstract: While multi-degree-of-freedom (DOF) robots are prevalent in upper-limb rehabilitation, their mechanical complexity and high costs often hinder clinical deployment. This study investigates the biomechanical efficacy of a compact 1-DOF robotic system designed to strategically modulate complex muscle activation patterns. We developed a low-cost rehabilitation device utilizing a quasi-direct drive actuator, implementing a control architecture that dynamically transitions between proportional-derivative (PD) position control and resistive force control based on estimated external torque. To evaluate the system, four healthy subjects performed planar arc reciprocal movements under seven phase- and angle-dependent loading conditions. Surface electromyography (sEMG) analysis demonstrated that localized resistive loading during the forward movement phase selectively amplified targeted muscle activity, such as the biceps or triceps brachii. Conversely, the backward phase primarily induced joint-stabilizing co-contraction. These findings empirically demonstrate that strategically applied 1-DOF haptic interventions can effectively isolate and recruit specific muscle groups. This validates the potential of simplified robotic architectures to deliver accessible, task-specific precision rehabilitation without the need for redundant mechanical complexity.

Keywords: Human-Machine Systems, Intelligent Systems, Medical and Welfare Systems
Abstract: Charts are critical for data communication, but remain largely inaccessible to visually impaired users who rely on screen readers. While automatic chart-to-text generation systems exist, deploying such systems requires understanding which chart types can be reliably automated and which require human verification. We systematically evaluate five open-source vision models (Pix2Struct, DePlot, MatCha, UniChart, ChartGemma) released between 2022 and 2024 across four chart types using the balanced dataset of FigureQA. Our temporal analysis reveals significant progression: all 2022-2023 models completely fail on pie charts (0% accuracy), whereas ChartGemma (2024) achieves 75%. Bar charts and single-line charts reach 85-90% accuracy across recent models. Our results align with known architectural limitations: patch-based vision transformers fragment circular pie-chart structures into disconnected rectangular patches, while single-state decoders cannot simultaneously track multiple line trajectories. We implement a dual-mode accessibility system that combines annotation-based evaluation with vision-based deployment, identifying which chart types achieve sufficient accuracy for automation: bar and single-line descriptions (≥85% reliability), pie charts requiring verification (75% reliability), and multi-line charts needing manual description (low reliability). This work establishes reproducible baselines for the deployment of accessibility systems.

Keywords: Biological and Physiological Engineering, Identification and Estimation, Medical and Welfare Systems
Abstract: Electroencephalography (EEG) signals are strongly affected by electrical transfer within the skull. However, experimental evaluations have largely relied on spontaneous brain activity or induced responses, which exhibit limited reproducibility, rather than on controlled sources. We developed a prototype phantom using an ex vivo porcine skull and conductive agarose pseudobrain to determine the temporal and spatial transfer functions of the skull during EEG recording. In our system, internal electrical activity was induced by optically stimulating embedded phototransistors, thereby creating artificial current sources that emulate neuronal functions in the brain. Surface electrodes mounted on the skull successfully recorded periodic voltage fluctuations corresponding to the applied stimulations. The results demonstrate the feasibility of a controlled experimental evaluation of skull-mediated EEG signal transmission using biological skull specimens. The proposed framework may contribute to the validation of EEG forward models and the investigation of spatial attenuation characteristics.

Keywords: Sensors and Transducers, Medical and Welfare Systems, Biological and Physiological Engineering
Abstract: This study investigates a bed-integrated sensing system for unconstrained sleep monitoring using an inductance-based pressure sensor. Because the sensor detects the deformation of an elastic body via a coil installed at its base, no electric circuitry is required in the contact area, ensuring both user comfort and device durability. The objective of this study is to evaluate the feasibility of detecting respiration and heartbeat signals from an individual lying on a bed. Experiments were conducted with 11 subjects. The results demonstrate that respiration signals were successfully detected in all subjects, while heartbeat signals were detected in approximately half. These findings suggest that the proposed method is effective for respiration monitoring, although further improvements are necessary for robust and stable heartbeat detection. This study was conducted in cooperation with INOAC CORPORATION.

