Keywords: Robotic and Automation Systems, Network System Integration
Abstract: As robots take on critical roles in logistics, disaster response, and infrastructure, their dependence on remote communication channels exposes them to a dangerous and largely overlooked class of cyberattacks. This talk presents a framework revealing that the kinematic and dynamic symmetries engineers have long exploited to simplify robot design and control create provably undetectable vulnerabilities. This talk shows that an attacker can apply coordinated affine transformations to both sensor measurements and control commands, inducing arbitrary deviations in the robot’s physical behavior while rendering all monitoring systems blind to the intrusion, requiring neither cryptographic compromise nor deep system knowledge. This talk further demonstrates that the same structural framework that characterizes the attack also reveals its detection signature: by designing state monitoring functions that deliberately break the attacker’s symmetry, optimal detection becomes achievable. Findings call for a fundamental shift toward security-aware control architecture design in cyber-physical robotic systems. The research, supported by NEDO, will also be embedded within an industry–university collaborative department established by Kobe University and Kawasaki Heavy Industries to integrate advanced robotics research, doctoral education, and real-world technology implementation.

Keywords: Information Management Systems
Abstract: This lecture presentation provides an overview of the scope, use cases, standardization and technical background of the Asset Administration Shell (AAS), which is an approach of an interoperable digital twin in the fields of smart manufacturing and beyond. In the standardization part, the relation to IEC 61987 (list of properties), IEC 62832 (digital factory) (see the SICE2023 presentation about digital factory), IEC 61406 (identification link) and IEC 63278 (AAS) will be described and the use organization IDTA (industrial digital twin association) will be introduced. Typical use cases of the AAS are provided, before a technical overview of the AAS meta model and tool chain is given, including authoring and repository tools (AASX Package Explorer and AASX Server). Finally, interfacing with IEC 62541 (OPC UA) is covered as well as the approach of industrial data spaces in value chains.

Keywords: Networked Sensor System, Analytical Measurement
Abstract: This paper investigates a sensor placement problem for distributed control systems under stealthy cyber attacks. We derive sufficient conditions under which the impact of such attacks remains bounded based on invariant zeros and graph-theoretic properties of the network. In particular, these conditions can be characterized in terms of a vertex cover, which significantly reduces the number of candidates for sensor placement. The effectiveness of the proposed approach is demonstrated through simulations on a DC microgrid system. Furthermore, we show that the vertex-cover-based condition substantially reduces the number of candidate sensor placements.

Keywords: Biological and Physiological Engineering, Human Interfaces, Medical and Welfare Systems
Abstract: Traditional affective computing has focused on decoding emotional states from individual physiological signals. Recently, interpersonal physiological synchrony measured via inter-subject correlation (ISC) has emerged as an objective marker for engagement. However, most existing studies rely on pulse intervals, potentially overlooking the rich morphological information embedded within the photoplethysmogram (PPG) waveform. This study investigates the ISC of six pulse-wise morphological features among 24 participants viewing 12 music videos (MVs). Our results reveal that viewing identical MV content induces significant synchronization not only in pulse rate but also in multiple morphological features. Interestingly, for the Max rise slope (the steepest upward slope of the pulse), the overall group-level synchronization was not significant. However, an inter-subject representational similarity analysis (IS-RSA) revealed that the synchronization patterns of this specific feature exhibited the strongest correspondence with individual differences in subjective emotional similarity. These findings suggest a functional dissociation between different PPG features. This study highlights the potential of PPG morphological synchronization as a novel complementary tool for affective computing, enabling systems to better understand and facilitate the interconnected human experience by leveraging accumulated data from groups of individuals.

Keywords: Transportation Systems, Robotic and Automation Systems, Signal and/or Image Processing
Abstract: This study proposes a self-localization method that uses onboard cameras as an alternative to short-range LiDAR-based localization, which constitutes one component of redundant localization systems in automated driving. Although camera-based self-localization offers advantages such as lower cost and improved system redundancy, its estimation accuracy can deteriorate under varying environmental conditions due to its sensitivity to local brightness changes. To address this issue, this study introduces the Census transform and Hamming distance into the similarity calculation so that local structural information can be utilized instead of raw intensity values. Experimental results obtained from real-vehicle driving tests demonstrate that the proposed method suppresses degradation in localization accuracy even under strong sunlight and nighttime conditions, where local brightness variations are significant.

Keywords: Intelligent Systems, Process Control, Computational Intelligence
Abstract: Accurate prediction of the remaining useful life (RUL) of laptop hinges, defined based on torque degradation at a predefined health indicator (HI) threshold, is critical during the design-for-manufacturing (DFM) stage. However, conventional evaluation approaches largely rely on empirical experience, often leading to inaccurate and inefficient assessments. To address this limitation, this study proposes a data-driven framework called Pressure-Marked Attention-Based Bidirectional Gated Recurrent Unit (PM-AB-BiGRU). The proposed framework explicitly incorporates the pressure value per unit friction surface area (pressure value) as a conditional feature to characterize hinge torque degradation behaviors under different loading conditions. A sequence-to-sequence BiGRU architecture is employed to model temporal dependencies in hinge torque degradation, while an attention mechanism is used to adaptively emphasize critical degradation features associated with varying pressure values. Experimental results based on hinge torque degradation datasets provided by Company S demonstrate that the proposed PM-AB-BiGRU effectively captures degradation trends under different pressure conditions and accurately predicts the corresponding hinge RUL.

Keywords: Computational Intelligence, Intelligent Systems, System Engineering
Abstract: This paper proposes a surrogate-assisted evolutionary algorithm for discrete black-box optimization problems. While uncertainty-aware methods are well established for continuous optimization, handling uncertainty in discrete decision spaces remains nontrivial. To address this gap, we utilize the prediction variance of Random Forest surrogates as an uncertainty measure, leveraging their suitability for discrete variables to enhance global search performance. Using black-box optimization experiments on small-scale 0--1 integer programming problems, we study the uncertainty weight and quantify its impact through statistical hypothesis testing toward scalable surrogate-assisted optimization.

Keywords: Computational Intelligence, Intelligent Systems, System Engineering
Abstract: This paper proposes an adaptive constraint-switching multimodal multi-objective evolutionary algorithm (ACSMMEA), a novel approach for solving constrained multimodal multi-objective optimization problems (CMMOPs). Although some existing algorithms employ two search strategies: constraint-free and constraint-aware search, their transition timing is dictated by a static parameter, necessitating problem-specific tuning. To address this limitation, ACSMMEA introduces an adaptive switching strategy that monitors population convergence in the decision variable space and transitions from constraint-free to constraint-aware search upon detecting convergence. Furthermore, the proposed method incorporates an ϵ-constraint method that gradually tightens constraint tolerance after switching, allowing information from promising solutions acquired in the constraint-free search phase to be inherited and utilized in the constraint-aware phase. Experimental results on 14 CMMOP benchmark problems demonstrate that ACSMMEA effectively maintains solution diversity in the decision space while accurately approximating the Pareto front, exhibiting performance comparable to or better than state-of-the-art methods.

Keywords: Sensors and Transducers, Mechanical Systems Control
Abstract: In this study, we achieved the levitation and manipulation of lightweight objects larger than the wavelength by using multiple ultrasonic Laguerre-Gaussian (LG) beams with a common focal point. A phased array composed of ultrasound speakers was used to generate the sound field, and both the sound field and the focal position were electronically controlled. We demonstrated that multiple LG beams complement the sound field, enabling more stable levitation and manipulation. This method was experimentally verified; using 40 kHz ultrasound, we demonstrated the ability to levitate a 20 mm diameter expanded polystyrene sphere and manipulate its movement horizontally and vertically at speeds of up to 100 mm/s.

Keywords: Robotic and Automation Systems, Mechanical Systems Control, Adaptive and Optimal Control
Abstract: Animals generate adaptive and efficient gaits in complex environments by integrating proprioception from their musculoskeletal systems. This paper investigated a control system for the bounding gait of a quadruped robot equipped with feline-inspired elastic leg mechanisms. We evaluated the effectiveness of a Central Pattern Generator (CPG) controller incorporating elastic force feedback, as well as the impact of toe trajectory parameters on locomotion performance. Through dynamic simulations and physical experiments, we demonstrated that the proposed control scheme autonomously adjusted inter-limb phases, resulting in enhanced locomotion speed and energy efficiency.

Keywords: Robotic and Automation Systems, Factory Automation
Abstract: When AMRs operate in a smart factory, they often encounter keep-out zones, namely areas that are unsuitable or restricted for robot entry. These zones cannot always be detected by onboard sensors and may frequently change due to factory rearrangement or site management requirements. In such situations, neither remapping nor directly editing the navigation map is an economical solution, as both approaches incur additional time and cost. To address these issues, this study proposes a Keep-Out-Aware AMR Navigation System. The system integrates layout–SLAM map fusion, real-time importation of keep-out zones into the global costmap, and a virtual LiDAR signal generation mechanism, allowing manually defined restricted areas to influence both global path planning and MPC-based local path planner. Through this approach, engineers can define and adjust keep-out zones from existing layout drawings or manual point-based marking, reducing the need for remapping and improving site management flexibility. The demonstration results show that the proposed system enables effective alignment between the layout and the SLAM map for keep-out zone transfer. Virtual LiDAR synthesis incurs a computational delay of less than 0.015 s, which is far below one LiDAR scan cycle. The system can also effectively prevent AMRs from entering restricted spaces that are difficult for onboard sensors to perceive while maintaining stable dynamic obstacle avoidance.

Keywords: Intelligent Systems, Computational Intelligence, Adaptive and Optimal Control
Abstract: Association rule mining and class association rules have been widely used to extract meaningful subgroup characteristics from database. However, existing approaches are limited in their ability to represent statistically distinctive subgroup behavior when multiple categorical variables must be treated jointly, particularly when the underlying class probability distribution is required rather than a single target class outcome. This limitation creates a gap in expressive yet interpretable pattern representations for multi-variable categorical distribution analysis. This study proposes itemset with class probability distribution (ICPD), which is an itemset representation in which each itemset is associated with the probability distribution of categorical variables as its statistical background. ICPDs capture the overall class distribution underlying a target class. Furthermore, ICPDs enable representations in which multiple categorical variables are treated as target variables and their joint probability distributions. This study focuses specifically on individuality-based ICPD, and presents methods for discovering such patterns and for analyzing data using the discovered ICPDs.

Keywords: System Engineering, Computational Intelligence, Manufacturing Systems
Abstract: Two-stage stochastic programming is useful for production scheduling under uncertain troubles and delays, but the number of scenarios becomes very large and computation takes too much time. We model a metal recycling plant as a mixed-integer program and extend it to a two-stage stochastic program with uncertain delays and overtime. To reduce computation time, we sample a small set of scenarios, but random sampling does not always give good solutions. We propose a framework that selects scenarios from all candidates using metaheuristics, and we develop methods based on Simulated Annealing (SA) and a Genetic Algorithm (GA). Using data from an actual plant, we compare SA, GA, and random sampling by re-evaluating the expected objective value over all scenarios.

Keywords: Guidance and Flight Control, Autonomous Decentralized Systems, Intelligent Systems
Abstract: This paper studies real-time tactical missile evasion against a faster interceptor by treating speed as a maneuverability resource. Because minimum turn radius varies quadratically with speed, temporary deceleration can provide additional turning authority when it is more valuable than forward progress. The resulting finite-horizon nonlinear optimal-control problem is solved in receding-horizon form by Sequential Convex Programming (SCP), with the speed profile in the decision vector and an aerodynamic load-factor constraint coupling admissible lateral acceleration to instantaneous speed. Under the stated regularity and penalty assumptions, the SCP iterates converge locally to a first-order stationary point. Across four engagement scenarios the method succeeds in 100% of cases, versus 75% for Augmented Proportional Navigation, 50% for reactive, and 25% for geometric evasion; in 100 randomized engagements it achieves 72% and 60% success against proportional-navigation and augmented-proportional-navigation interceptors, respectively. On the tested desktop-class Intel i7-9700K using one thread, the mean update time is 12.3 ms (81 Hz), supporting real-time computational feasibility on that platform.

Keywords: Adaptive and Optimal Control, Nonlinear Control, Modeling, System Identification and Estimation
Abstract: Model Predictive Control (MPC) is an effective feedback control method for various engineering applications. MPCs using gradient-based optimization have difficulty handling collisions due to discontinuous dynamics. Monte Carlo MPC (MCMPC) can handle collisions through sampling-based optimization without gradient information. Research on parameter adaptation in MCMPC remains insufficient, and developing an MPC framework capable of handling both collisions and model parameter uncertainty remains a challenge. We propose Sampling-based Moving Horizon Estimation (SMHE) by applying the MCMPC algorithm to parameter estimation, and develop Sampling-based Adaptive MPC (SAMPC) by combining SMHE with MCMPC. This paper experimentally validates SMHE and SAMPC on swing-up and stabilization of a pendulum on a cart with cart stoppers under large initial model parameter errors. SMHE achieves a weighted RMSE averaging 73% of that of a joint Extended Kalman Filter (EKF) with collision data rejection. SAMPC achieves higher control performance and adaptability than the joint-EKF-based adaptive MCMPC.

Keywords: Power Systems Control, Robust Control, Adaptive and Optimal Control
Abstract: This study proposes a resilient distributed load frequency control (LFC) method for frequency stability and generation--cost optimality under False Data Injection Attacks (FDIAs). The proposed method introduces an Independent System Operator (ISO) that coordinates supply--demand mismatch and marginal costs, while a dynamic compensation layer mitigates the influence of FDIAs without explicit attack detection. Theoretical analysis shows that a suitable design gain can bring the closed-loop equilibrium close to the optimal solution and guarantee practical stability with respect to the generation commands. Numerical simulations using the IEEJ EAST 30--machine system model verify the effectiveness of the proposed method.

Keywords: Computational Intelligence, Intelligent Systems
Abstract: This study focuses on crime action recognition using surveillance camera footage captured in urban environments, with the objective of detecting and classifying criminal activities from a limited number of anomalous video samples. To achieve this, we propose a method based on the self-supervised learning model VideoMAE (Video Masked Autoencoder), combined with Transformer layers that capture temporal relationships across entire video sequences, thereby enabling effective extraction of spatiotemporal features specific to criminal behaviors. The novelty of the proposed method lies in integrating general-purpose video representations learned through self-supervised learning into a few-shot learning framework. In particular, Transformer layers are introduced to explicitly model temporal feature relationships and are combined with a distance metric-based classification framework to improve crime action recognition performance under limited data conditions. This approach enables the discrimination of criminal activities across different domains, which is difficult for conventional average-pooling-based feature representations and simple few-shot classification methods. The usefulness of the proposed method is its applicability to real-world environments where labeled crime data are extremely limited. In particular, the method can learn from only a few samples of rare criminal incidents, allowing flexible adaptation to newly emerging crime classes. To evaluate the effectivenes

Keywords: Robotic and Automation Systems, Intelligent Systems, Factory Automation
Abstract: Inspection of finished products is essential for ensuring product quality in manufacturing. Although visual inspection based on geometric information is widely used, some abnormalities are difficult to detect from appearance alone. To address this issue, this paper proposes a haptic inspection method using a high-speed multi-fingered robot hand for versatile use in high-mix low-volume production. The system grasps and excites a target component, and detects anomalies from fingertip force data acquired at 1,000[Hz]. As a case study, an inspection was conducted for a simple component consisting of two blocks fastened by a screw, in which the screw condition was classified as either loosened or fixed. Analyses in the time, frequency, and cepstral domains revealed differences in peak-to-peak value, peak-frequency characteristics, and spectral-envelope-related features. A SVM-based model using features selected by statistical tests and SVM-RFE achieved 92.5% accuracy in real-time inspection experiments. These results indicate that the features examined in this study, such as peak-to-peak value and LFCC, are effective for inspections of screw loosening and similar defects, and also demonstrate the usefulness of robot-based part inspection based on force sensing.

Keywords: Robotic and Automation Systems, Intelligent Systems, Intelligent Control
Abstract: Automating electrical facility inspection requires robots to operate diverse equipment autonomously. Conventional imitation learning faces challenges with complex architectures and large-scale data collection. To address these, we propose a high-precision one-shot imitation learning method using trajectory transfer. While conventional transfer suffers from accuracy degradation in cluttered environments, our approach leverages prior knowledge of control panels. By employing component scooping to identify individual components, the system achieves robust estimation and task execution. Experimental results with emergency stop buttons and circuit breakers validated the estimation accuracy and demonstrated the feasibility of autonomous equipment operation.

Keywords: Robotic and Automation Systems, Intelligent Systems, Human-Machine Systems
Abstract: 人間からロボットへの模倣学習は課題です 模倣を適用する際に再訓練が必要な場合に発生します 別のロボットに。本研究は単一のモデルを提案します 潜在的を使った人間から複数ロボットへの模倣学習 ImitationNetに基づく空間。さらに、 習得された潜在空間が 複数のロボットの動きにおける一貫した対応 人間とは異なる具現化を持つ。その 提案された方法は、学習効率を向上させることを目的としています。 人間の模倣学習

Keywords: Robotic and Automation Systems, Intelligent Systems, Intelligent Control
Abstract: This paper studies how ordinary monocular video can be converted into a navigation-oriented and executable scene representation for a Unitree Go2 quadruped. In the broader project, an RGB-D baseline using RTAB-Map, ROS 2 communication, occupancy-grid planning, and Go2 command execution had already been physically deployed. This paper develops a complete video-grounded pipeline. It first extracts traversable regions from monocular frames using SegFormer, then aligns these frame-wise predictions through ORB-SLAM3-based camera-motion estimation together with fallback odometry logic, and accumulates them into a global traversability map. The accumulated representation is further converted into an executable MuJoCo scene, where route planning, motion validation, and replay-based robustness experiments can be carried out in a closed-loop manner. The framework also includes a Go2–MuJoCo bridge that rejects motion updates violating scene geometry, as well as controlled degradation tests under harsh lighting, communication delay, jitter, and packet loss. Overall, the results show that ordinary monocular video can be transformed into a clear, executable, and diagnostically useful navigation representation. Compared with the previous obstacle-insertion formulation, the proposed traversable-region-first pipeline provides a more interpretable basis for global map generation, executable scene construction, and robustness-aware navigation validation.