Keywords: Sensors and Transducers, Medical and Welfare Systems, Biological and Physiological Engineering
Abstract: Humans exhibit various biological rhythms, such as heartbeat and respiration, which indicate autonomic nervous system activity and physical condition. Recently, non-contact, non-restrictive measurement methods using tactile sensors have gained significant attention as they eliminate the need for body-attached electrodes or devices. This study investigates the feasibility of measuring respiratory and heart rates in a seated posture using a sheet-type tactile sensor. In the experiments, the sensor was attached to a chair sheet to acquire pressure signals at 20 Hz from five adult male subjects across three backrest angles. Simultaneously, electrocardiogram (ECG) signals were recorded to evaluate the accuracy of the sensor-estimated heart rate. The respiratory rate was estimated from the peak frequency within the respiratory band via FFT analysis, whereas the heart rate was determined from the intervals between sharp peaks in the pressure signals. The results demonstrated that while respiratory signals were stably detected under most conditions, heart rate estimation accuracy varied depending on the subject and posture.

Keywords: Mechanical Systems Control, Medical and Welfare Systems, Biological and Physiological Engineering
Abstract: This paper addresses a trajectory tracking control problem in a teleoperated Functional Electrical Stimulation (FES) system designed for remote rehabilitation therapy. In remote healthcare settings, enabling therapists to control a patient's paralyzed limbs via network commands is important for effective treatment. However, patient-specific muscle characteristics and posture-dependent torque variations introduce nonlinear model errors that degrade tracking performance. To tackle this issue, we propose a novel FES control framework that incorporates a pre-learned muscle force model to improve trajectory tracking performance in the presence of posture-dependent muscle force variations. The pre-learned model utilizes experimentally obtained muscle force, arm angle, and pulse width. The usefulness and practical performance of the proposed approach are evaluated through experimental verifications.

Keywords: Modeling, System Identification and Estimation, Biological and Physiological Engineering, Medical and Welfare Systems
Abstract: A central challenge in neuroscience is that we can only record a small fraction of the brain, yet we want to model its activity and estimate its states from such limited data. This poster takes a control-theoretic view of this problem. We discuss how Hankel Dynamic Mode Decomposition (Hankel DMD) provides a principled route to recover hidden modes from such observations. By stacking delayed copies of the recorded signal, Hankel DMD constructs a subspace that approximates the Koopman operator, exposing dominant eigenmodes. We then connect this methodology to clinical settings such as brain-computer interfaces (BCIs) for patients with amyotrophic lateral sclerosis (ALS), where only a few electrodes are usable, signals are noisy, and the underlying neural processes change as the disease progresses. We outline how Hankel DMD-based modeling could enable BCIs that adapt to the changing neural dynamics of individual patients.

Keywords: Modeling, System Identification and Estimation, Electrophysiology and Kinesiology, Medical and Welfare Systems
Abstract: In recent years, the quantification of body movement has been increasingly emphasized in the fields of sports and rehabilitation. Inertial measurement units (IMUs) have attracted significant attention as a motion measurement method for calculating joint angles, which are standard quantitative indicators. The calculation of joint angles using IMUs is performed based on the relative relationship between the coordinate systems of adjacent segments. Therefore, to obtain interpretable joint angles, it is essential to attach one or more IMUs to each segment and identify the relative relationship between the IMU coordinate system and its corresponding segment coordinate system. Previous studies have defined this relationship by performing rotational movements around the target joint axes and estimating the rotation axes from the resulting angular velocities. However, it has been pointed out that estimation accuracy decreases when the calibration movement does not follow a pure single-axis motion. To address this issue, this study proposes a rotation axis estimation method that is applicable to rotational movements involving multiple axes by incorporating a kinematic model in addition to IMU data during calibration. The estimation accuracy of the proposed method was verified using the thumb carpometacarpal (CM) joint, where performing strict single-axis movements is particularly difficult．

Keywords: Human Interfaces, Medical and Welfare Systems, Ubiquitous Healthcare
Abstract: Our previous pilot study used airborne ultrasound phased arrays to deliver stroking stimulation to the forearms during short nap sessions, while EEG and ECG were recorded to assess responses related to sleep. The results suggested a possible reduction in sleep onset latency in some participants, but the system was limited to forearm stimulation and required a relatively constrained sleeping posture.