Keywords: Robotic and Automation Systems, Intelligent Systems, Autonomous Decentralized Systems
Abstract: In multi-robot exploration, pheromone-based coordination can reduce revisits, but it relies primarily on local infor-mation and may leave isolated unexplored regions, particularly during the later stages of exploration. This limitation is especially important in disaster environments, where direct inter-robot communication may be unreliable or unavaila-ble. To address this problem, this paper proposes a cooperative exploration method that combines stigmergy with Gaussian process regression (GPR). Each robot uses locally observed pheromone information within its sensing region and the GPR predictive mean and uncertainty beyond that region to generate a synthesized motion vector. This ap-proach is designed to suppress revisits while providing motion guidance beyond the local sensing range. To evaluate the feasibility and effectiveness of the proposed method, computational simulations were conducted in a fixed obsta-cle-free two-dimensional field, with the initial positions of the three robots randomized in each of 10 trials. The pro-posed PGMU method completed exploration in fewer steps on average than the Pheromone-only method, although the difference was not statistically significant. These results suggest that combining the GPR predictive mean and uncer-tainty with pheromone-based guidance has the potential to improve exploration efficiency while maintaining complete exploration under communication-constrained conditions.

Keywords: Robotic and Automation Systems, Intelligent Systems
Abstract: This paper evaluates the scalability of the Spreading Activation Policy Network (SAP-net), a knowledge-selection method for transfer reinforcement learning, implemented on a Raspberry Pi 4B computing cluster. SAP-net selects policies via Spreading activation on a policy graph, but computational and communication overhead becomes a critical bottleneck as the number of policies increases. We implemented SAP-net on a 40-node cluster connected in star topology to a NetGear ProSafe GS752TS switch and evaluated reinforcement learning tasks in the Webots simulator under 80, 160, and 320 policies. Performance was assessed using task success rate, latency ratio, and power consumption. Results show a sharp degradation in success rate and a latency ratio reaching 98% at 320 policies, while power consumption saturates beyond 160 policies. This indicates that performance degradation is primarily caused by communication and synchronization delays in the cluster rather than hardware resource exhaustion. These findings highlight the importance of low-latency interconnects for maintaining temporal consistency and suggest practical advantages of Raspberry Pi-based edge clusters in cost- and power-constrained environments.

Keywords: Robotic and Automation Systems, Intelligent Systems, Guidance and Flight Control
Abstract: In this paper, we propose a hierarchical hybrid framework for safe Unmanned Aerial Vehicle (UAV) navigation that integrates low-frequency semantic reasoning with high-frequency control through a shared cost field. A Vision-Language Model (VLM) performs low-frequency semantic risk assessment, while a Model Predictive Path Integral (MPPI) controller generates trajectories based on a unified cost field that integrates both distance-based and anticipatory risks. We introduce a semantic grounding mechanism that converts VLM outputs into dynamic, anisotropic cost fields aligned with object motion, aiming at proactive avoidance of potential hazards.

Keywords: Robotic and Automation Systems, Intelligent Systems
Abstract: As the demand for automation in manufacturing grows, the need for deep-learning-based models for robot control increases. These models must achieve highly precise control while remaining robust to environmental variations. In this paper, we propose a learning method for robot control, called Grasping-Error-Aware Minimization, to improve robustness to grasping errors. We introduce a constraint into the loss function that enforces sufficient separation between feature representations of input images whose supervisory signals for robot actions differ significantly. We evaluated the proposed method on a peg-in-hole task under grasping-error conditions. The deep-learning-based model trained with the proposed method achieved the highest or tied for the highest success rate under all four unseen grasping-error conditions. These results indicate that the proposed method enables high-precision control even under unseen grasping errors.

Keywords: Robotic and Automation Systems, Human Interfaces
Abstract: Underwater robot development is often hindered by the complexities of waterproofing and wiring, which significantly delay the rapid prototyping process. This paper presents MarineCraft, a modular toolkit designed to accelerate the development cycle through structural reconfiguration. The system features self-contained, waterproof propulsion modules that integrate power, wireless communication, and actuation. By eliminating centralized wiring and the need for repeated sealing, MarineCraft allows diverse robot geometries to be assembled and tested in minutes rather than days. Experimental results demonstrate that this reconfigurable architecture enables fast, iterative design cycles while maintaining reliable operation and leak-free performance at depths of up to 2.5 meters. Our toolkit effectively lowers the barrier to underwater robotics by transforming modularity into a vehicle for rapid physical prototyping.

Keywords: Robotic and Automation Systems, Autonomous Decentralized Systems, Safety, Environment and Eco-Systems
Abstract: In logistics warehouses, the deployment of autonomous mobile robots (AMRs) that collaborate with human workers has been increasing. This paper proposes a hybrid fleet coordination architecture for multi-agent systems that combines distributed collision avoidance control with a supervisory layer that locally supports inter-agent coordination, achieving high throughput while ensuring safety and scalability. The proposed architecture is evaluated through simulation and compared with centralized and decentralized architectures. The results demonstrate that the proposed approach avoids the throughput degradation observed in centralized systems as the number of AMRs increases. In addition, limitations of decentralized architectures, particularly insufficient inter-agent coordination, are alleviated, and the occurrence of deadlocks is significantly reduced.

Keywords: Robotic and Automation Systems, Mechatronics Systems, Mechanical Measurement
Abstract: This paper presents a variable-reach robotic gripper with extendable flexible fingers in a fixed-base experimental setup. The gripper uses four slit straw-based fingers whose effective length is varied through a rolling and unrolling mechanism, allowing the end-effector to reach objects located at different distances from the palm without moving its base. During retraction, the fingers are compactly rolled, whereas during extension they recover an approximately tubular shape suitable for grasping. A second tension-driven mechanism bends the fingers inward to generate grasping forces. The study analyzes the operating principle of the gripper, introduces simplified analytical models for the extension--ribbon geometry and the pressing force, and experimentally validates two main functions. First, single-finger tests characterize the pressing force at different finger lengths and show a trade-off between reachable length and contact force. Second, grasping experiments demonstrate that the gripper can grasp objects positioned both near and far from the palm while maintaining a fixed base. The results support the feasibility of the proposed mechanism as a variable-reach gripper for constrained-access manipulation.

Keywords: Robotic and Automation Systems, Mechatronics Systems, Intelligent Control
Abstract: This paper presents the design and implementation of a closed-loop fuzzy logic control system for a flapping-wing flying robot. Flapping-wing flight exhibits highly nonlinear and uncertain dynamics, making traditional control methods challenging. The proposed approach utilizes fuzzy logic to handle these uncertainties, enabling attitude and trajectory control. An Inertial Measurement Unit (IMU) provides real-time feedback on orientation and motion, which is processed by the fuzzy controller to generate appropriate wing actuation commands. Experimental results demonstrate that the closed-loop fuzzy logic controller achieves stability, maneuverability, and responsiveness. The findings suggest that fuzzy logic control is a promising solution for bio-inspired aerial robots operating in dynamic environments.

Keywords: Robotic and Automation Systems, Mechatronics Systems
Abstract: Roof snow removal is a physically demanding task with a high risk of falls. While fixed, automated snow removal systems have been proposed, their application to existing houses is difficult. Therefore, this research proposes automating snow removal using a small robot. The proposed method employs a mechanism in which the robot burrows into the snow layer, excavates the snow with an auger, and moves forward through peristaltic motion. Based on the design concept of artificially inducing an avalanche to remove snow from the roof, a dynamic model was constructed that simul taneously satisfies the“slip condition” to prevent the robot itself from slipping on the roof and the“avalanche condition”to ensure that only the snow on the roof collapses. Furthermore, based on this dynamic model, design parameters were examined using multi-objective optimization. Finally, a prototype was manufactured based on the optimization results, and the operation of the prototype was verified.

Keywords: Robotic and Automation Systems, Mechatronics Systems, System Integration
Abstract: In recent years, the declining working population due to a low birthrate and aging society has heightened the importance of automation using robots for the maintenance and inspection of infrastructure facilities. Tasks such as operating and verifying breakers and switches on control panels, and inserting parts into each other necessitate the establishment of complex mechanical contact and grasping with the target components. This renders motion acquisition for automation and manipulation difficult in real world due to the risk of collision while testing. This research introduces the methodology for the development of a physics simulation environment specialized for maintenance and inspection of circuit breakers, and for acquiring motions using reinforcement learning. The reinforcement learning with the physics environment successfully generated autonomous manipulation policies for control panel circuit breakers.

Keywords: Robotic and Automation Systems, Networked Sensor System, Network System Integration
Abstract: In recent years, the declining population and aging of the core farmershave become issues in Japan. To solve this issue, autonomous mobile robots are being deployed in agriculture to support farmers. However, in overhead trellis environments, the deployment of mobile robots is difficult because satellite positioning systems and similar position estimation method cannot be utilized. Consequently, autonomous mobile robots have not been deployed in overhead trellis environments, and crop transport tasks remain dependent on manual labor. To solve this issue, we proposed a harvest transportation system utilizing autonomous mobile robots. The mobile robot autonomously navigates in the overhead trellis environments. The workers follow the robot from behind and loads the harvested crops onto its cargo bed. This system employs UWB for the mobile robot’s position estimation. Therefore, we must define UWB positioning methods for the mobile robots. This study proposes a TWR-based UWB positioning and navigation method with lightweight packet design for implementation in autonomous mobile robots. The proposed method defines the packet format between UWB devices and describes specific methods for applying the system to mobile robots. In this experiment, a mobile robot performed autonomous navigation utilizing the proposed TRW-based UWB positioning system. The results showed that the proposed UWB positioning system is effective for the autonomous navigation of mobile robots.

Keywords: Robotic and Automation Systems, Networked Sensor System, Network System Integration
Abstract: We propose an autonomous navigation method for mobile robots to automate floor cleaning tasks in factory environments. In factories handling large metal products, it is difficult to use conventional position estimation methods, which existing cleaning robots applied due to environment-specific constraints. In previous work, Ultra-wideband (UWB) positioning was adopted as a method of applicable to the environment. In UWB positioning, a tag is a node to be estimated, and an anchor is a node with known coordinates. For position estimation of mobile robot, tags are mounted on the robot, and its position is estimated by multilateration using UWB ranging measurements between the tags and anchors and the anchor coordinates. When UWB positioning is used in this environment, non-line-of-sight (NLOS) conditions caused by metal structures reduce the performance of the positioning. To avoid NLOS conditions, anchors must be installed according to obstacle arrangements. However, accurate determination of anchor coordinates increases workload and limits practical deployment. To solve this issue, we propose autonomous navigation method without anchor coordinates by utilizing relative coordinate between tags and ranging measurements between each tag and anchor. The experiments using motion capture showed the effectiveness of the proposed method, as the robot completed the specified path while maintaining an error of less than 1.0 m, with a mean error of 0.115 m from the path.

Keywords: Robotic and Automation Systems, Networked Sensor System, Network System Integration
Abstract: This paper proposes a consistency-based relative state estimation method for multi-robot flocking under sensor failures. Each robot is assumed to measure only the relative position of neighboring robots, and relative orientations are estimated from observed motion. The proposed method integrates multiple relative state estimates using weights derived from mutual consistency among observations. The method does not require explicit fault detection or predefined fault models. Numerical simulations compare the proposed method with raw observations and a median-based robust fusion method under stuck, noise-increase, and drift faults. The results demonstrate that the proposed method provides balanced robustness across different fault models and maintains flocking behavior without identifying faulty sensors.

Keywords: Robotic and Automation Systems
Abstract: This paper reports on the experimental validation of remote positioning and heavy object handling technologies for fusion devices. Specifically, this paper addresses the Blanket Remote Handling System (BRHS) for ITER, the international experimental fusion device, which is used for the maintenance and replacement of Blanket Modules (BMs) installed on the inner wall of the ITER vacuum vessel. The ITER BRHS requires high-precision handling of heavy objects using a large‑scale manipulator under severe constraints, such as those imposed by the high‑radiation environment. To address these requirements, the authors developed and experimentally implemented key technologies for remote positioning and heavy object handling, including structural-simulator-based virtual reality, robot‑vision‑based positioning, and peg-in-hole and load transfer using force feedback. Experimental validation using a full-scale prototype manipulator confirmed the applicability of each technology and demonstrated the feasibility of the entire BM handling process from positioning to BM removal. These results establish a technical foundation for BM handling, particularly with respect to remote positioning and heavy object handling.

Keywords: Mechatronics Systems, Mechanical Systems Control, Modeling, System Identification and Estimation
Abstract: The purpose of this study is to clarify and control the deformation motion of a tri-star-type electric wheelchair capable of stair climbing and to realize its wheel-lift control, with particular attention given to the transition to a two-wheel-contact inverted stable mode.

To address this issue, the deformation principle is divided into two phases: the motion from a four-wheel-contact state to the moment when the wheels are lifted, and the motion that transitions from a two-wheel-contact state to an inverted posture.

Based on this principle, a control system that coordinates the wheel and seat mechanisms is constructed. The effectiveness of the proposed approach was verified through simulations, which confirm smooth wheel lifting with suppressed acceleration. Furthermore, the feasibility of the wheel-lifting motion was demonstrated through experiments.

Keywords: Mechatronics Systems, Robotic and Automation Systems, Adaptive and Optimal Control
Abstract: This paper addresses the challenge of maintaining high trajectory-tracking accuracy and robustness for Unmanned Aerial Vehicles (UAVs) operating in turbulent environments. While Nonlinear Model Predictive Control (NMPC) can handle constraints and optimize performance, its reliance on fixed weighting matrices Q and R makes it susceptible to degradation under unpredictable disturbances and model uncertainties. To enhance robustness without sacrificing optimality, we propose an adaptive weighting NMPC in which the objective weights are updated online. The core contribution is an adaptation law for the terminal cost weight Qf and the input-rate penalty weight Rrate that trades off tracking accuracy and control effort based on the measured error and disturbance intensity. The method is validated via MATLAB simulations on a quadrotor subject to von K´arm´an turbulence with mean wind speeds and benchmarked against a conventional fixed-weight NMPC. Results show that the proposed controller reduces the RMSE of trajectory-tracking by an average of 8.82% (with per-wind-level reductions ranging from 2.09% to 9.69%) compared to the fixed NMPC, while also yielding consistently lower tail errors (95th percentile) and smoother, lower-energy actuator commands. These findings indicate that online weight adaptation offers a practical path to robust and efficient NMPC for UAVs operating in turbulent winds.

Keywords: Mechatronics Systems, Components and Devices, Factory Automation
Abstract: This study develops a multi-tool end effector driven by a single motor by extending the MaGDri mechanism proposed in our previous work, which was limited to gripper modules. The MaGDri mechanism enables torque-path switching between multiple tool modules using magnetic decoupling and gravity, eliminating the need for additional actuators and complex wiring. To extend the functionality beyond grasping, a screwdriver module with a ratchet mechanism is introduced, enabling one-way rotation and intentional generation of overtorque required for torque-path switching. In addition, a reduction mechanism is incorporated within the tool module to amplify the output torque without increasing the motor size, enabling flexible torque design at the module level. Experimental results demonstrate that the reduction mechanism increases the output torque in accordance with the gear ratio, while the ratchet mechanism enables intentional decoupling by constraining rotation. Furthermore, operational experiments confirm that a sequence of tasks—grasping, screw tightening, and loosening—can be executed continuously through selective switching of tool modules. These results demonstrate that the proposed system achieves reliable and versatile multi-task operation using a single motor.

Keywords: Mechatronics Systems, Robotic and Automation Systems
Abstract: Coping with the challenge of collecting high-quality expert data for industrial robotics, this paper introduces ECPGym. This ECPGym features a grid-based modular fixture mechanism. It allows easy reconfiguration of components, facilitating diverse inspection scenarios. An experimental setup was constructed using two manipulators and dual cameras to capture synchronized front view and wrist-view image streams along with joint state data, enabling monitoring of the operation and image information necessary for the maintenance and inspection of electrical equipment. By streamlining the collection of diverse datasets, ECPGym provides a practical and flexible environment for training autonomous inspection robots in complex industrial scenarios.

Keywords: Mechatronics Systems, Modeling, System Identification and Estimation
Abstract: Force control is essential for the tasks which require the interaction with the environment to achieve desired interaction force without damaging the environment. To design the force control system, the environmental impedance information is quite useful to guarantee the responsiveness. In this context, Kalman filter is one of the effective method to identify the environmental impedance. However, conventional Kalman filter-based environmental impedance estimation scheme did not include the environmental position. Environmental position is independently estimated by using the threshold of the contact and non-contact states. Hence, this study proposes a unified approach to simultaneous state and environmental impedance estimation using an extended Kalman filter including environmental position. The effectiveness of the proposed method is verified through numerical simulations.

Keywords: Mechatronics Systems, Mechanical Systems Control, Robotic and Automation Systems
Abstract: This study focuses on a leg equipped with a self-weight support mechanism that mechanically supports the self-weight to improve the energy efficiency of quadruped robots. This mechanism constrains the foot height relative to the body to maintain a constant body height. While this geometric constraint is maintained, elastic deformation of the leg components occurs when load is applied, resulting in sinkage that decreases the body height. To compensate for this sinkage, a previously proposed variable constraint angle method is also introduced. Furthermore, an alternative compensation method for sinkage, referred to as the lift-up method, is considered, in which the hip is lifted by a prescribed amount immediately after contact before applying the constraint. Through dynamic simulations, it is confirmed that the self-weight support mechanism contributes to reducing energy consumption, and the characteristics and trade-offs of the two sinkage compensation methods are clarified.

Keywords: Mechatronics Systems, Robotic and Automation Systems, System Integration
Abstract: This paper investigates vibration suppression in resonant systems based on time delay feedback. Conventional study achieved both fast response and vibration suppression in two-mass systems. However, steady-state error remained, limiting the accuracy of transient response evaluation. To address this issue, a vibration suppression method combining time delay feedback of reflected wave rejection with a load-side disturbance observer (LDOb) is proposed. In addition, the definition of the time delay parameter is revised for two-mass systems, and its influence on stability and vibration characteristics is analyzed through pole placement analysis and experiments. Comparative experimental results confirm that the proposed method significantly reduces the steady-state error and improves the settling performance under the nominal load condition while preserving vibration suppression. As a result, the proposed method provides an improved design framework for resonant systems accounting for time delay dynamics.

Keywords: Mechatronics Systems, Modeling, System Identification and Estimation, Transportation Systems
Abstract: Mobile robots for logistics and service applications must traverse not only flat floors but also steps and stairs. We are developing a transformable leg-wheel robot that uses wheeled locomotion on flat terrain and changes to a legged configuration on uneven terrain. This paper evaluates whether the robot possesses the kinematic workspace and statically stable postures required for stair-climbing control. A sequence of 12 quasi-static postures was constructed in a Autodesk fusion environment for stairs with a step height of 200 mm and a tread depth of 300 mm. Joint and linear-actuator limits, foot-contact conditions, swing-foot clearance, and the relationship between the center of mass and the projected support region were evaluated. Under both unloaded and 12 kg payload conditions, all postures were realizable within the mechanical range, and the center-of-mass projection remained inside the corresponding support region. These results demonstrate the mechanical, kinematic, and quasi-static feasibility of applying stair-climbing control to the developed robot.

Keywords: Mechatronics Systems, Robotic and Automation Systems, System Integration
Abstract: Tendon-driven robots actuate their joints by transmitting motor torque through tendons. These robots offer several advantages over conventional rigid-body robots, particularly regarding weight reduction, which is crucial for safe human-robot interaction. However, they face challenges in motion control due to specific tendon properties; notably, tendon elasticity causes vibrations that can be critical during high-speed motion. In this paper, we apply delayed-feedback control to stabilize the motion of tendon-driven mechanisms. First, closed-loop system identification is performed while regulating output motion. Consequently, the identification results characterize the tendon-driven mechanism as a two-mass model with damping terms. The stabilizing effect of the delayed-feedback control is validated through experiments using a test bench comprising a drive motor, a load, and transmission tendons. Vibration damping and steady-state error suppression are achieved by incorporating a high-pass filter (HPF) into the delayed-feedback loop, alongside a proportional-derivative (PD) controller and a disturbance observer (DOB).

Keywords: Robotic and Automation Systems, System Integration, Remote Sensing
Abstract: Chemical, Radiological, Biological, Nuclear, and Explosive (CRBNE) incidents demand immediate, high-fidelity situational awareness to preserve human life. While small multirotor Unmanned Aerial Vehicles (UAVs) offer superior close-range sensing, their utility is restricted by the "energy-endurance-payload" trilemma, often exhausting 70% of their battery life during the initial transit to a standoff distance. This paper introduces a novel two-tier hierarchical architecture designed to bridge the gap between range-limited consumer drones and logistically complex strategic military assets. The system utilizes a 25 kg class micro-turbojet powered fixed-wing carrier to transport and deploy multiple low-cost, attritable multirotor scouts. By decoupling strategic transit from tactical sensing, the carrier preserves the scouts' energy for mission-critical maneuvering and data collection within the hazard zone. Furthermore, the carrier orbits at a safe altitude to serve as a high-bandwidth Relay-UAV (RUAV), significantly increasing data throughput to ground stations in disrupted environments. Comparative analysis against global programs such as Jiutian and Project Vanquish highlights this tactical-scale specialization as a unique and underserved paradigm. System-level modeling under attritable mission assumptions indicates a major cost-effectiveness improvement over conventional single-platform architectures, enabling high-risk telemetry acquisition without risking high-value assets.

Keywords: Robotic and Automation Systems, Mechatronics Systems, Networked Sensor System
Abstract: Reliable communication is critical for effective coordination in emergency and search-and-rescue operations; however, such missions are often conducted in areas with limited or disrupted communication infrastructure. Current field practices, such as deploying mobile relay stations on elevated terrain, are often time-consuming and sometimes infeasible. This paper presents the design and experimental validation of a grid-powered tethered unmanned aerial vehicle (UAV) platform intended for rapid deployment as a temporary communication relay. The proposed system is designed to be low-cost and easily integrable with equipment commonly available to rescue teams. A key aspect of the system design is the use of 230 V AC power transmission through the tether, leveraging standard portable generators typically used in field operations. Comparative analysis of power transmission losses for DC and AC configurations was conducted to justify this approach and optimize cable requirements. The system’s calculated maximum deployment altitude exceeds 40 m under ideal conditions.A fully functional prototype was developed and experimentally evaluated. Test flights demonstrated stable operation and confirmed the system’s ability to carry payloads of up to 2 kg while maintaining continuous operation.The proposed solution enables rapid deployment of aerial communication infrastructure, significantly reducing setup time in critical scenarios.

Keywords: Computational Intelligence, Intelligent Systems, Advanced Computics
Abstract: To support molecular search for organic thin-film solar-cell materials, this paper compares a model based on Bidirectional Encoder Representations from Transformers (BERT) with a recurrent neural network (RNN) for masked-symbol prediction on strings written in Simplified Molecular Input Line Entry System (SMILES) notation. The corpus combined 21,786 QM8 strings and 652,617 ChemTS-generated strings; 674,186 strings remained after preprocessing. In 10-fold cross-validation, mean terminal-symbol prediction accuracy was 0.900 for the BERT-based model and 0.778 for the RNN. For random masks of one, two, and three consecutive characters, the respective mean accuracies were 0.909, 0.887, and 0.841 for the BERT-based model and 0.120, 0.005, and 0.000 for the RNN. The BERT-based model therefore achieved consistently higher accuracy in the masked-symbol prediction tasks examined. End-to-end evaluation is still required to determine whether this advantage improves molecular search.

Keywords: Human-Machine Systems, Robotic and Automation Systems, Human Interfaces
Abstract: The Fabric Robotic Suit (FRS) integrates pneumatic artificial muscles into clothing, and users wear this device for sports motion instruction. However, major concerns about guidance accuracy make the practical use of the FRS difficult in sports. This limitation is particularly pronounced during deep elbow flexion (at target angles less than 90◦) and elbow motion requiring large angular displacements (more than 90◦). In this study, we propose a new pneumatic control method, “Asymmetric periodic haptic feedback”. We designed this method based on human perceptual characteristics to help users perceive haptic cues clearly. Experiments with human participants show that the proposed method improves the accuracy of elbow-joint motion guidance under the conditions of deep elbow flexion and elbow motion requiring large angular displacements.

Keywords: Intelligent Control, Robotic and Automation Systems, Mechanical Systems Control
Abstract: Conventional robotic automation often relies on dedicated jigs to ensure positioning accuracy. Visual servoing can achieve high precision; however, it generally requires pre-prepared desired images, which reduces adaptability to changes in objects or environments. This study proposes a high-precision visual servoing method that remains robust when object conditions vary. The method introduces a ”Feature Imaginator”, a module that generates desired image features online based on the current visual input. Because the system produces the desired features dynamically, it removes the need for advance preparation of desired images. In addition, the method employs multi-scale feature maps to support both coarse alignment and fine positioning. The proposed method enables high-precision positioning with errors within 0.5 mm along each axis, even when the target object varies.

Keywords: Robust Control, Modeling, System Identification and Estimation, Intelligent Systems
Abstract: Although diffusion models provide powerful generative priors, controlling them during sampling to satisfy task-specific objectives and safety constraints remains challenging. A key failure mode of standard training-free guidance is distributional drift: naive task-driven updates can push reverse-process states into atypical regions where the pretrained model is unreliable, degrading sample plausibility and increasing hyperparameter sensitivity. We propose Typicality-Threshold Projection Guidance (TTPG), a training-free framework for plausibility-preserving control of pretrained diffusion models. TTPG introduces an explicit typicality-admissibility mechanism by calibrating a time-dependent threshold rev{on a score-norm-based typicality surrogate} from forward-noised training data and enforcing it through projection at each reverse step. This yields a practical correction rule that minimally removes inadmissible deviations from the unconstrained guided state while preserving task-oriented guidance. Experiments on a safety-constrained benchmark show that TTPG preserves plausibility, remains effective under aggressive guidance, and improves robustness to hyperparameter choice.

Keywords: Modeling, System Identification and Estimation, Nonlinear Control, Adaptive and Optimal Control
Abstract: We propose a probabilistic extension of exactly linearizable deep dynamics models for state-dependent uncertainty quantification. Existing neural network-based models that map nonlinear dynamics into a latent linear space allow optimal control problems to be solved as convex quadratic programs, but they are inherently deterministic and cannot quantify modeling uncertainty.

To address this limitation, we introduce state-dependent Gaussian modeling errors in the latent linear dynamics and learn their covariance jointly with the system model via a negative log-likelihood objective. A Jacobian-based linear approximation is employed to map the latent uncertainty to the physical state space.

The proposed framework represents state-dependent modeling uncertainty and provides approximate multi-step uncertainty bounds. Numerical experiments on a hybrid powertrain system show that the proposed model can estimate state-dependent uncertainty, suggesting its potential use in future uncertainty-aware control.

Keywords: Modeling, System Identification and Estimation, Adaptive and Optimal Control
Abstract: Neural networks have attracted attention as highly expressive prediction models for model predictive control (MPC) of nonlinear systems. However, their direct use in MPC leads to a nonlinear optimization problem, which incurs a high online computational cost and may suffer from local optima. In this paper, we propose a method that formulates the resulting MPC problem as a quadratic program (QP) by constructing an identified model comprising a nonlinear state estimator and an output predictor that is affine in the future input sequence. Because the proposed approach does not rely on prior physical knowledge, it is applicable to nonlinear systems that are difficult to model accurately from first principles. A quadrotor is a representative example of such a system, since complex aerodynamic effects hinder accurate first-principles modeling. Therefore, we consider a position-controlled quadrotor as the target system and experimentally demonstrate high-accuracy trajectory tracking on a real platform using the proposed QP-based MPC.

Keywords: Intelligent Control, Robust Control, Intelligent Systems
Abstract: Learning-based neural controllers show promising performance, but formal closed-loop stability guarantees remain difficult to establish and are critical for real-world deployment of physical AI. Existing neural control methods often rely on conservative constraints or post hoc stability certification, limiting flexibility and the ability to improve stability properties during training. In this study, we propose a learning-based control framework that integrates stability certification directly into the training loop by formulating stability conditions as a differentiable semidefinite program (SDP). This formulation enables stability certificates to be optimized with respect to neural controller parameters via gradient-based methods, treating stability constraints as learnable objects. We further examine the joint optimization of stability and task performance, highlighting its benefits and numerical challenges.

Keywords: Signal and/or Image Processing, Identification and Estimation
Abstract: This paper studies online tensor completion for third-order tensors whose column blocks arrive sequentially over time. While tensor nuclear norm minimization based on t-SVD provides accurate recovery, its computational cost increases as the observed tensor grows. Building on a distributed nuclear norm minimization scheme for matrices that synchronizes blockwise subspaces by inheriting the left singular vectors, we extend this framework to online tensor completion and introduce a temporally persistent synchronization that shares the dominant left singular subspaces of all blocks across time. We further accelerate the blockwise decomposition by replacing exact SVD with randomized SVD. Numerical experiments on synthetic tensors demonstrate that the proposed synchronized methods improve both recovery accuracy and computational efficiency over conventional online solvers.

Keywords: Guidance and Flight Control, Nonlinear Control, Signal and/or Image Processing
Abstract: This paper presents a sim-to-real framework for vision-based drone obstacle avoidance by integrating real-time YOLOv8 object detection with Model Predictive Path Integral (MPPI) control. Building on our earlier YOLOv8--MPPI obstacle avoidance study and its virtual experimentation extension, we explicitly address the major causes of sim-to-real degradation: perception uncertainty, control delay, and model mismatch. A repeatable drone simulator with a Python SDK is used to generate obstacle scenarios, inject observation noise, dropout, and communication delay, and sweep control constraints and cost weights. Based on risk metrics including minimum distance, avoidance margin, and reachability, the proposed framework designs conservative safety margins before transfer to a real DJI Tello platform. The controller core is preserved while only the input/output modules are replaced. The resulting study provides a practical design and validation process for uncertainty-aware visual obstacle avoidance and clarifies how virtual tuning can support safe real-flight deployment.

Keywords: Multivariable Control, Guidance and Flight Control, Transportation Systems
Abstract: This paper proposes a stability-preserving design method for synthesizing linear state-feedback gains using large sets of disturbance time-series data. Closed-loop stability is guaranteed through characteristic polynomial parametrization, while controller parameters are optimized to improve disturbance response performance. The proposed framework allows direct treatment of non-quadratic performance indices, such as peak-to-peak values of physical quantities, and can incorporate actuator constraints through a penalty-based formulation. The effectiveness of the approach is illustrated through numerical simulations of aircraft gust alleviation control.

Keywords: Guidance and Flight Control, Transportation Systems
Abstract: This study investigates the factors governing the whirling motion of a water-supply hose used in a helicopter‑mounted firefighting system (Fire Attack System). The whirling motion emerges as the flight speed increases and imposes operational limitations on firefighting missions with this system. Identifying the mechanisms responsible for this motion is an essential first step toward developing effective countermeasures. In this study, numerical simulations are conducted with reference to previous research on probe whirling observed in drogue-and-probe aerial refueling systems. The results indicate that unlike in aerial refueling systems, where the whirling motion is caused by wing-tip vortex or vortices shed from the pods, the whirling motion observed in the Fire Attack System is likely to be dominantly influenced by a low-velocity flow region formed downstream of the helicopter fuselage.

Keywords: Guidance and Flight Control
Abstract: In this report, an orbital transition with thrust vector control of plasma pulse thruster (hereafter called as PPT) for ultra-small satellite is described. Now the authors have been constructing ultra-small satellite of 2Usize(100mm×100mm×200mm) with propulsion system to adapt so strict condition as extremely small and low electric consumption. Then a PPT for the ultra-small satellite to transit the orbit is conducted, the authors propose the PPT with two functions such as thrust control and thrust vector control. As one function for the PPT, the thrust control system is achieved by varying operating frequency of the PPT, the other function as the thrust vector control system is behaved by the total thrust vector generating from sum of each different four thrusts control arranged appropriately. To clarify some performances of the proposed PPT system, some experiments and numerical calculations are executed and obtained data are evaluated.

Keywords: Guidance and Flight Control, Identification and Estimation, Standard of Measurement
Abstract: In this report, an attitude control system with IMU (Inertial Measurement Unit) is described. To detect the attitude angle, IMU as magnetic field and angular velocity for 3-axes with versatility and availability called as “COTS” is treated. Then, an attitude detection system without drift component by time passage is proposed by using Jacobian matrix with changing ratio for norm and argument of magnetic field vector. Moreover, the attitude detection system brought by IMU measured angular velocity and magnetic field with zero-shift is constructed. Then to confirm effectiveness of the proposing attitude detection system, some experiments with angular velocity and magnetic field and attitude calculation by the formulated Jacobian matrix are executed, the obtained results are evaluated. At the results, it is found that the constructed attitude detection system by using Jacobian matrix of time derivative as norm and argument of magnetic field vector is effective.

Keywords: Standard of Measurement, Identification and Estimation, Guidance and Flight Control
Abstract: In this report, an attitude detection system with polar coordinate transformation under magnetic field is described. To detect the attitude angle on inertial space for 3-axes, a transformation matrix of rate between orthogonal 3-axes and polar coordinate system is formulated as Jacobian matrix, rate data of magnetic field and angular velocity by IMU (Inertial Measurement Unit) with versatility and availability called as “COTS” is treated. Then, an attitude detection system on magnetic field without bias component by time passage is proposed by the polar coordinate transformation for orthogonal coordinate system of 3-axes. Moreover, to verify validity of the proposing attitude detection system, some experiments with perturbation of 3-axes under magnetic field are executed, the obtained results are estimated. At the results, it is found that the proposed attitude detection system by using polar coordinate transformation as Jacobian matrix is effective.

Keywords: Robotic and Automation Systems
Abstract: The rapid advances in AI technologies in recent years, including large language models, are beginning to transform our daily lives. In other words, the coexistence of humans and AI as artificial entities is becoming a natural part of everyday society. Supported by NEDO, we conducted the “Research and Development of Cyborg-AI” project from July 2020 to March 2025. In this project, we developed fundamental technologies that are also important for today’s Physical AI, including data-driven autonomous control methods, analysis of high-speed, high-degree-of-freedom human movements, and imitation learning by artificial agents (robots) based on human demonstrations. As an evaluation of these technologies, we demonstrated humanoid robot skateboarding performances at the robotic skateboard park installed within ATR. In this talk, after introducing the achievements of the Cyborg-AI project, I would like to discuss possible future forms of a society in which humans and Cyborg-AI coexist and coevolve, as well as the technologies that will be required to realize such a society, including perspectives and predictions for the future.

Keywords: Modeling, System Identification and Estimation
Abstract: Model Predictive Control (MPC) for quadruped robots relies heavily on accurate dynamics prediction models. Traditional explicit methods suffer from high derivation cost and linearization errors, while early data-driven sequential models fail to exploit the robot's inherent topological structure, leading to error accumulation in multi-step rollouts. This paper proposes a data-driven spatio-temporal dynamics prediction model with structural priors. A spatial topology graph is constructed based on joint states and contact observations, with learnable local relation edge weights introduced on a fixed skeleton topology. Graph Convolutional Networks (GCN) extract spatial features, while Temporal Convolutional Networks (TCN) model temporal dynamics. A multi-head output design separately predicts continuous body states and discrete contact events. Evaluations in MuJoCo show that our model outperforms MLP, LSTM, and Linear baselines in multi-step ADE/FDE, and achieves low prediction error on both body states and contact-related components.

Keywords: Modeling, System Identification and Estimation
Abstract: In practical industrial settings, operators often have access to future-known information, such as planned control inputs, which can be naturally incorporated into multi-step forecasting models. However, naive architectures may violate temporal causality, leading to information leakage. This issue is particularly problematic in decision-support applications that require reliable predictions. This paper proposes a causality-aware deep learning framework for incorporating future-known inputs while preserving temporal causality. We introduce a masked fully connected structure to eliminate non-causal dependencies and a structured parameterization inspired by temporal convolutional networks to impose a causal Toeplitz-like form. The proposed method is applied to a real-world gasification-based waste treatment plant for multi-step prediction of control-related variables. Experimental results show that, although a slight degradation in prediction accuracy is observed, the proposed approaches maintain a practically acceptable level of performance while reducing model complexity. These results indicate the potential of causality-aware design for practical industrial forecasting.

Keywords: Modeling, System Identification and Estimation
Abstract: In this paper, an offline disturbance estimation scheme using a Finite Impulse Response (FIR) filter has been proposed. System parameters and unknown disturbances are simultaneously identified through an optimization-based framework. The effectiveness of the proposed scheme has been validated via numerical simulations. These results demonstrate high-precision estimation under specific disturbance assumptions, providing a robust solution for environments with such disturbances.