To overcome these limitations, we developed a large arched AUTD matrix for whole body mid-air haptic stimulation. The matrix is arranged around the sleeping space in a curved configuration, allowing ultrasound field to be generated over a broader body area while keeping the participant physically unobstructed. In this research, we describe the transition from a feasibility system focused on forearm stimulation to a whole body stimulation platform that allows more flexible sleeping postures. And the hardware arrangement, system concept, and future strategy for designing tactile stimuli in sleep studies. By expanding the stimulation area and reducing postural constraints, this platform provides a basis for investigating tactile patterns with low arousal that are optimized for sleep support.

Keywords: Human Interfaces, Medical and Welfare Systems, Ubiquitous Healthcare
Abstract: In recent years, issues surrounding hyperacusis have gained significant attention. Hyperacusis is defined as a condition in which individuals experience an abnormally painful loudness in response to sounds that are well-tolerated by most people. Although various symptoms have been reported beyond merely feeling distress from environmental sounds in daily life, many aspects of its actual state remain unclear, and individual variations are substantial. Consequently, it is difficult to accurately grasp the specific sound characteristics and environmental factors that contribute to the condition. Currently, as public awareness of developmental disorders such as Autism Spectrum Disorder (ASD) and Attention-Deficit/Hyperactivity Disorder (ADHD) grows, issues of sensory hypersensitivity and hyposensitivity are becoming more prominent, with hyperacusis emerging as a particularly serious concern as a representative auditory symptom. This study investigates a system designed to support cognitive behavioral therapy for hyperacusis.

Keywords: Biological and Physiological Engineering, Medical and Welfare Systems, Human-Machine Systems
Abstract: While joint state estimation using electromyography (EMG) has been widely explored for wearable assistive devices and prosthetic limbs, explicitly accounting for the statistical properties of both states and observations may further improve estimation framework design. In this study, we derived equations that express muscle activities as functions of joint kinematics and stiffness based on a musculoskeletal model, and designed an extended Kalman filter to estimate these states using the derived equations as the observation model. This approach enables probabilistic joint state estimation by incorporating the statistical properties of both states and observations. Experimental validation using surface EMG signals demonstrated the successful simultaneous estimation of wrist joint position, velocity, and stiffness.

Keywords: Adaptive and Optimal Control, Components and Devices, Medical and Welfare Systems
Abstract: Silicon photomultipliers (SiPMs) are ideal for portable real-time PCR devices due to their high gain and compactness, yet their performance is highly susceptible to temperature fluctuations. To address this, we propose a novel adaptive compensation technique using a pair of solid Polymethyl Methacrylate (PMMA) polymer reference samples. Unlike traditional liquid references, this solid-state method eliminates complex hardware while effectively decoupling the SiPM gain from ambient temperature changes. Experimental results demonstrate that the proposed model maintains SiPM output fluctuations within 1%​ even under temperature variations exceeding 10°C. Furthermore, accelerated aging tests confirm the long-term stability of the PMMA references. This work provides a low-cost, robust, and practical calibration solution for enhancing the reliability of grassroots molecular diagnostics.

Keywords: Human Interfaces, Medical and Welfare Systems, Biological and Physiological Engineering
Abstract: Estimation of joint motion from electromyography (EMG) has been widely explored as a non-invasive interface for wearable devices and cursor control. While many existing approaches rely on nonlinear neural networks that achieve high accuracy, their complex structure leads to high computational cost and limited interpretability. In contrast, the linear Kalman filter offers a probabilistically grounded and interpretable framework, but its application to non-invasive wrist angle estimation remains has not been sufficiently investigated. In this study, we designed a Kalman filter that decodes forearm surface EMG signals into the angles of two degrees of freedom of a wrist: flexion/extension and radial/ulnar deviation. The state vector comprises the two joint angles and their angular velocities, and the observation and transition matrices are identified from training data by least squares. Experimental validation using simultaneously recorded EMG and goniometer signals demonstrated that the Kalman filter was able to reconstruct continuous wrist joint angles, confirming the feasibility of interpretable, non-invasive joint angle estimation.