Keywords: Modeling, System Identification and Estimation
Abstract: Multirate sensing makes system identification structurally difficult: conventional lifting-based approaches yield only products of the original state-space matrices and cannot recover A,B,C,D individually. The authors’ recent work introduced a cyclic reformulation approach that casts a multirate system as an enlarged linear time-invariant system and recovers A,B,C,D directly, applicable to both SISO and MIMO cases. Building on this framework, the present paper investigates the minimum data length required for exact identification, a question left open in the original work. We restrict ourselves to the SISO noise-free case for analytical tractability, and adopt a two-stage procedure combining MOESP-type subspace identification with convolution-based least squares. We derive the observability index of the cycled system, which can be significantly larger than the state dimension under sparse output observation, and numerical experiments show that the minimum data length is governed by the MOESP stage and exceeds its algebraic lower bound by exactly the cycled state dimension, independently of the observation density. For the tested settings, this yields an empirical closed form prediction of the minimum data length that is substantially tighter than the naive estimate obtained by treating the cycled system as an unstructured LTI system.

Keywords: Modeling, System Identification and Estimation
Abstract: Nonlinear dynamics of a pipe conveying fluid have attracted much attention for the several decades from the point of view of not only engineering applications but also mathematical interest of the mechanisms. In this study, we consider the dynamics of a cantilevered pipe conveying fluid with an end mass. In order to capture the qualitative characteristics of the nonlinear dynamics, we apply the method proposed by the authers. Different from the analysis by conventional methods, the method can carry out the nonlinear and bifurcation analysis on the destabilized mode. Using a projection operator constructed properly we can capture the qualitative characteristic in the bifurcation diagram, which cannot be shown in conventinal methods. At last, it is shown that the theoretically predicted bifurcation phenomena are in qualitatively good agreement with the experimental ones

Keywords: Modeling, System Identification and Estimation
Abstract: In hot rolling processes of steel, strip walking—i.e., lateral deviation of the strip in finishing mills—is primarily caused by a left-right thickness difference (wedge), which induces a lateral moment. The roll gap difference (leveling) and differential rolling force (differential force) are adjusted at each stand to control the wedge and suppress strip walking. This paper proposes a system identification method for modeling the relationship between these inputs and the exit wedge, based on the physical interpretation that the seven finishing stands are arranged sequentially. Two physics-informed approaches are investigated: (1) a method using stand-specific regularization weights newly introduced in this paper, and (2) a sequential parameter estimation procedure presented in a previous study. Experimental results using plant data demonstrate that the sequential estimation method improves generalization performance compared with the conventional uniform-regularization baseline.

Keywords: Modeling, System Identification and Estimation
Abstract: In this research, the authors extend the discrete-time Fourier transform for a linear time-invariant system to the 2D (two-dimensional) discrete-time Fourier transform for a linear time-varying system. They describe the input-output relation for a linear time-varying system by means of integral equations in the frequency domain. Moreover, they show a simple example by extending a time-invariant constant system to a time-varying system.

Keywords: Modeling, System Identification and Estimation
Abstract: This paper derives the covariance matrix of the system matrix in the MOESP type closed-loop subspace identification method (CL-MOESP). The covariance matrix can be derived by column expansion of the system matrix. The accuracy of the estimated values in system identification is important. One way to evaluate the accuracy is to calculate the error covariance matrix of the system matrix.

Keywords: Modeling, System Identification and Estimation
Abstract: This paper studies controllability Gramian assignment for discrete-time linear systems by state feedback. An optimization problem is formulated to make the closed-loop controllability Gramian close to a desired Gramian. In addition, conditions for exact assignment and a corresponding state feedback gain are derived using a relationship between controllability Gramian assignment and state covariance assignment. The proposed approach is examined through a spring-mass-damper system.

Keywords: Modeling, System Identification and Estimation, Mechanical Systems Control
Abstract: This study proposes a vibration‑assisted legged rover designed to suppress slippage on soft planetary terrain by actively compacting the soil beneath its feet. To regulate the compaction process, the rover estimates ground stiffness from vibration acceleration measured at the leg and adaptively controls the duration of applied vibration. Experimental results show that the amplitude of the fundamental vibration frequency increases monotonically with vibration time, reflecting progressive soil densification. To quantify this behavior, an amplitude ratio, defined as the increase in fundamental frequency amplitude relative to the initial state, is introduced and found to correlate strongly with the bearing capacity of the ground. By applying a threshold to this ratio, the rover autonomously determines when sufficient compaction has been achieved and terminates vibration. The proposed mechanism enables more stable locomotion on deformable terrain and provides an effective strategy for slip suppression in planetary exploration.

Keywords: Modeling, System Identification and Estimation, Mechanical Systems Control, Robust Control
Abstract: This study proposes an online parameter identification framework based on the extended state observer (ESO) for mechatronic systems with uncertainties. The lumped disturbance estimated by the ESO is interpreted as an indicator of model mismatch, based on which parameter adaptation is designed to drive it to zero. The resulting adaptive ESO (AESO) with a state sensitivity equation (SSE) enables simultaneous state estimation, disturbance rejection, and parameter identification. Experimental results demonstrate that the proposed method achieves more consistent parameter identification than conventional recursive least squares (RLS) methods.

Keywords: Modeling, System Identification and Estimation, Guidance and Flight Control, Autonomous Decentralized Systems
Abstract: Accurate modeling of aerodynamic disturbances is essential for enabling multirotor flight in close proximity to environmental structures such as the ground, walls, and tunnel. However, these disturbances are highly nonlinear and difficult to model using conventional analytical approaches. In this work, we propose a hybrid modeling framework that combines a nominal multirotor dynamics model with a data-driven disturbance estimator. The disturbance forces induced by nearby surfaces are learned using a permutation-invariant deep neural network based on the deep sets architecture. Spectral normalization is employed to ensure Lipschitz continuity and improve generalization. Experimental data is collected using Crazyflie 2.1 platform in environments with ground and wall interactions. Visualization of the learned force fields demonstrates that the model captures key aerodynamic effects, including ground effect, wall-induced disturbances, and their nonlinear coupling. Furthermore, the model successfully captures velocity-dependent interactions.

Keywords: Modeling, System Identification and Estimation, Human-Machine Systems, Mechanical Systems Control
Abstract: This study investigates how differences in acceleration response between electric vehicles (EVs) and internal combustion engine vehicles (ICE vehicles) affect driver speed-control behavior. To this end, a driver–vehicle model based on internal model control (IMC) was developed. In this model, the vehicle characteristics assumed by the driver were represented as the nominal plant, and the actual vehicle characteristics were represented as the actual plant. The difference between them was treated as model error. In the numerical simulations, the nominal model was fixed as an ICE vehicle. Then, the actual plant was set as either an ICE vehicle or an EV, and the matched and mismatched conditions were compared. In addition, the effects of perceptual delay and command update delay were examined under the mismatched condition. The results showed that oscillations appeared in both the speed response and the control input under the mismatched condition. The oscillations were amplified as the delays became larger. These results indicate that, for the representative parameter settings considered in this study, vehicle-characteristic mismatch and driver-side delays can induce oscillations during speed tracking.

Keywords: Modeling, System Identification and Estimation, Intelligent Control, Identification and Estimation
Abstract: State estimation in high-dimensional dynamic systems is challenging due to the curse of dimensionality. This research proposes a numerically stable filtering framework grounded in Dynamical Low-Rank approximations to achieve real-time tracking in such regimes. We first address the stability for numerical computation of the proposed nonlinear filter by employing the basis update and Galerkin integrator. Additionally, to decouple computational complexity from ambient dimensionality, we introduce Neural Discrete Empirical Interpolation Methods (Neural-DEIM) for efficient DEIM reconstruction. By leveraging a Set Transformer architecture, Neural-DEIM achieves permutation-invariant spatial node selection, thereby overcoming the severe basis degeneration encountered by traditional singular-value-decomposition-based greedy algorithms in convection-dominated regimes.

Keywords: Modeling, System Identification and Estimation, Process Control, Computational Intelligence
Abstract: 。地熱発電は効果的な再生可能エネルギーです。 カーボンニュートラルに寄与するエネルギー源;しかし、 その収容能力率は年々減少しています。かなり この減少の多くは地下の条件に起因しています。 生産井戸での蒸気噴出停止により その一例です。これらの異常に対処するために、非常に高い 正確な異常前駆検出モデルが求められます。オン 一方で、地熱発電所は異なります 地下環境および貯留層においても重要

Keywords: Modeling, System Identification and Estimation, Signal and/or Image Processing, Simulation of Large Systems
Abstract: This paper studies a low-rank Kalman-Bucy filtering framework for linear time-varying systems through the tracking analysis of Oja's principal component flow. Under structured assumptions on the eigenspaces and their time variation, we show that the Oja flow can remain in a neighborhood of the time-varying dominant subspace by tuning a parameter of the flow, rather than tracking it exactly. These restrictive assumptions identify a tractable class of linear time-varying systems and provide a theoretical basis for low-rank filtering, as illustrated by a numerical experiment.

Keywords: Nonlinear Control, Mechanical Systems Control, Guidance and Flight Control
Abstract: This paper focuses on a trajectory-tracking method designed to enable parking, verified through experiments using an electric wheelchair. The results show that while the simulation exhibited oscillatory behavior, the actual wheelchair achieved significantly smoother and more stable motion. Furthermore, the vehicle continued to converge toward the reference after the target movement ceased. These findings confirm the practical potential of the proposed method for autonomous parking tasks.

Keywords: Nonlinear Control, Mechanical Systems Control
Abstract: In this paper, we evaluate the tracking performance of a system using nonlinear homogeneous finite-time P-PI control with a distorted-wave reference. Firstly, we guarantee the frequency response of the closed-loop system with homogeneous finite-time P-PI control and clarify its mathematical properties. Secondly, based on the properties of frequency response, we assess the tracking performance for the distorted-wave reference.

Keywords: Nonlinear Control, Mechanical Systems Control
Abstract: The design of attitude observers and output feedback circular path following controllers is considered for sampled-data four-wheeled mobile robots (4WMRs) with steering angle saturation. First, attitude observers are designed by applying the design framework of nonlinear sampled-data control systems. Then the designed observers are combined with the state feedback circular path following controllers to construct output feedback circular path following controllers. It is shown that the designed controllers achieve sampled-data circular path following control in the semiglobally practically input-to-state stable (SP-ISS) sense. Finally, simulation and experimental results are given to show the effectiveness of the designed observers and controllers.

Keywords: Nonlinear Control, Mechanical Systems Control
Abstract: Level sets of a Lyapunov function produced by feedback control　design are often squashed heavily for linear systems, and bent and twisted further for nonlinear systems. Such Lyapunov functions do not cause disadvantage in linear control and global nonlinear design. However, those Lyapunov functions confirm the virtual uselessness of the designed controllers when achievable stabilization is not global. Recall that level sets are essentially the only means of estimating stability regions. Since controllers can alter systems' vector field, specifying Lyapunov functions in desired shape a priori is more suitable in practice than accepting posteriori Lyapunov functions produced along with controllers. Given a local Lyapunov function of desired shape, this paper proposes the framework of designing a controller that provides the largest achievable domain of attraction estimate in the presence of the largest admissible uncertainties. The idea and its usefulness are illustrated through feedback control of an inverted pendulum on a cart.

Keywords: Nonlinear Control, Multivariable Control, Guidance and Flight Control
Abstract: Berthing maneuvers at low speeds are challenging, yet most existing automatic berthing methods assume continuous-valued propeller commands that cannot be directly applied to vessels with fixed pitch propellers, which are subject to strict actuator operational constraints. This study proposes a berthing control method based on discrete engine order telegraph commands. To enable online coordinated control, the method separates the search for a propeller command sequence that governs speed, from the optimization of rudder commands that governs route tracking. The propeller search results are provided as states to a model predictive controller for the rudder, accounting for changes in rudder effectiveness caused by speed variations and propeller operations. The effectiveness of the proposed method was verified through real-time simulations and full-scale sea experiments.

Keywords: Nonlinear Control, Robotic and Automation Systems
Abstract: In recent years, two-wheeled mobile robots, such as electric wheelchairs, have increasingly adopted autonomous driving technologies. Trajectory planning for the systems has become an important research topic. This paper proposes a trajectory optimization method for two-wheeled mobile vehicles based on the conjugate gradient method. The proposed method improves convergence speed compared with conventional gradient descent by incorporating previous search directions.

Keywords: Nonlinear Control, Computer Aided Design, Modeling, System Identification and Estimation
Abstract: We propose an improved numerical method for the inverse problem of nonlinear optimal control based on the stable manifold method. The stable manifold method provides an exact numerical solution to the Hamilton-Jacobi (HJ) equation by utilizing the Hamiltonian system associated with the HJ equation. In the inverse problem, the weight matrices in the cost function are computed at each point along a given reference trajectory, and therefore become time-varying. It is known that the HJ equation remains valid under perturbations of the Hamiltonian; however, the Hamiltonian tends to grow as the solution moves away from the origin in reverse time, resulting in deterioration of numerical accuracy. The proposed method incorporates a correction based on the St{"{o}}rmer–Verlet method, which is a symplectic numerical integration scheme. Numerical experiments confirm that the proposed method reduces the deviation of the Hamiltonian from zero.

Keywords: Nonlinear Control, Robust Control
Abstract: This is a continuation of our recent attempt at disproving absolute stability of discrete-time nonlinear feedback systems using integral quadratic constraints (IQCs) and linear matrix inequalities (LMIs). We have shown very recently that, if the dual solution of an IQC-based LMI ensuring absolute stability is of rank-one, then we can detect a destabilizing nonlinearity that generates a non-zero fixed point from the assumed class of nonlinearities. In the case of repeated and slope-restricted nonlinearities, we further extended this result by lifting and enabled the detection of destabilizing nonlinearities that generate periodic trajectories. In this paper, we extend these results to repeated, slope-restricted, and “odd” nonlinearities case.

Keywords: Nonlinear Control, Human-Machine Systems, Safety, Environment and Eco-Systems
Abstract: Automobiles are essential in our lives, while accidents due to human error are increasing. Tezuka et al. proposed a human assist control law using a zeroing control barrier function (ZCBF). Furthermore, Arai et al. proposed a human-assist control law using a hybrid control barrier function (H-CBF). In this paper, we implement the H-CBF based human assist control on an electric wheelchair and confirm its effectiveness through actual experiments.

Keywords: Nonlinear Control, Modeling, System Identification and Estimation, Guidance and Flight Control
Abstract: This paper presents a generic approach to controlling Takagi-Sugeno fuzzy systems with unknown disturbances affecting their inputs. First, the LMI design conditions for simultaneously obtaining the feedback and observer gains are derived by transforming the fuzzy system. Next, new conditions for imposing constraints on the estimation error and system input are derived. The proposed approach enables effective tuning of the gain magnitudes with low computational complexity. Finally, the proposed approach is applied to an unmanned aerial vehicle as a typical example of a complicated nonlinear system. Simulation and experimental results demonstrate the effectiveness of the proposed approach.

Keywords: Nonlinear Control, Human Interfaces, Virtual Reality Systems
Abstract: In this study, we addressed the issue of motor oscillation in a human-in-the-loop remote control system. We developed a VR-based head-tracking system that incorporates human dynamic behavior into the feedback loop. To achieve stable and responsive operation, we analyzed the oscillation conditions using delayed differential equations. We conducted experiments to determine whether human reaction time should be considered in the design of control gains. The results revealed that human reaction time does not directly affect the physical oscillation conditions of the motor. By designing gains based solely on hardware delay, we suppressed oscillation while maximizing responsiveness, thereby significantly improving operational comfort.

Keywords: Nonlinear Control
Abstract: Echo State Networks (ESNs) in the reservoir computing framework are capable of approximating nonlinear dynamical systems in a sample efficient way. ESNs have been mainly used as part of control systems for nonlinear system identification and control, but their standalone use for complex continuous control tasks remains limited. In this study, we apply ESN-based learning approach to quadcopter (drone) control tasks, where the system is high-dimensional and highly nonlinear. We demonstrate that, by increasing the flexibility of parameter settings and performing appropriate tuning, ESN-based model-free online control alone can achieve stable quadcopter flight control. In addition, under a unified evaluation protocol, we conduct a comparative analysis of the ESN-based approach with PID control and reinforcement learning-based control methods. We discuss the characteristics and limitations of the ESN-based method.

Keywords: Guidance and Flight Control, Identification and Estimation, Robotic and Automation Systems
Abstract: In this paper, we propose a low-cost ROS-based multi-sensor information fusion framework for 3D localization of quadrotors in GPS-denied indoor and outdoor environments. The proposed system integrates visual odometry and 3D map-based localization using the RTAB-Map ROS2 package together with MAVROS for ROS2 Humble. Visual odometry and 3D map localization data are transmitted directly to the flight controller, where the onboard Extended Kalman Filter (EKF2) performs multi-sensor fusion to estimate the final drone pose. The primary objective of this work is to achieve reliable localization in GPS-denied environments such as indoor spaces and dense forest areas. Experimental results demonstrate a localization accuracy of 0.4763 m along the indoor linear axis, while the same sensor fusion framework achieved an outdoor linear pose accuracy of 0.264 m under brighter lighting conditions in GPS-denied environments.

Keywords: Virtual Reality Systems, Human Interfaces, Entertainment Systems
Abstract: This study focuses on the productization of “haptic media,” which conveys the presence of environments and living creatures through touch. While previous work mainly explored research prototypes, this year we developed a product-oriented concept called “DIY Haptic Media,” aiming to create haptic content that can be easily produced, experienced, and distributed. We also explored the possibility of a platform where anyone can create and experience haptic content. Conventional haptic expressions mainly relied on changing physical materials such as beads or small objects. In this work, we expand the range of haptic representations and propose diverse haptic experiences, including feedback that conveys the sensation of flowing water, vibrations synchronized with bird songs and wing flapping, and haptic expressions representing the movement of insects or small animals. These approaches allow users to experience natural phenomena and living creatures through touch, as well as through vision and hearing. Furthermore, we compared haptic-only presentation with haptic presentation combined with additional information, such as sound, and investigated their effects on immersion and presence. We also discuss sensing and haptic presentation methods for education, entertainment, and interactive product applications.

Keywords: Mechanical Systems Control, Human Interfaces
Abstract: Swivel-caster carts are widely used in factories because of their high maneuverability in narrow spaces. However, when transporting large or heavy loads, they often have poor straight-line stability, especially when the caster orientations are not aligned with the desired traveling direction. In addition, when the center of gravity of the load is offset from the center of the cart, the resulting yaw moment can cause the cart to deviate from the intended straight path. This increases the operator’s burden and makes stable transportation difficult. This study proposes an assist system for large-load transportation using a swivel-caster cart. Based on a dynamic model of a four-wheel swivel-caster cart, an assist method using an additional active caster wheel is designed to improve start-up behavior and straight-line stability. Furthermore, to experimentally verify the effectiveness shown by numerical simulations, the development of an assist cart is considered, with particular focus on the control of the active caster wheel used as an actuator.

Keywords: Virtual Reality Systems, Human Interfaces
Abstract: Telexistence vehicles require intuitive, stable user interfaces to achieve effective remote mobility and immersive teleoperation. This study evaluates the operability of a telexistence vehicle previously developed by Kurimoto et al. using three input interfaces: a joystick controller, a keypad, and a steering wheel-type controller. The objective is to clarify how differences in operation interfaces affect maneuverability, operational accuracy, usability, and user experience in remote vehicle control. In the evaluation experiments, participants operate the telexistence vehicle using each controller under identical navigation tasks. The experiments compare travel performance, task completion time, trajectory accuracy, and subjective usability. The joystick controller is expected to provide flexible multidirectional operation, while the keypad interface offers simple and low-cost input control. In contrast, the steering wheel-type controller is expected to improve intuitive operability by replicating driving sensations similar to those of real vehicles. In addition, the study investigates learning efficiency, operator workload, and immersion associated with each interface. The results are expected to clarify suitable operating methods for different application scenarios and operator characteristics. This research contributes to the design of user interfaces for telexistence systems, remote mobility platforms, and teleoperated robotic vehicles.

Keywords: Modeling, System Identification and Estimation, Human-Machine Systems
Abstract: Positive emotions, including feelings of happiness, are significantly involved in sustaining continuous behaviors such as exercise and studying. However, since such emotional states are difficult to observe directly, a dynamic model that represents the temporal changes in emotions is necessary to quantitatively analyze how these emotions evolve during the course of sustained behavior. This poster aims to verify the representational capabilities of a dynamic model of happiness for future use in control-based behavior support systems. First, building on an existing hedonic treadmill model based on the approach–avoidance framework, the model is reformulated into a state-space representation to enable systematic analysis and integration with model-based control methodologies. Secondly, to verify the representational capacity of the proposed model, parameter identification is conducted using time-series data obtained from strength training experiments. Under the hypothesis that variations in physical performance reflect underlying motivational and affective dynamics, the model parameters are estimated and the resulting behavior is compared with observed data. The results provide that the state-space representation of the happiness model is capable of expressing key features of sustained behavioral dynamics. This work constitutes a first step toward the design of control systems for maintaining human motivation and sustained engagement.

Keywords: Virtual Reality Systems, Human Interfaces, Medical and Welfare Systems
Abstract: Telexistence vehicles require intuitive, stable user interfaces to achieve effective remote mobility and immersive teleoperation. This study evaluates the operability of a telexistence vehicle previously developed by Kurimoto et al. using three input interfaces: a joystick controller, a keypad, and a steering wheel-type controller. The objective is to clarify how differences in operation interfaces affect maneuverability, operational accuracy, usability, and user experience in remote vehicle control. In the evaluation experiments, participants operate the telexistence vehicle using each controller under identical navigation tasks. The experiments compare travel performance, task completion time, trajectory accuracy, and subjective usability. The joystick controller is expected to provide flexible multidirectional operation, while the keypad interface offers simple and low-cost input control. In contrast, the steering wheel-type controller is expected to improve intuitive operability by replicating driving sensations similar to those of real vehicles.

Keywords: Intelligent Systems, Signal and/or Image Processing, Human Interfaces
Abstract: The efficiency of guide dog training is often related to the animal’s psychological state. However, identifying these states remains challenging, as such assessments often rely on implicit, experience-based knowledge acquired through training practice. This study proposes a deep learning-based attention estimation framework utilizing multi-camera 3D skeletal reconstruction of both trainers and dogs. By quantifying spatiotemporal features and relative geometric relationships from the 3D skeletal data, the framework provides an objective metric for guide dog attentiveness. This approach allows us to validate the correlation between training outcomes and attention levels, while systematically analyzing the underlying factors that influence a guide dog's focus. The proposed framework is expected to support the objective evaluation of guide dog attentiveness and contribute to a better understanding of training dynamics.

Keywords: Sensors and Transducers, Ultra-High Precision Measurement, Human-Machine Systems
Abstract: Airborne Ultrasound Phased Array (AUTD) devices require precise synchronization to generate stable acoustic interference for mid-air haptic applications. Although the current AUTD3 system can achieve synchronization using EtherCAT, it is mainly designed for local, dedicated, short-distance, and master-slave control environments. To extend synchronization capability to more general network environments, this study proposes a UDP/IP-based distributed clock synchronization system for multiple AUTD/FPGA devices. The proposed method compensates for propagation delay and system time offset to reduce data-loop latency and timing skew, while a D-PLL-based algorithm is employed to mitigate clock drift. Through experiments and simulations, the synchronization system has been implemented on three FPGA boards to verify the timing correction performance. The experimental results show that the time error between two boards increases almost linearly before drift compensation, with an estimated drift rate of 62,574 ns/s, equivalent to approximately 62.6 ppm. This confirms that oscillator drift is a major source of long-term synchronization error. Future work will focus on improving the D-PLL algorithm to accelerate convergence and support scalable AUTD systems for haptic applications over general UDP/IP networks.

Keywords: Human-Machine Systems, Social Systems, Virtual Reality Systems
Abstract: Urban Air Mobility (UAM), commonly known as “flying cars,” has attracted worldwide attention as a next-generation transportation infrastructure. However, compared with technological development, research on social implementation, including legal frameworks and operational systems, remains insufficient. One major challenge is the lack of objective and quantitative methods for evaluating public acceptance of new aerial infrastructures. Since acceptance toward UAM may vary depending on regions and stakeholders, establishing evaluation criteria is essential for future deployment in Japan. This study aims to develop a social acceptance evaluation framework for UAM based on objective physiological measurements. In particular, a simple EEG-based “Kansei Analyzer” is employed to evaluate emotional responses toward UAM. The analyzer estimates real-time emotional indices using an algorithm trained on large-scale datasets linking EEG patterns and hormone-related evidence. By collecting large-scale experimental data, the proposed framework seeks to identify acceptance thresholds for different stakeholders, such as acceptable noise levels and flight altitudes. The study further aims to provide bottom-up indicators for institutional design and operational standards of UAM systems. Ultimately, this research contributes to consensus building among industry, academia, government, and citizens toward the realization of sustainable aerial mobility.

Keywords: Mechatronics Systems, Human Interfaces, Virtual Reality Systems
Abstract: The authors have been conducting research on the application of pneumatic actuators, a type of soft actuator, to force display systems. Initially, an actuator using a bicycle inner tube, named the Balloon Actuator, was developed. This actuator was capable of generating relatively large forces of approximately 100 N; however, its response time was about 300 ms, making it unsuitable for force display applications. Subsequently, a new Balloon Actuator was developed using a 3D printer capable of printing soft polymer materials with a hardness of approximately Shore A5. Although the maximum force decreased to approximately 20 N, the response time improved to about 200 ms. However, the actuator could withstand only low internal pressures and was prone to rupture.

To further improve responsiveness and robustness, a new Balloon Actuator was fabricated using a self-healing polymer gel material and a 3D printer. This latest Balloon Actuator achieved a larger output force than the previous version, although its response time also increased. These results and their implications will be discussed in detail in the poster presentation.

Keywords: Human Interfaces, System Integration, Robotic and Automation Systems
Abstract: This study develops a remote-operated path-following robot system capable of navigating around obstacles and pedestrians. The robot autonomously tracks a predefined path, and an operator observes its movement from a remote location. If the robot detects obstacles or people within a sensor range of 1.0 m horizontally and 1.5 m in depth from its center, it halts and waits for the operator’s control. To facilitate remote situational awareness, the robot is equipped with a 180-degree field-of-view camera. Upon detecting a pedestrian on the path, the system plays an audio message requesting them to clear the way. Once the remote operator confirms safety and ensures the path is clear, autonomous path-following resumes. If a static obstacle persists, the operator manually steers the robot to bypass it before restoring autonomous navigation to the destination. Experimental results confirmed that the robot successfully stopped for obstacles, adapted via remote intervention, and resumed path-following to reach its destination safely, demonstrating the feasibility of the proposed remote navigation guidance system.

Keywords: Signal and/or Image Processing, Human-Machine Systems, Entertainment Systems
Abstract: Recent advances in haptic feedback devices with high spatial resolution, such as airborne ultrasound phased arrays, have enabled the presentation of detailed pressure stimuli. This study investigates a communication framework for remote haptic interaction using pressure stimulus data with high spatial resolution. The proposed framework represents haptic stimuli by associating a human hand mesh with pressure values that vary over time at each vertex. Control signals tailored to each device are then generated from this stimulus representation, allowing the data to serve as a common haptic content format for various haptic devices, including airborne ultrasound tactile displays and VR controllers. To support immersive interaction with low latency, we implemented compression based on wavelet transforms. For data transmission, we adopted packet communication through WebRTC DataChannel to achieve low latency networking and demonstrated data exchange between multiple PCs. We further evaluated the overall processing latency, including encoding, transmission, and decoding. Finally, we discuss the trade off between spatial resolution and communication latency when transmitting large volumes of detailed haptic data.

Keywords: Virtual Reality Systems, Human-Machine Systems, Human Interfaces
Abstract: In recent years, video content and games that provide strong audiovisual rewards have increasingly been used in childcare environments. However, concerns have been raised that excessive or compulsive use of such media may negatively affect children’s mental health. In this study, we propose an interactive system that enables children to spontaneously enjoy pure tactile stimulation itself. The proposed AR system consists of an airborne ultrasound phased array for presenting contactless and unconstrained tactile feedback, and an ELF-SR2 autostereoscopic display. In the proposed method, virtual soap bubbles emerge from the bottom of the autostereoscopic display. When a user interacts with these bubbles, tactile feedback is presented together with visual effects indicating that the bubbles have burst. To investigate how tactile feedback attracts children’s interest, we conducted a user experiment comparing children’s behavior with and without haptic feedback, using behavioral indicators such as the time spent interacting with the system. This study may contribute to the development of tactile applications designed for children as alternatives to conventional audiovisual entertainment.

Keywords: Human Interfaces, System Integration, Virtual Reality Systems
Abstract: Midair haptics using airborne ultrasonic phased arrays is a promising method for presenting tactile feedback without physical contact. In virtual button interaction, a midair click sensation can provide users with tactile confirmation that a pressing action has been performed or completed. Previous studies on midair virtual buttons have often employed stage-based feedback strategies, such as changing stimulation patterns or modulation frequencies according to discrete interaction phases. In this study, we focus on press-dependent feedback, in which the tactile stimulus is continuously adjusted according to the user’s fingertip motion. We propose a real-time midair click feedback method based on fingertip position tracking and spatial pressure distribution control. A  depth camera acquires the three-dimensional fingertip position, and the height information of the fingertip is mainly used to control the acoustic pressure distribution generated by an airborne ultrasonic phased array. The aim is to present tactile cues that correspond to the downward pressing motion and indicate the progression or completion of a virtual button press. This work presents the system configuration, fingertip tracking, RealSense measurement characteristics, and hardware arrangement of the ultrasonic phased arrays.

Keywords: Human Interfaces, Virtual Reality Systems, Human-Machine Systems
Abstract: We propose a system that combines a CG image of a virtual arm presented on a two-dimensional remote screen with contactless tactile stimulation delivered to the participant’s real hand using an ultrasound phased array. The purpose of this study is to examine whether such a simple visual–tactile setup can function as a platform for inducing and evaluating body ownership, even when the position of the real hand and that of the displayed virtual arm do not physically coincide. The rubber hand illusion is used as an evaluation framework to assess the effectiveness of the proposed system. In addition, to clarify the conditions and tolerance range under which the system functions, we investigate the effect of changing the presentation angle of the virtual arm.

Keywords: Human-Machine Systems, Human Interfaces, Virtual Reality Systems
Abstract: Human error continues to be a significant factor in railway accidents and operational disruptions, despite the increasing automation and digitalization of railway systems. This study presents a multi-user virtual collaborative training simulator developed to support railway workers in responding to abnormal and emergency situations. The system combines a virtual driver’s cab, head-mounted display-based XR training, a training control and monitoring module, M&S-based situation simulation, and high-fidelity 3D railway environments. The simulator allows ten or more railway workers, including drivers, traffic controllers, station staff, train crew, and track workers, to participate in the same virtual scenario. Each participant can practice coordinated response procedures according to their assigned role. The training scenarios are built on more than 300 event elements derived from domestic and international railway accident cases. They are further extended using an AI-assisted scenario generation model that incorporates random events, human-error checkpoints, and accident precursor conditions. Repeated scenario-based training is expected to improve situation awareness, communication, procedural compliance, and collaborative decision-making among railway workers. These improvements of this study can contribute to reducing various railway incidents related to human error.

Keywords: Intelligent Systems, Human Interfaces, Virtual Reality Systems
Abstract: Recent advances in artificial intelligence have enabled the automation of agricultural crop grading. However, practical deployment of such technologies requires effective user interfaces that allow workers to access grading results without disrupting their workflow. In conventional crop grading systems, workers must repeatedly shift their gaze between crops and display devices, which may reduce work efficiency and increase workload. To address this issue, this study proposes a smart glass-based augmented reality (AR) system that presents grading results directly within the user’s field of view. The proposed system employs a camera mounted on smart glasses and a convolutional neural network (CNN) to classify cucumber grades and display the classification results through AR. To evaluate the effectiveness of the proposed system, a comparative study was conducted against a conventional PC-based grading system. The results showed that the proposed system achieved higher evaluation scores across all assessment items. In particular, notable improvements were observed in perceived work efficiency and gaze movement. These findings demonstrate that the proposed system can enhance the usability of crop grading tasks and highlight the potential of smart glass-based AR interfaces for practical AI-assisted agricultural operations.

Keywords: Virtual Reality Systems, Human Interfaces
Abstract: Previous studies have suggested that spatial expansion induced by amodal completion may not exacerbate VR sickness. To investigate why this phenomenon occurs, we are currently conducting an experiment. In contrast to previous studies, participants are exposed to the stimuli for a longer duration while VR sickness severity and vection intensity are measured. This design enables us to examine temporal changes in subjective ratings and compare these changes between conditions with and without amodal completion. The maximum stimulus exposure duration is 8 min. Subjective ratings are collected every 2 min using a Numerical Rating Scale (NRS) during stimulus presentation while participants perform a target-search and counting task. The NRS is adopted because it provides an intuitive measure of VR sickness symptom severity while minimizing the burden associated with repeated subjective reporting. Nevertheless, obtaining subjective ratings may temporarily divert participants' attention from the VR stimulus. Therefore, future work will explore the feasibility of using objective physiological measures to evaluate VR sickness.

Keywords: Human Interfaces, Virtual Reality Systems, Entertainment Systems
Abstract: Airborne ultrasound tactile display (AUTD) can create touch sensations in mid-air by focusing ultrasound on the skin without direct contact. Previous studies have suggested that mid-air ultrasound tactile stimuli can evoke pleasant sensations. However, it is still unclear whether synchronized tactile feedback can improve pleasantness and what kinds of sounds are suitable for audio-tactile presentation. In this study, we developed an audio-synchronized mid-air ultrasound tactile system using an AUTD device. The system detects sound events and converts them into synchronized mid-air ultrasound tactile patterns. We tested several sound types, including environmental sounds such as wind, rain, and thunder, and touch-related sounds such as stroking and brushing sounds. User evaluations were conducted to investigate the effects of synchronized tactile feedback on the pleasantness of auditory experiences. Preliminary observations suggest that synchronized audio-tactile presentation may improve perceived pleasantness These findings demonstrate the potential of synchronized mid-air tactile feedback to enrich auditory experiences and contribute to the development of design guidelines for pleasant and effective multimodal interaction systems.

Keywords: Virtual Reality Systems, Human Interfaces
Abstract: Virtual reality (VR) sickness is influenced by visual motion speed. While increasing motion speed enhances vection, it can also increase discomfort, creating a trade-off between immersion and user comfort. Recent studies using simple optical-flow stimuli have suggested that superimposing optical flows with different velocities can reduce discomfort while preserving vection strength. However, it remains unclear whether this effect extends to more complex VR environments. This study investigates whether superimposing optical flows with different velocities can reduce discomfort while preserving vection strength in an outer-space VR environment. One optical flow velocity was fixed, while the other was reduced by 0, 50, 100, or 200 m/s, yielding four experimental conditions. Vection strength, perceived speed, and discomfort were compared across these conditions. Vection strength and perceived speed were evaluated using magnitude estimation to capture subtle perceptual differences associated with different optical-flow velocity combinations. Discomfort was evaluated using a numerical rating scale, which provides an intuitive assessment of symptom severity and facilitates consistent subjective reporting, even among participants with limited experience in psychophysical evaluations. To complement subjective reports and capture temporal changes in VR sickness, future studies will explore the use of physiological measures.

Keywords: Virtual Reality Systems, Human Interfaces
Abstract: Providing a fixation point is considered an effective strategy for preventing visually induced motion sickness (VIMS). This effect is thought to result from the fixation point capturing the gaze, thereby promoting visual fixation, as well as suppressing saccades to the surrounding areas. If this theory is correct, symptoms should be suppressed as long as there is a fixation target that captures the gaze, even if it is merely an object embedded within the video. To verify this hypothesis, we will compare the severity of VIMS symptoms among fixation points used in previous studies, small animal-shaped objects, and objects with mosaic textures. Furthermore, to verify the extent to which each stimulus was actually fixated upon, we will use a spectacle-type eye-tracking device to obtain gaze characteristics during video observation. Specifically, we will compare the fixation rate on the fixation target, the frequency of saccades directed outside the fixation target, and optokinetic nystagmus across the different fixation targets. Additionally, by examining VIMS symptoms in conjunction with gaze characteristics, we will verify whether the results support our hypothesis.