Keywords: Robotic and Automation Systems, Intelligent Control, Biological and Physiological Engineering
Abstract: Current AI paradigms, including Large Language Models and Physical AI, share a common assumption: intelligence is first constructed in cyber-space through large-scale data, then compressed into smaller models — what we term compressive distillation. This approach faces two fundamental problems: large intelligence must be built first, and there is no guarantee that genuine principles are embedded rather than statistical correlations. We propose AI-ed Physics as an alternative, grounded in how biological systems develop intelligence. Living organisms never begin with large intelligence. Instead, intelligence seeps out from the interaction between body and environment — gravity, friction, contact, and physical constraints — and larger cognitive structures gradually wrap around this core. Crucially, the source of this intelligence is guaranteed by physical law: gravity always acts, contact forces always arise, dissipation always occurs. True machine intelligence should follow this biological order: not distilling intelligence from data, but letting it seep from embodied physical interaction.

Keywords: Intelligent Control, Mechanical Systems Control, Robotic and Automation Systems
Abstract: This paper addresses the improvement of maneuverability by refining the passive wheel of a differential-drive vehicle in the AI-Formula challenge. The caster-type passive wheel used in the AI-Formula vehicle provided by Honda R&D Co., Ltd., exhibited significant rotational resistance around the yaw axis, which hindered path-tracking performance. This paper proposes a passive wheel unit utilizing omni-wheels. While the omni-wheels improved maneuverability, the resulting reduction in lateral constraint led to degradation of straight-line stability, causing frequent weaving and spinning. To resolve these issues, this paper proposes a passive wheel mechanism that also incorporates measurement wheels for estimating the vehicle’s pose. Steering experiments revealed that the instability originates from the rearward center of gravity (COG), which generates an oversteer-inducing yaw moment during cornering. By shifting the COG forward, the positioning error was reduced by approximately 92%. Furthermore, the spin limit was evaluated based on a measured limit yaw moment, and a cornering test at 20 km/h with a turning radius of 15 m confirmed that no spin occurred, as predicted. However, an oversteer tendency remained, demonstrating that mechanical COG adjustment alone is insufficient and that feedback control utilizing the proposed measurement wheels is required for accurate path tracking in the high-speed range.

Keywords: Signal and/or Image Processing
Abstract: White-line detection is important for autonomous driving in the AI-Formula competition. Conventional methods based on the Hough transform often produce discontinuous results under changes in illumination and road conditions. In this study, white-line extraction is formulated as a binary semantic segmentation problem, and an encoder-decoder convolutional neural network is applied. A composite loss function combining weighted cross-entropy and Focal Tversky Loss is introduced to address class imbalance. Experimental results showed that the proposed method can extract white-line regions from forward-facing camera images under class-imbalanced conditions.

Keywords: Nonlinear Control, Multivariable Control, Robotic and Automation Systems
Abstract: This study proposes a cooperative control method for automated automobiles to achieve smooth and safe traffic flow at highway merging sections. Applying principles from air traffic control, an optimization problem was formulated using Mixed-Integer Linear Programming (MILP) that considers the trade-off between merging time and fuel efficiency. Simulation results demonstrate that flexible merging control, tailored to the priority between merging time and fuel efficiency, can be achieved simply by adjusting a single weighting factor w. Specifically, it was confirmed that the system allows for a qualitative switch from an aggressive cut-in strategy via acceleration to a rear-entry strategy via deceleration.

Keywords: Adaptive and Optimal Control, Transportation Systems, Modeling, System Identification and Estimation
Abstract: In this study, a vibration modeling approach is proposed to realize vehicle control that suppresses cargo vibrations on a truck bed. A neural network-based vibration model is constructed to estimate the vertical acceleration and pitch angular velocity of the cargo bed from the vehicle control inputs and state variables. To obtain training data, driving experiments are conducted using a truck model car. Finally, the predicted results of the neural network are compared with experimental data to evaluate the effectiveness of the proposed vibration model.

Keywords: Transportation Systems, Adaptive and Optimal Control, Modeling, System Identification and Estimation
Abstract: This study addresses highway forced merging for autonomous vehicles (AVs) in mixed traffic with human-driven vehicles (HVs). In conventional research, Bayesian inference was employed to estimate the driving-style parameters of mainline drivers, allowing their possible behaviors to be predicted and an optimal merging trajectory to be planned while accounting for all candidate behaviors. However, this strategy is computationally expensive. To enable real-time implementation, which remains a major challenge in traffic engineering, a hierarchical control architecture is proposed in this study to reduce computational cost. Specifically, the upper layer optimizes the merging time and target gap, and the lower layer computes an optimal cubic-polynomial trajectory to track the selected strategy with minimal deviation. Using data collected from 20 computer-generated and 20 real-world merging maneuvers, both the conventional and proposed methods were applied to generate optimal trajectories for the merging vehicle in all 40 merging maneuvers. The results show that the hierarchical controller substantially reduces computation time compared with the conventional method and also improves the smoothness of the post-merge behavior of the merging vehicle.

Keywords: Robotic and Automation Systems, Mechanical Systems Control, Mechatronics Systems
Abstract: This study presents a dragonfly-inspired flapping robot driven by micro servo motors and investigates the effects of flapping frequency and angle on lift and pitch stability. The robot employs four independently driven wings with sinusoidal control and a phase difference between forewings and hindwings. Experiments show that higher flapping frequency increases peak lift but not necessarily average lift, while flapping angle strongly influences both lift and stability. These results highlight the importance of optimizing motion parameters for efficient flapping flight.

Keywords: Biological and Physiological Engineering, Analytical Measurement, Electrophysiology and Kinesiology
Abstract: A prosthetic leg with a variable stiffness mechanism is an artificial limb that replaces the missing part of the leg to support the mobility of users. However, a commonly used prosthetic leg does not include the mechanism that changes the rigidity during different walking situations, which can cause long-term health issues. Variable stiffness mechanisms are designed to solve the following problems. Many studies have focused on the hardware of the variable stiffness mechanisms, but to achieve a complete prosthetic leg, the control system is required to respond with movement intent precisely and naturally by applying electromyography (EMG) sensors at the walk-related muscles to gather data and find relationships between muscle activation patterns and gait cycle sequences. The results from the experiment and analysis suggest that five sequences of the gait cycle have relationships and can be defined by muscle activation patterns, while the remaining three sequences showed no significant relationship and required further study.

Keywords: Robotic and Automation Systems, Mechatronics Systems, Modeling, System Identification and Estimation
Abstract: The relationship between observability and estimation accuracy in mobile robot localization has not been fully established. This study examines this connection through the information matrix obtained from the linearized observation model. The analysis shows that its positive definiteness corresponds to local observability and that the minimum eigenvalue is related to the estimation error covariance. Two prediction models within an Unscented Kalman Filter (UKF), a kinematic model based on control inputs and a Visual Odometry (VO) model using frame-to-frame motion, are evaluated under slip disturbances. Although the observation geometry remains unchanged, estimation accuracy is strongly influenced by model mismatch, and the VO model provides improved robustness. The minimum eigenvalue exhibits a threshold-like behavior, indicating potential degradation rather than a continuous performance measure.

Keywords: Mechanical Systems Control, Nonlinear Control, Robust Control
Abstract: This paper presents a control strategy for a specialized "half-drone wheeled inverted pendulum" designed to navigate structured indoor environments, such as stairs and standardized doorsteps. The control architecture integrates feedforward compensation with a robust feedback law synthesized through a polytopic LMI (Linear Matrix Inequality) approach. While the feedforward term accounts for the nominal system dynamics and thrust requirements, the polytopic feedback ensures stability across the highly nonlinear operational workspace by modeling the system within a convex hull of parametric planes. Gazebo simulations demonstrate that this dual-layered control effectively manages the transition from horizontal movement to step climbing, successfully surmounting obstacles up to 14 cm. The results confirm the system's ability to recover from initial state errors and emphasize the critical role of thrust saturation in maintaining tracking performance during geometric transitions.

Keywords: Robotic and Automation Systems, Nonlinear Control, Mechatronics Systems
Abstract: This paper discusses simulation results under actuator constraints for an omnidirectional wheeled mobile robot (Omni-WMR) equipped with steered active single-wheel caster mechanisms. The dynamic control allocation problem is formulated as a constrained quadratic programming (QP) problem with input saturation and rate constraints. We detail the transformation from physical actuator constraints (forces, steering angles, and their rates) to constraints on the auxiliary variables used in the optimization. Simulation results for three different weight designs show that increasing the number of past control steps in the cost function reduces energy consumption while satisfying constraints, albeit with a slight increase in tracking error. The effectiveness of the proposed constrained allocation approach is demonstrated through numerical simulations.