Keywords: Robust Control, Autonomous Decentralized Systems, Modeling, System Identification and Estimation
Abstract: This paper presents a distributed attitude control framework for the formation of multiple spacecraft, each modeled as a hybrid system coupling rigid-body and flexible structural dynamics. A multi-agent sliding mode controller is designed by embedding a consensus protocol into the sliding surface, enabling cooperative attitude regulation and synchronization under a predefined communication topology. The actuation system relies on a dual-gimbal variable-speed control moment gyroscope, whose electromechanical dynamics are considered. A hybrid steering strategy dynamically switches between gyroscopic and reaction-wheel operating modes depending on proximity to kinematic singularities. To enhance autonomy and reliability, an onboard fault detection and isolation module is integrated at the actuator level, implementing a single-failure-tolerant strategy: upon detection of a primary actuator fault, a redundant unit is activated; if the redundant unit also fails, the faulty spacecraft is automatically excluded from the communication network and the topology is reconfigured to preserve formation stability. Numerical simulations in Matlab/Simulink validate the proposed architecture in fault scenarios, demonstrating robust performance under external disturbances.

Keywords: Robust Control, Intelligent Control, Mechatronics Systems
Abstract: This paper proposes a decentralized control method that simultaneously achieves self-localization and formation tracking for a multi-UAV system using bearing information. Unlike conventional bearing-based approaches developed for first-order models, the proposed method is designed for second-order agent dynamics and explicitly handles unobservable situations, including collinear configurations and temporary sensor anomalies, through an auxiliary input. Finally, hardware experiments using a physical multi-UAV system demonstrate the practical effectiveness of the proposed approach.

Keywords: Modeling, System Identification and Estimation, Autonomous Decentralized Systems
Abstract: This paper addresses the design and analysis of a minimal-order distributed state observer for general linear systems over connected undirected networks. Each agent in the network constructs a local estimate based solely on its own measurements and exchanges it with neighboring agents to cooperatively reconstruct the full system state. A key feature of the analysis is the consideration of information processing delays affecting the exchange of estimates among neighboring agents. Leveraging the Small Gain Theorem, a sufficient condition for convergence of the distributed observer is derived, providing an explicit bound on the admissible delay in terms of the spectral properties of the network Laplacian, the coupling gain, and the undetectable system dynamics. Numerical results illustrate the theoretical findings and confirm the effectiveness of the proposed approach.

Keywords: Autonomous Decentralized Systems, Identification and Estimation
Abstract: Distributed state estimation has attracted significant attention in the context of multi-agent systems, where a group of agents cooperatively estimates the state of a dynamical system through local interactions over a communication network. A distributed observer can be interpreted as a synchronization problem over a network of interconnected sub-observers, in which each agent updates its local estimate based on its own measurements and information exchanged with its neighbors. Under suitable conditions, these local interactions enable all agents to asymptotically reach a common estimate of the true state. However, in practical situations, the estimation process is significantly affected by various constraints, including asynchronous and aperiodic sampling of sensor measurements, measurement delays, and inter-agent communication delays. To address these challenges, this paper proposes a distributed observer that explicitly accounts for asynchronous sampling and communication delays. Based on the estimated states, a state-feedback controller is constructed to stabilize the overall system. The effectiveness of the proposed method is validated through numerical simulations.

Keywords: Nonlinear Control, Multivariable Control
Abstract: This paper investigates the convergence properties of multi-objective min-Smith dynamics in population games. Agents revise strategies only if all objectives improve simultaneously, which yields a nonsmooth vector field. Without assuming a potential structure, we introduce a Pareto-gap Lyapunov candidate and a dissipativity condition based on the Clarke directional derivative. We prove that this condition drives the distance from the population state to the weak Pareto Nash equilibrium set to zero. For affine payoffs, finite LPs characterize coexistence of active Pareto-gap and positive-flow pairs. The two examples are LP-certified within stated numerical tolerances.

Keywords: Modeling, System Identification and Estimation, Social Systems, Autonomous Decentralized Systems
Abstract: This paper proposes an adaptive incentive design for population games. In situations where agents act out of self-interest, the Nash equilibrium and the socially optimal outcome do not necessarily align. To address this issue, we dynamically design incentives based on the agents' externalities, and completely align the Nash equilibrium with the socially optimal strategy. Furthermore, the proposed mechanism is designed to satisfy steady-state budget balance, ensuring that the sum of the incentives equals zero at the equilibrium.

Keywords: Robotic and Automation Systems, Signal and/or Image Processing, Networked Sensor System
Abstract: An implementation and an application of the combination of the genetic algorithm and Newton's method for solving a system of nonlinear equations is presented. The method first uses the robustness of the genetic algorithm to estimate the rough location of the roots, then uses the good convergence rate of Newton's method. An effective application of the method for the positioning problem of multiple small rovers, proposed for use in asteroid exploration, is demonstrated.

Keywords: Robotic and Automation Systems, Signal and/or Image Processing, Sensors and Transducers
Abstract: Aiming at bypassing standard simultaneous localization and mapping (SLAM) libraries and preserving critical hardware resources for other high-level tasks, this paper presents a custom-coded lightweight 2D LiDAR SLAM system designed specifically for resource-constrained mobile robots, with a practical implementation on the Raspberry Pi Mouse V3. The primary challenges involve balancing mapping accuracy with limited on-board processing power, while simultaneously compensating for significant odometry errors caused by wheel slippage. To address this, we investigate an optimization strategy based on data decimation to simultaneously reduce computational overhead and error accumulation. Experimental results reveal that a decimation factor of 5 is the optimal configuration that balances temporal noise Filtering and spatial overlap, achieving a precision of 20.84 mm while reducing processing time by 98.7%. Offline validation on a Raspberry Pi 4 confirms the system’s capability to process SLAM tasks within feasible time limits for this architecture. Our custom-coded approach demonstrates that high-fidelity SLAM is achievable on resource-constrained platforms through strategic data management, providing a robust foundation for future autonomous navigation.

Keywords: Robotic and Automation Systems, Signal and/or Image Processing
Abstract: This paper proposes a robust human tracking system for quadruped robots using asynchronous camera-LiDAR fusion. Monocular vision-based tracking suffers from inaccurate distance estimation due to target posture changes and occlusions. To address this issue, we integrate 2D object detection with 3D LiDAR point clouds without strict time synchronization. We introduce an extrinsic calibration method for sparse LiDAR and a data association scheme for asynchronous sensor fusion. Experimental results demonstrate highly accurate calibration with a mean reprojection error of 0.94 pixels. In addition, the proposed approach provides stable distance estimation independent of the worker’s posture, ensuring reliable real-time human tracking.

Keywords: Robotic and Automation Systems, Mechanical Systems Control, Laboratory Automation
Abstract: This paper discusses the dynamic modeling and feedback regulation system design of a Cartesian robot with a T-Bot motion structure suitable for pick-and-place task applications. The kinematics of the robot is derived using geometric analysis to determine the Cartesian translational motion of the robot along the X, Y, and Z axes with respect to variation in the joint variables of the robot. The dynamic model of the robot is then subsequently derived using the Euler-Lagrange equation to capture the energy-related relationship between force/torque acting on the system and the system response to given input. A feedback regulation scheme based on PID control framework is finally proposed to regulate the system response according to the desired motion profile and performance criteria. Simulation of the closed loop system performance are conducted and the results are presented to illustrate how the closed-loop system successfully reaches a desired target positions accurately with minimal position error and fast response time.

Keywords: Robotic and Automation Systems, Mechanical Systems Control
Abstract: The application of reinforcement learning to robots has advanced, making it possible to handle not only balance maintenance and walking but also more dynamic motions. This progress is largely due to the rise of robots equipped with DD / QDD motors, which are easy to model thanks to technological improvements. Robots have long been expected to assist humans, and cooperating with people requires high-precision control of posture and interaction forces. However, such behaviors are currently difficult to achieve with DD / QDD-equipped robots or purely learning-derived motions. On the other hand, model-based control using conventional machines with joints composed of DC motors and gears remains useful, but has limitations in dealing with complex environments and large disturbances. Therefore, the integration of the precision of conventional methods and the disturbance robustness of new methods would be a significant contribution. In this paper, we propose a framework for acquiring policies using widely used learning frameworks, using a conventional rigid-joint machine as a platform. The novelty of this paper is in a hybrid framework that applies reinforcement learning to a system with high-impedance joints, and the way roles are distributed within that framework. Furthermore, we demonstrate that Sim2Sim can be realized using this method.

Keywords: Robotic and Automation Systems, Factory Automation
Abstract: Accurate estimation of target grasp poses is essential for robotic object grasping. If the estimation is accurate, the robot can execute reaching motions and successfully grasp the object. In our previous work, grasp poses were accurately estimated for objects placed at different positions; however, the estimation accuracy degraded when the object orientation changed. To address this issue, we propose a two-stage grasp pose estimation approach. In the first stage, global grasp poses are estimated from an RGB image. In the second stage, hand displacements toward the grasp poses are estimated from local images generated around the target object to compensate for estimation errors in the first stage. Experiments using a dual-arm robot demonstrate that the proposed two-stage estimation method improves estimation accuracy and enables successful reaching and grasping of a known large object placed at different positions and orientations under the tested conditions.

Keywords: Robotic and Automation Systems, Robust Control, Nonlinear Control
Abstract: This paper presents the design, fabrication, and experimental validation of a hybrid flying robot integrating both a water nozzle module and a propeller module. The proposed platform employs four water nozzles to generate lift, while four propellers provide the control forces necessary for stability and maneuverability. To achieve robust angular motion control, a sliding mode control (SMC) scheme combined with a control allocation strategy is developed, enabling effective mapping of desired control inputs to feasible actuator commands. The stability of the closed-loop system is analytically guaranteed based on the uniformly ultimately bounded (UUB) property using Lyapunov theory. A prototype of the hybrid flying robot was constructed, and yaw angle tracking experiments were performed under both sinusoidal and step-like trajectories to comprehensively evaluate its dynamic performance. Experimental results demonstrate that the proposed system achieves stable flight and accurate trajectory tracking, while maintaining robustness against unpredictable disturbances and system uncertainties. These findings confirm the effectiveness and practical applicability of the proposed hybrid actuation configuration and control framework.

Keywords: Robotic and Automation Systems, Computational Intelligence, Intelligent Control
Abstract: The development of autonomous driving models based on Vision-Language-Action (VLA) frameworks requires a large volume of driving data. Even when a model with fundamental driving capabilities has been successfully developed, deployment to regions with different traffic regulations necessitates the collection of regional data, which poses a significant barrier to global scalability. In this paper, we propose a text-centric augmentation approach for generating data that reflects diverse traffic rules. The proposed method enables low-cost generation of such data by using text-printed images in place of camera images, thereby facilitating the training of VLAs under varying traffic rules.

Keywords: Medical and Welfare Systems, Signal and/or Image Processing, Computational Intelligence
Abstract: Chest CT classification remains challenging because diagnostically relevant abnormalities are often localized and volumetric annotation is costly. This paper proposes an Autoregressive-Masked Hybrid Mamba-Transformer (AR-HMT) for entry-based 3D CT classification. AR-HMT learns slice-level representations with a hybrid Mamba-Transformer encoder and an AR-masked decoder, and reuses the pretrained encoder for downstream classification. On a patient-level internal test set, AR-HMT achieves 93.06% accuracy and 97.43% AUC, outperforming representative CNN, Transformer, Mamba, and self-supervised baselines on accuracy, macro-F1, and AUC. Ablation results indicate that masked pretraining, autoregressive decoding, and hybrid encoding each contribute to the final performance.

Keywords: Medical and Welfare Systems, Human Interfaces, Sensors and Transducers
Abstract: Motor function in patients with severe upper-limb impairment is clinically evaluated based on the presence of slight voluntary muscle contractions at the origin of the pectoralis major below the clavicle, as assessed by palpation. In this study, to mimic a therapist's palpation, we developed a photo-reflector array to measure body-surface deformation at the origin of the pectoralis major below the clavicle. Using the measured body-surface deformation data, we also proposed a method for detecting voluntary movement based on the reconstruction error of principal component analysis (PCA). In the experiment, the body-surface deformation was measured during shoulder joint adduction/abduction under passive, semi-voluntary, and voluntary movement conditions. A detection threshold was derived from measurement data obtained during passive movement. The threshold exceedance rate was then calculated within movement ranges centered on full adduction for each of the three movement conditions. As a result, the exceedance rate was higher for semi-voluntary and voluntary movements than for passive movement. These results suggest that body-surface deformation measured with the developed photo-reflector array can be used to detect movement which includes voluntary muscle contraction.

Keywords: Medical and Welfare Systems, Human-Machine Systems, System Integration
Abstract: Individuals with total hearing loss have difficulty in recognizing their environmental sounds even when using cochlear implants (CIs). Environmental sounds are essential for detecting hazards and understanding surrounding situations; however, difficulty in perceiving these sounds directly affects both safety and independence. To improve independence, this study aims to develop a portable system to support environment recognition by providing sensory feedback on sound direction and type selectively. The system consists of two units. The sound localization feedback unit estimates the sound direction and presents it as tactile feedback. The sound classification feedback unit classifies environmental sounds and presents the results through visual and tactile feedback. To confirm the feasibility of the proposed system for supporting environment recognition, experiments were conducted with one participant with total hearing loss using CIs. The experiments consisted of three daily-life scenarios: rear danger avoidance, direction-of-arrival (DOA) search, and sound type recognition. The results show the accuracy of rear danger avoidance and sound type recognition were 66.7% and 90%, respectively, whereas the participant was unable to complete the tasks without the system. Moreover, DOA search with the system was approximately 2 s faster than without the system. These results indicate that the proposed portable system has potential for supporting environment recognition.

Keywords: Medical and Welfare Systems, Robotic and Automation Systems, System Integration
Abstract: In Somatic Cell Nuclear Transfer (SCNT), an oocyte is enucleated by cytoplasmic suction. As the nucleus cannot be seen under an optical microscope, determining suction volume depends on expert tacit knowledge. We developed a multi-task learning-based CNN–LSTM for a suction pressure control system in enucleation. The CNN–LSTM predicts the optimal suction stop timing based on expert enucleation data. We measured the prediction error of the CNN–LSTM’s suction stop timing against the ground truth and confirmed that the system successfully automated suction pressure control for porcine oocytes.

Keywords: Medical and Welfare Systems, Robotic and Automation Systems, System Integration
Abstract: The demand for intracytoplasmic sperm injection (ICSI) is increasing, resulting to a shortage of trained embryologists. In particular, detecting and tracking the polar body during ICSI remains a key challenge for novice embryologists. To address this challenge, we developed a high-speed visual assistance system that detects and tracks the polar body in the oocyte using YOLO11. The proposed system achieved 32 inference frame per second (FPS) in real-time performance, higher than the U-Net-based predecessor. The performance has been validated through recorded video analysis and a real-time experiment.

Keywords: Medical and Welfare Systems, Mechatronics Systems, Human Interfaces
Abstract: This study proposes a novel dressing assistance approach utilizing a clothes hanger. The proposed hanger with a reconfigurable geometry reshapes a garment (e.g., a jacket) so that the shoulder regions and armholes are expanded while the garment remains on the hanger. As a result, the user can put on the jacket as easily as if it were being assisted by another person. This approach eliminates the need to remove the jacket from the hanger or to manually arrange it prior to dressing, and reduces the effort required to locate armholes that are often obscured. To achieve this, the hanger reconfigures the garment from a stable hanging state into a dressing-ready configuration by modifying its support condition. A prototype hanger was developed, and its performance was evaluated through operation force measurements and dressing experiments using different types of upper garments. The results show that the required input force remains within a manageable range and that the system enables smooth arm insertion. These findings suggest that the proposed approach provides a simple and practical solution for dressing assistance without relying on complex systems.

Keywords: Medical and Welfare Systems, Intelligent Systems, Identification and Estimation
Abstract: This study proposes a method to estimate the internal stress distribution of a phantom from ultrasound echo images for neurosurgical training support, such as Sylvian fissure retraction. To estimate internal stress distribution, the Generative Adversarial Network (GAN)-based image synthesis model pix2pixHD was applied. The pix2pixHD trained adversarially with dataset consisting of ultrasound images capturing phantom deformation as input and corresponding structural simulation images obtained via non-linear finite element analysis (FEA) as ground truth. During training, to reduce blurring at the phantom edges, functions such as the VGGNet-based feature extractor of pix2pixHD architecture were disabled. The pix2pixHD generated high visually and quantitatively similar stress estimation images in terms of shape and color. Quantitative evaluations showed that the Intersection over Union (IoU) exceeded 0.99 for shape consistency. Moreover, the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) for stress values remained below 3% relative to the maximum stress of 0.252 MPa. However, challenges regarding the architecture of pix2pixHD and the method of coloring the stress values for visualization remain.

Keywords: Medical and Welfare Systems, Human-Machine Systems, Mechatronics Systems
Abstract: Transfemoral amputees must effectively control their prosthetic limbs, but the lack of sensory feedback makes walking acquisition challenging, especially for older individuals, leading to low success rates. One major obstacle to regaining gait abilities is knee buckling, which necessitates proper gait training. Current training methods heavily rely on the oral guidance of specialized staff, which can be difficult for some individuals to understand, potentially leading them to discontinues their training. The present study developed a system that provides real-time visual feedback on ground reaction force which is a mechanically key factor in preventing knee buckling. A real-time visual feedback system displaying hip joint moment and its target values for the prosthetic leg was implemented to facilitate understanding of gait mechanics and prevent knee buckling. A preliminary evaluation with two healthy participants wearing a simulated transfemoral prosthesis suggested that the proposed system may help users exert hip joint moments less likely to cause knee buckling. However, further evaluation with transfemoral amputees and improvement of the visual feedback interface are required.

Keywords: Safety, Environment and Eco-Systems, System Integration, System Engineering
Abstract: This study proposes a new method that enables the Functional Resonance Analysis Method (FRAM), which is used to detect accidents caused by functional resonance, to be applied to large-scale systems. First, the proposed method identifies functional couplings that are likely to generate functional resonance and represents them as directed graphs, referred to as functional resonance graphs. Next, it constructs a directed graph equivalent to the FRAM analysis diagram, called the system graph. Then, a subgraph isomorphism detection algorithm for directed graphs is applied to identify substructures that are isomorphic to the functional resonance graphs within the system graph. Furthermore, the method evaluates whether variations in timing or precision applied to the detected substructures may lead to accidents. When applied to the task analysis of a training ship’s departure operations, the proposed method successfully revealed accident possibilities caused by functional resonance that could not be detected using conventional FRAM.

Keywords: Safety, Environment and Eco-Systems, System Integration, Modeling, System Identification and Estimation
Abstract: Control based on a representative point is widely used for Vapor Pressure Deficit (VPD) regulation in a solar light type plant factory, but it remains insufficiently examined to what extent the control state of the whole space can be evaluated solely from tracking the target VPD at that point. This paper proposes an evaluation framework that combines vertical differences and residuals within each layer rather than relying only on tracking at the representative point, and presents one time series measured on March 5, 2026 as an example. We analyzed the VPD at the representative point, derived from the recorded dry-bulb temperature and relative humidity, together with VPD derived from SHT31 measurements at 27 spatial points arranged in a 3×3×3 grid. The mean absolute error (MAE) and root-mean-square error (RMSE) at the representative point were 0.0699 kPa and 0.0807 kPa, whereas the corresponding spatial values were 0.1698 kPa and 0.2148 kPa, with a maximum absolute error of 0.81 kPa. Comparison among heights indicated that the 1.5 m layer had smaller RMSE than the 0.5 m and 2.5 m layers, and residual RMS values exceeding the screening width based on the SHT31 typical accuracy were frequently observed after removing the planar trend in x and y at each height. This case study indicates that evaluating proportional (P) control based on a representative point should combine indices for vertical differences and residuals within each layer.

Keywords: Safety, Environment and Eco-Systems, Mechatronics Systems
Abstract: In this paper, we aim to actively exploit the high response characteristics of Magnetorheological (MR) fluid. The unidirectional constant-speed rotation generated by the motor is divided into two opposite rotational directions by a bevel gear mechanism. A novel control system is developed, in which the arm angle is regulated by switching between two fabricated MR brakes and adjusting their viscosity. The proposed system is evaluated through continuous positioning experiments and sinusoidal trajectory tracking experiments.

Keywords: Safety, Environment and Eco-Systems, Social Systems
Abstract: Recently, the number of traffic accidents in Japan has been decreasing; however, the proportion of fatalities involving vulnerable road users has been increasing. Accordingly, prediction and prevention of personal injury accidents involving vulnerable road users have become increasingly important. In this study, traffic accident data obtained from the Toyama Prefectural Police are used to examine a method for extracting locations with a high risk of personal injury accidents from property damage accident locations characterized by personal injury accident features derived from Formal Concept Analysis (FCA), and to visualize the types of personal injury accidents that are likely to occur. This report presents a similarity analysis of personal injury accidents that occurred in the vicinity of property damage accident locations possessing personal injury accident attributes extracted using FCA. The similarity between the personal injury accident attributes and accident attributes of personal injury accidents that occurred in the vicinity was calculated, and its relationship with an index TP’ based on the normalized numbers of objects and attributes in the implication structure was analyzed. As a result, it was confirmed that higher values of TP’ tend to correspond to higher similarity. Furthermore, focusing on the average value, the similarity in the upper range of the index was higher than that in the lower range.

Keywords: Virtual Reality Systems, Human Interfaces, Signal and/or Image Processing
Abstract: In augmented reality and aerial display systems, reproducing accurate occlusion between dynamic physical hands and virtual objects is essential to ensure spatial consistency. While point cloud-based occlusion estimation provides geometric accuracy, applying it directly incurs an increased computational load due to exhaustive per-pixel evaluations. To address this issue, we propose a hybrid rendering pipeline that integrates dynamic measured point clouds and polygon-based virtual objects into a unified depth representation. Our system introduces an architecture utilizing a geometry buffer to identify the data source of each pixel. By framing the occlusion calculation as a visibility determination based on Hidden Point Removal (HPR) for sparse measured point clouds, this architecture restricts the computational targets strictly to the boundary regions where virtual objects overlap measured point clouds. This approach skips calculations for non-target regions, optimizing the computational load to scale proportionally with the overlap area of the heterogeneous data. Comparative experiments confirm that our method reduces the occlusion processing time by up to 51.6% compared to the conventional method. The proposed pipeline achieves a real-time performance of 60 fps (processing time < 16.6 ms) while maintaining occlusion accuracy equivalent to the conventional exhaustive per-pixel evaluation methods and suppressing self-occlusion artifacts.

Keywords: Virtual Reality Systems, Human Interfaces, Robotic and Automation Systems
Abstract: Autonomous mobile robots are increasingly deployed in environments shared with humans, where their motion and appearance can influence perception and behavior. However, previous studies have generally examined appearance and motion separately, and platforms that independently manipulate both factors within the same scenario remain limited. This study proposes a virtual reality (VR)-based platform for the integrated evaluation of robot motion and appearance. The platform independently controls robot height and deceleration time during approach, allowing quantitative measurement of psychological responses under consistent conditions. A preliminary experiment examined arousal and pleasure in six participants. Robot height significantly affected both measures, whereas deceleration time affected arousal only. No statistically significant interaction was detected. These results demonstrate the feasibility of the platform, but do not establish independent effects. Because VR does not fully reproduce physical mass, collision risk, or bodily responses, the findings should be interpreted as preliminary results obtained under controlled virtual conditions.

Keywords: Network System Integration, Mechatronics Systems, Robotic and Automation Systems
Abstract: In this paper, we proposed a controller-based prediction method for data compression in bilateral control. For position prediction, we employed a method combining the Position Verlet method and a disturbance observer (DOB), while for force prediction, we utilized the reaction force observer (RFOB) directly. Evaluation experiments using data from an actual bilateral control system confirmed the effectiveness of the proposed method. The experimental results demonstrated that the proposed method achieved higher compression ratios compared to Linear Prediction (LP) and Adaptive Linear Prediction (ALP).

Keywords: Network System Integration, Mechatronics Systems, Human Interfaces
Abstract: Packet loss and stochastic delays in real networks significantly impede the transparency of bilateral teleoperation systems. Traditional interpolation methods, such as first-order hold (FOH), suffer from significant overshoot when predicting nonlinear signal peaks. This study presents error-feedback adaptive first-order hold (EFB-AFOH), a novel interpolation framework that dynamically corrects linear extrapolation using a feedback mechanism based on previous prediction errors. This study provides a rigorous mathematical definition of the algorithm and its physical interpretation as a hybrid of inertial prediction and restorative compensation. Comprehensive sensitivity analyses demonstrate that EFB-AFOH consistently outperforms conventional methods by adapting to signal complexity and communication quality.

Keywords: Network System Integration
Abstract: Continuous monitoring of body-surface temperature distributions is useful for wearable health and thermal-condition monitoring, but distributed sensing across garments remains difficult because multiple nodes require both power delivery and data collection. Conventional wiring increases complexity, limits placement flexibility, and degrades wearability. We present a conductive-textile-based wearable temperature sensing system in which the textile serves as a shared bus for DC power and data communication. Each sensor node is a battery-free circular board, 21 mm in diameter and 0.9 g in weight, integrating a commercial temperature sensor and custom circuitry for UART/DC superposition. Address-based half-duplex UART with master polling enables collision-free multi-node communication. An 18-node array was implemented on a garment worn by a mannequin together with a real-time temperature mapping viewer. In a thermal scanning experiment using warm air from a heat gun set to approximately 50 °C, the system captured spatiotemporal temperature variations over the garment surface at a refresh rate of 99.2 Hz. The results demonstrate that the proposed conductive-textile bus supports flexible node placement and real-time, high-refresh-rate distributed temperature mapping, providing a practical basis for garment-integrated body-surface monitoring and future multi-sensor wearable systems.

Keywords: Simulation of Large Systems, Autonomous Decentralized Systems, Social Systems
Abstract: Social networks are typical complex systems in which individual interests can produce undesirable collective outcomes. Distributed graph coloring has been used to study how coordination emerges, but the extent to which network topology affects difficulty independently of the feasible-coloring landscape remains unclear. We conduct agent-based simulations of 17,000 graphs generated by the Barabási–Albert (BA) model to distinguish degree heterogeneity from a structural proxy for solution-space constraint: the size of the maximum uniquely colorable subgraph (UCS). Across the generated graphs, a larger maximum UCS is associated with longer convergence time. Degree heterogeneity has almost no marginal correlation with difficulty; however, after stratifying graphs by maximum UCS size, higher degree heterogeneity is associated with faster convergence. Thus, within the BA ensemble and local update rule considered here, the apparent difficulty of hub-rich networks cannot be attributed to degree heterogeneity alone. This interpretation complements prior topology comparisons by identifying solution-space constraint as a distinct explanatory factor.

Keywords: Simulation of Large Systems, Intelligent Systems, Computational Intelligence
Abstract: Urban growth simulation at country-wide scales requires balancing spatial detail with computational efficiency. We present a GPU-oriented cellular automata framework for simulating urban expansion and population redistribution over millions of cells while preserving a 100 m base resolution. The objective is to provide a framework substantially faster than typical CPU-oriented urban CA workflows while retaining enough spatial detail for regional analysis, enabling national grids, scenario sweeps, and future hyperparameter search. The model combines Global Human Settlement Layer products, terrain constraints, water context, and macroeconomic and demographic drivers. CUDA kernels update the cellular state in parallel and maintain aggregation layers, matching the local stencil and per-cell scoring structure of the author-defined model to massively parallel hardware. Experiments show that the current calibration is most reliable for aggregated built-up surface structure: in a 2010–2025 validation run, relative mean absolute error reaches 0.304 at 10 km for built-up surface and 0.903 at 10 km for population.

Keywords: System Integration
Abstract: Heterogeneous multi-robot systems often combine robots based on different middleware frameworks, making unified monitoring and upper-layer integration difficult. This paper presents a middleware abstraction architecture for monitoring robots based on ROS1, ROS2, and RT-Middleware through a common interface. Middleware-specific state data are converted into a unified internal representation and transmitted to a server-side monitoring framework. The primary contribution is the separation of middleware-dependent robot processing from middleware-independent monitoring and upper-layer interfaces. VDA5050 is adopted as an example of a fleet-level standard to which the unified representation can be partially mapped, rather than as a claim of full protocol compliance. A prototype demonstration showed that heterogeneous middleware data could be normalized and visualized through the same web-based interface. The current scope, limitations, and future extension toward quantitative performance evaluation and broader VDA5050 support are also discussed.

Keywords: Entertainment Systems, Human Interfaces, Human-Machine Systems
Abstract: In recent years, DJ entertainment has become a prominent global trend, serving as a vital space for social interaction through music. However, with the widespread popularization of DJing, events have increasingly shifted toward the passive appreciation of live performances. This shift has resulted in a one-way flow of information from the DJ to the audience, leading to a significant lack of interpersonal interaction among participants. To address this issue, this study proposes an embodied-expression interface designed to measure and visualize the body movements of all participants at a DJ event. We established a method to calculate the tempo of body movements synchronized with music from acceleration waveforms and developed a standalone wearable device that changes light colors in real time based on this analysis. Furthermore, this paper evaluates the performance of the device by considering its optimal placement on the body. We also conducted verification tests in a multi-user environment during music playback to confirm the system's operational stability. The results suggest that the proposed interface effectively visualizes physical expression and has the potential to facilitate rhythm synchronization across the entire venue.

Keywords: Virtual Reality Systems, System Integration, Human-Machine Systems
Abstract: This study investigates how accurately human walking direction can be predicted using a minimal set of skeletal points. While previous approaches often rely on full-body data, we show that only three points, the chest and both ankles, are sufficient for machine learning models to predict left/right turns with high accuracy. Using Kinect-based 3D skeletal data and a simple neural network, we achieved 90% accuracy approximately 70 ms before the stance leg makes contact. A psychophysical experiment confirmed that humans can perform above chance with the same limited information. These results suggest that a minimal skeletal input set can be effective for gait prediction, with implications for efficient motion modeling in robotics and avatar systems.

Keywords: System Integration, Human-Machine Systems, Virtual Reality Systems
Abstract: This study proposes a system for quantitatively evaluating guide dog training using a handle-mounted force sensor and a shoulder-mounted 360-degree camera. Interaction through the handle is difficult to observe externally, making skill transfer to novice trainers challenging. The proposed system records interaction force and handle motion, enabling objective comparison between experienced and novice trainers. Frequency analysis revealed distinct characteristics in both the force and handle-angle signals, suggesting that trainer skill is reflected in the dynamics of physical interaction with the guide dog. These results demonstrate the feasibility of quantitative evaluation of guide dog trainer skill.

Keywords: Modeling, System Identification and Estimation, Nonlinear Control
Abstract: Accurate dynamic modeling of robotic arms is essential for achieving high-precision control. However, traditional physics-based modeling often struggles to capture complex nonlinearities, such as joint friction and cable tension, which typically require tedious parameter identification. To address this challenge, this study utilizes Koopman operator theory as a data-driven alternative. This theory can represent nonlinear dynamics in a higher-dimensional linear space, thereby allowing the application of linear control techniques to complex nonlinear systems. In this study, we present a data-driven modeling and control framework for a physical robotic arm. First, the nonlinear dynamics of the arm are identified from experimental data using the Koopman operator to construct a linear predictor of the system states. Next, leveraging this identified model, we implement the computed torque method to achieve precise trajectory tracking. By embedding the Koopman-based model into the control loop, the system can compensate for unmodeled dynamics without relying on explicit physical equations. The proposed framework was evaluated and validated through real-world experiments on a multi-joint robotic arm. The results demonstrated that the Koopman-based computed torque control successfully reduces tracking errors, confirming its effectiveness and practicality for complex robotic systems.

Keywords: Intelligent Systems, System Integration
Abstract: Vapor Pressure Deficit (VPD) is an important environmental parameter in sunlight-type plant factories because it strongly affects transpiration, stomatal behavior, and crop productivity. In this study, experiments were conducted using a mist-spraying VPD control system developed by Kurimoto et al., which enables stable humidity and VPD control inside the cultivation environment. However, conventional environmental control systems generally rely on a single representative sensor, making it difficult to accurately estimate spatial microclimate distributions. Previous studies by Kurimoto employed a high-density 3×3×3 sensor array with 27 sensors installed at different heights and estimated three-dimensional VPD distributions using Kriging interpolation. The results showed that three-dimensional estimation reduced VPD estimation errors by approximately 50% compared with conventional two-dimensional approaches. In addition, localized VPD deviations of up to 0.81 kPa were observed even when the central sensor maintained the target value of 1.0 kPa. Furthermore, Kurimoto developed an HPC-based thermofluid simulation capable of reproducing spatial VPD distributions, forming the basis of a digital twin environment. This study investigates AI-integrated spatial VPD estimation methods using measured environmental data and digital twin simulation data. Kriging interpolation and AI-based approximation models are compared to evaluate estimation accuracy and spatial reproducibility.

Keywords: Mechanical Systems Control, Modeling, System Identification and Estimation, Nonlinear Control
Abstract: The Torque-Unit Manipulator (TUM) is a conceptual design for a space manipulator in which all joints are passive and the links are driven by actuators called torque-units mounted on the links. A torque-unit can be simply constructed using a motor and a rotating disk, and since no motor is required inside the joints, high reliability can be achieved. In the position control of the TUM, when attention is focused only on the link angles, the system can be represented by equations of motion similar to those of conventional manipulators, allowing existing control laws to be applied. However, residual angular velocities remain in the disks after the links reach their target positions. Since the TUM can be regarded as a nonholonomic system, this problem may be resolved by appropriately planning the motion trajectories of the links. Previous studies theoretically demonstrated that this problem can be resolved for a TUM restricted to planar motion. To extend this result to three-dimensional motion, a TUM structure capable of three-dimensional motion was proposed and its equations of motion were derived. Furthermore, the dynamic characteristics were investigated based on the derived equations of motion.

Keywords: Modeling, System Identification and Estimation, Signal and/or Image Processing, Advanced Computics
Abstract: Traditionally, cosmetic prototyping has required considerable time and cost due to repeated experimental fabrication and evaluation processes. In addition, controlling nanoparticle size remains a significant challenge in cosmetic material development because particle size strongly affects optical properties and the resulting skin appearance. This study aims to reduce development time and cost by introducing ray-tracing simulations for nanoparticle size control and early-stage evaluation of skin appearance. By incorporating simulation-based evaluations into the cosmetic development process, the visual effects of cosmetic materials can be predicted prior to physical prototyping. Furthermore, a skin model for evaluating the appearance of foundation-applied skin was developed. The proposed model reproduces the optical characteristics of human skin by considering light scattering and absorption within the skin structure. In particular, the model represents skin tone using particles associated with pigments such as melanin and hemoglobin, while also reproducing skin blemishes and surface irregularities. Through these representations, a more realistic simulation of skin appearance was achieved, contributing to the efficient development of cosmetic products.

Keywords: Adaptive and Optimal Control, Modeling, System Identification and Estimation, Transportation Systems
Abstract: Train Automatic Stop-position Controller (TASC) system is used to automatically stop train at the desired position. TASC system generates discretized brake-notch outputs that are more finely notch level than those available in manualoperation. Because of that, older trains require additional customization to accommodate more finely divided discrete brake-notch commands. Moreover, train brake conditions depend on driving situation such as number of passengers and external environment such as the weather, and thus control systems using train physical models have limitations for required and detailed dynamics. This paper proposes that the application of Data-Enabled Predictive Control (DeePC) to automatic train stopping. Without using physical model, DeePC can predict the future train behavior using measured input-output data. The proposed approach is designed to achieve accurate stopping performance even when only small number of brake-notch commands are available. In addition, the controller is expected to maintain the optimal trajectory even though disturbances existing such as changes in adhesion caused by rain. Numerical validation results demonstrate the feasibility of DeePC-based automatic stopping control for trains with limited brake-notch resolution.

Keywords: SCADA and Network Systems, Intelligent Systems, Process Control
Abstract: Proportional-Integral-Derivative (PID) control methods are widely applied in process systems. However, their performance deteriorates due to variations in system parameters caused by plant aging and environmental changes. A conventional method based on the variance criterion of input/output data (EMV-index) suffers from delayed detection and false alarms because the detection threshold and evaluation data length are difficult to tune. To address this problem, this study proposes a novel neural-net based control performance assessment mechanism. The proposed method determines the system state by simultaneously calculating multiple evaluation indices using various evaluation lengths and feeding them into a pre-trained neural network. This approach reduces the need for complex parameter tuning, such as determining thresholds and data lengths for control performance assessment. Furthermore, the effectiveness of the proposed method is quantitatively verified through numerical studies using a linear time-varying system, demonstrating that it can detect system variations significantly earlier and with fewer false alarms than the conventional method.

Keywords: Signal and/or Image Processing, Intelligent Systems, Robotic and Automation Systems
Abstract: We propose a structure-based visual feedback method for autonomous navigation of a patrol imaging vehicle in elevated strawberry cultivation. Unlike conventional autonomous navigation approaches that rely on artificial markers, magnetic tape, rails, or range sensors, the proposed method uses existing overhead structures as visual cues, reducing the need for additional infrastructure. Images are captured by an upward-facing camera mounted on the vehicle, and two main overhead pipes are detected as navigation references. In contrast to our previous single-pipe angle-based approach, this study estimates the horizontal coordinate v_x of the vanishing point formed by the two pipes and uses it as a heading-error signal for differential-drive steering. Line candidates are extracted from edge images and refined using angle filtering, pipe-width verification, and robust line fitting. The vanishing point is smoothed by a moving average and predicted using a Kalman filter to maintain a continuous steering reference during short-term detection failures. Arch pipes installed at 0.5 m intervals are also counted for coarse traveled-distance estimation. Evaluation using 1,776 real greenhouse frames showed that both pipes were detected in 81.3% of the frames and at least one pipe in 95.4%, demonstrating the feasibility of low-infrastructure autonomous navigation using existing cultivation structures.

Keywords: Network System Integration, Networked Sensor System, Intelligent Systems
Abstract: In body-linked interaction and remote physical collaboration systems using functional electrical stimulation, monitoring the synchronization state among users is important. One possible approach is real-time streaming of point cloud data representing body movements. However, point cloud data require large bandwidth, and bandwidth contention among multiple users can make it difficult to maintain real-time performance. Furthermore, conventional methods with a fixed number of transmitted points require prior tuning for each network condition, making them unsuitable for network environments where congestion conditions change dynamically. This study investigates an adaptive point cloud streaming control method that does not require prior tuning. The proposed method dynamically adjusts the number of transmitted points based on the application delay requirement and the network congestion state. Through simulation evaluations in an environment where multiple users share the same network bandwidth, the ratio of point cloud data arriving before the delay deadline is investigated. The results demonstrate that the proposed method maintains a high arrival ratio without prior configuration. The proposed approach is expected to serve as a real-time monitoring framework that supports synchronization-state monitoring among users in future body-linked interaction systems.

Keywords: Robust Control, Modeling, System Identification and Estimation, Network System Integration
Abstract: Networked control systems offer physical flexibility but are susceptible to security risks, communication overhead, and potential disconnections. To address these challenges simultaneously, this poster presents an encrypted event-triggered tracking control system capable of detecting network disconnections while maintaining data privacy. In this work, two distinct encryption methodologies are developed to secure the tracking control system. The first method encrypts the conventional event-triggered tracking system by integrating the ElGamal encryption scheme. In contrast, to mitigate the performance degradation caused by quantization errors due to encryption, the second method reformulates the control structure by altering the insertion point of the high-gain feedback. This structural modification successfully suppresses the impact of the quantization errors introduced into the system. Numerical simulations demonstrate that while both methods maintain privacy and successfully detect network failures, the second method significantly reduces state oscillations caused by encryption errors. Furthermore, the tracking performance and accurate disconnection detection of the proposed system are successfully verified even under external disturbances.

Keywords: Mechanical Systems Control, Factory Automation, Identification and Estimation
Abstract: Cyber-physical systems, including networked industrial control systems, can be exposed to cyberattacks that eavesdrop on operational status and may cause physical damage. To protect control parameters and communication signals from unauthorized observation and manipulation, this poster presents an encrypted control system for an industrial linear motor. The target system is an iron-cored permanent-magnet linear synchronous motor whose position-tracking performance is affected by the nonlinear attractive force between the magnets and the cores. The position and velocity controllers are implemented with a homomorphic encryption scheme, enabling encrypted control parameters and encrypted communication signals during operation. Experimental results show that the encrypted controller achieves position-tracking performance comparable to that of the unencrypted controller. The poster also summarizes cyberattack-detection experiments in which falsification and replay attacks are detected by monitoring the decrypted control input.

Keywords: Identification and Estimation, Signal and/or Image Processing
Abstract: Guide dog training involves continuous and subtle interactions between a dog and its trainer. Previous studies have explored assistive technologies and canine behavior analysis. However, quantitative analysis of guide dog interaction remains difficult because real training scenes involve occlusion, multi-person ambiguity, and complex motion patterns. While 2D pose estimation itself is generally stable, trainer identity can become ambiguous in scenes involving multiple people. To improve temporal consistency, this study uses the dog as a temporally consistent interaction anchor, enabling more reliable identification and tracking of the corresponding dog–trainer pair from multi-view videos. The method reconstructs synchronized 3D skeletons of both the dog and the trainer by combining robust 2D pose estimation, target tracking, temporal association, and multi-view triangulation. By providing temporally consistent 3D skeletal data from real training scenes, this method offers a basis for objective evaluation of guide dog–trainer interaction and further analysis of how dogs and trainers coordinate their movements during training.

Keywords: Adaptive and Optimal Control, Robotic and Automation Systems, Modeling, System Identification and Estimation
Abstract: This study proposes an online model predictive control (MPC) framework for collision avoidance among multiple differential-drive robots operating in constrained indoor environments. The method is intended for dynamic situations, such as restaurants or service facilities, where unexpected obstacles, blocked passages, or timing deviations require robots to recompute avoidance trajectories during operation rather than follow only preplanned routes. Conventional MPC-based collision avoidance typically resolves conflicts by modifying the trajectories of directly interacting robots. However, such local avoidance maneuvers may induce secondary collisions or deadlocks with nearby robots and obstacles, particularly in narrow passages, corners, and bottleneck regions. To address this issue, the proposed framework predicts pairwise conflicts, generates multiple avoidance candidates with different burden allocations, and evaluates the secondary collision risk associated with each candidate. Based on this risk assessment, the avoidance burden is assigned to the robot whose maneuver is expected to minimize secondary conflicts while maintaining goal-directed navigation performance under MPC. Simulation studies are conducted to assess the proposed method in comparison with a conventional MPC-based approach. Performance is evaluated in terms of travel distance, computational cost, and successful task completion.

Keywords: Intelligent Control, Autonomous Decentralized Systems, Intelligent Systems
Abstract: Autonomous attacking aerial swarms using nonlinear, evading dynamics present a severe asymmetric threat to critical infrastructure, frequently overwhelming standard constant-velocity intercept logic. This study presents a rigorous comparative evaluation of task assignment architectures; spanning centralized optimization, distributed auctions, and emergent heuristics in a dynamic many-to-few interception scenario. Using a custom simulation environment with reproducible, paired-seed statistical validation, we benchmarked these methods against ’boids’ attackers that employ continuous acceleration, flocking, and evasion behaviors. Specifically, we investigated the adaptation of the Consensus-Based Bundle Algorithm (CBBA) for this highly dynamic scenario by tuning hyperparameters and introducing feasibility fallback bids and adaptive time horizons. Through comprehensive ablation studies, we demonstrate that these auction modifications actually degrade performance unless paired with a ”Guard Layer”—a reactive spatial decomposition strategy that locally reserves defenders at the protected targets. The synthesized boids-aware CBBA with the Guard Layer architecture emerged as the dominant strategy. It achieved a 55.5 percent perfect-defense rate and 0.83 mean target hits over 200 baseline runs, reducing breaches by 60 percent compared to standard centralized and distributed baselines.

Keywords: Multivariable Control, Modeling, System Identification and Estimation, Simulation of Large Systems
Abstract: This study proposes a structure-preserving model reduction method for multi-agent systems with identical subsystem dynamics and Laplacian-type interconnections. In large-scale multi-agent systems, model reduction is useful for reducing computational complexity in analysis, simulation, and controller design. This study focuses on balanced truncation as the model reduction method. If balanced truncation is directly applied to the whole system, the Kronecker product structure, which reflects the network structure of agents, may be destroyed, and the resulting reduced-order model may lose its original network interpretation. To preserve this structure, this study proposes the use of common generalized Gramians. In the first step of the proposed approach, the whole system is decomposed into a number of subsystems through the diagonalization of the graph Laplacian. Then, common generalized controllability and observability Gramians are computed by solving linear matrix inequalities for all of these subsystems. The resulting common Gramians yield a single balancing transformation that can be applied to all the subsystems uniformly. Consequently, the reduced-order model preserves the original Laplacian network structure while reducing only the internal state dimension of each agent. Numerical simulations demonstrate the effectiveness of the proposed method.

Keywords: Signal and/or Image Processing, Identification and Estimation
Abstract: Dynamic mode decomposition (DMD) has been commonly used as a method to extract spatiotemporal features from multidimensional time-series data. DMD is a data-driven decomposition technique that represents multidimensional time-series data as a superposition of spatiotemporal modes, each associated with distinct temporal dynamics characterized by a single frequency and a single growth or decay rate. DMD has been shown to be effective as a feature extraction method for predicting individual differences from brain activity. Among several DMD algorithms, exact DMD (exDMD) is one of the most widely used. However, exDMD is sensitive to noise in the data and may therefore fail to accurately reconstruct multidimensional time-series data. To overcome the limitation, optimized DMD (opDMD) was developed. Although opDMD is known to reconstruct time-series data more accurately than exDMD, whether it is effective for predicting individual differences from brain activity remains unclear. In this study, we used publicly available data from the Human Connectome Project to compare exDMD and opDMD as feature extraction methods for predicting individual differences from brain activity. Our preliminary results showed that exDMD and opDMD achieved nearly comparable prediction accuracy. These findings suggest that exDMD remains a viable feature extraction approach for individual-difference prediction based on brain activity.

Keywords: Modeling, System Identification and Estimation
Abstract: Koopman operator theory enables nonlinear dynamics to be represented as linear evolution of observables. Most previous studies, however, target smooth and continuous systems, and hybrid systems involving contact, where velocity reverses discontinuously, remain challenging. We investigate whether Convolutional Koopman Autoencoders (CKAEs) can capture such hybrid dynamics directly from image sequences without explicitly splitting the data at mode-switching, contact events. As a baseline, we compare the CKAE against Hankel DMD on pendulum image sequences using delay embedding. The CKAE achieves better image reconstruction and long-horizon prediction. We then apply the CKAE to images of an elastic ball bouncing inside a two-dimensional box. Preliminary results show that sequences with and without contact events exhibit different characteristics with respect to the eigenvalues of the Koopman operators. These findings indicate that learned Koopman embeddings can unify continuous and discrete behavior for hybrid dynamical systems.

Keywords: Modeling, System Identification and Estimation, Mechatronics Systems, Adaptive and Optimal Control
Abstract: This study aims to develop a crawler-type rescue robot for high traversability in unstructured environments. Conventional rigid crawlers lack terrain adaptability, while previous internal flexible mechanisms suffer from poor reliability and maintainability. To address these issues, we propose "RODEM" (Real wOrlD Elastic Mobile robot), which features an external flexible mechanism. RODEM combines active curving via linear actuators and passive compliance using springs. An extensible crawler belt with a slotted slide mechanism was newly developed to absorb circumference variations during significant curving motions. A mathematical model of the multi-link arm system was constructed for quasi-static equilibrium analysis based on potential energy minimization. Simulations verified that the extensible belt operates successfully within permissible circumference changes. Experimental validation with a prototype confirmed that the belt remained securely attached and drove smoothly even at maximum curvature. Moreover, RODEM successfully overcame a 100-mm step—a challenging obstacle for rigid crawlers—through active shape adaptation. In conclusion, integrating an external flexible mechanism with an extensible belt creates a novel mobile system that achieves both high maintainability and exceptional rough-terrain traversability.

Keywords: Autonomous Decentralized Systems, Computational Intelligence
Abstract: In recent years, Graph Neural Networks (GNNs) have attracted significant attention for learning graph-structured data and have been successfully applied across various fields. For large-scale graphs, optimizing GNNs through multi-agent distributed optimization offers substantial benefits in terms of computational load and robustness. However, conventional distributed optimization for GNNs suffers from high communication costs because it requires communication with neighboring agents during the backpropagation process. In this study, we formulate the distributed optimization of GNNs within a multi-agent system and propose an algorithm that utilizes approximately calculated gradients, eliminating the need for communication during backpropagation. Numerical experiments demonstrate that the proposed method effectively reduces communication costs while maintaining performance comparable to conventional methods, thereby validating its effectiveness.

Keywords: Modeling, System Identification and Estimation
Abstract: Large space structures (LSSs) are often controlled by direct velocity and displacement feedback (DVDFB), whose robust stability properties make it suitable for flexible space systems. This study investigates whether data-driven prediction of a DVDFB closed-loop response remains possible when an unknown constant disturbance is applied at the input port. Using initial input-output data and known DVDFB gains, we formulate a data-driven prediction problem that simultaneously estimates the closed-loop transfer function and the disturbance magnitude. However, exact reproduction of the measured output does not necessarily imply unique identification. We therefore analyze identifiability from the viewpoint of model order constraints and pole-zero cancellation, showing that false solutions can remain when additional pole-zero factors are canceled. The results clarify conditions under which disturbance-aware data-driven prediction is identifiable or non-identifiable for DVDFB-controlled LSSs.

Keywords: Adaptive and Optimal Control, Modeling, System Identification and Estimation, Identification and Estimation
Abstract: This poster presents results from a proof-of-concept hardware experiment on an online parameter estimation method based on simple adaptive control (SAC). The method, which is currently under development, estimates unknown plant parameters from the steady-state adaptive gains using only the plant input and output. Under the zero-matching condition that a SISO almost strictly positive real plant and its reference model have matching zeros, and assuming that both systems are represented in observable canonical form, the unknown plant parameters can be explicitly recovered from these adaptive gains. The key novelty of the method is the explicit recovery of unknown plant parameters from the steady-state adaptive gains of SAC. Because the estimation is embedded in the adaptive control loop, it can be performed online even for unstable plants while maintaining closed-loop stability. Previous numerical studies demonstrated accurate estimation for third-order systems satisfying the zero-matching condition. The poster outlines the estimation procedure and presents preliminary hardware results, focusing on practical feasibility and estimation accuracy.

Keywords: Mechanical Systems Control, Mechatronics Systems, Identification and Estimation
Abstract: Low-impedance control is essential for transparent physical interaction in rehabilitation robotics; however, maintaining low apparent impedance under fast disturbances remains challenging because of actuator dynamics, friction, and control delay. This study aims to establish active rendering of virtual impedance based on a zero-impedance-capable implementation of the Fusion Hybrid Linear Actuator (FHLA) concept, which integrates two power sources; the present implementation combines the high force capability of pneumatic actuation with electromagnetic responsiveness. The key concept is to use the FHLA-based actuator as a transparent actuator platform and to render active virtual impedance on top of its low-impedance baseline. Transparent interaction is realized through active feedback control of both pneumatic and electromagnetic components, including precise current control of the electromagnetic element for fast force compensation. As an initial evaluation of this concept, zero-impedance control is tested under fast disturbance conditions up to 150 mm/s by measuring residual interaction force, apparent damping, and response delay. The results clarify the operating range in which zero-impedance behavior can be maintained and provide preliminary design criteria for future active rendering of virtual impedance.

Keywords: Adaptive and Optimal Control, Guidance and Flight Control
Abstract: In recent years, the demand for autonomous wheelchairs has been increasing to support mobility. However, ground-level sensors often face limitations due to occlusions and blind spots. Utilizing a drone’s bird’s-eye view can significantly enhance safety by providing a broader environmental perspective. This research focuses on generating autonomous trajectories for a drone that tracks a moving wheelchair while navigating complex environments containing both aerial and ground-level obstacles. We employ Model Predictive Control to optimize the drone's flight path, ensuring it maintains a stable view of the wheelchair through camera field-of-view constraints. Obstacle avoidance is formulated using the Big-M method, enabling the drone to navigate safely through narrow spaces between tree trunks and under canopies. Simulation results demonstrate that the proposed system effectively generates a safe and smooth trajectory, allowing the drone to track the wheelchair continuously while avoiding all obstacles. This approach provides a robust framework for aerial-assisted navigation, contributing to safer mobility in realistic environments where ground-level visibility is restricted.

Keywords: Nonlinear Control
Abstract: Almost every bounded trajectory of a strongly cone invariant delayed system converges to an equilibrium. Relying on this, we derive a component-wise sufficient condition for multi-stability of cone invariant delayed networks. We first show that a delayed network is eventually strongly cone invariant if each cone invariant delayed subsystem is excitable and transparent, and if every delayed interconnection preserves cone invariance in a strong sense. Then, we present a component-wise multi-stability condition based on linearization.

Keywords: Social Systems
Abstract: Fairness is often introduced as a design objective in resource allocation problems. However, allocations that appear fair from an external perspective do not necessarily remain stable against coalitional deviations. This mismatch reveals a gap between fairness and stability in cooperative systems. This paper presents a framework to implement a prescribed fair allocation as a stable outcome through external interventions. We consider three intervention schemes under different principles: subsidy, tax, and combined intervention schemes, and examine their different roles in supporting stability. A numerical case study illustrates how the required intervention and the resulting allocations depend on the trade-off between two fairness criteria.

Keywords: Nonlinear Control
Abstract: Data-driven design problem of finding a state feedback controller guaranteeing λ-contractivity is considered for discrete-time linear systems subject to input saturation. λ-contractivity is a generalized property of robust positive invariance, and can be used for local stabilization of the system. Building on several results in model-based framework, we develop a data-driven controller design method for λ-contractivity using linear programming.

Keywords: Multivariable Control, Modeling, System Identification and Estimation
Abstract: This paper studies cyber attacks against informativity-based analysis in data-driven control. Focusing on strong observability, we consider an adversary who post-processes finite time-series data by an invertible linear transformation acting on the data matrices. We show that such transformations are capable of embedding malicious states into the invariant subspace explained by the transformed dataset. We provide a constructive attack method and derive feasibility conditions that characterize when such transformations exist. Moreover, we formulate an optimization problem to obtain the minimum-norm attack that quantifies the smallest data distortion required to destroy informativity. Numerical examples demonstrate that small and structured transformations can invalidate informativity certificates.

Keywords: Adaptive and Optimal Control, Nonlinear Control, Intelligent Control
Abstract: This paper presents a novel design method of finite-time switching signals in nonlinear switched systems. The design problem of switching signals is typically formulated as a combinatorial optimization. To address this, we propose a relaxation-based approach that yields a numerically tractable continuous optimization problem. The efficacy of the proposed methodology is then demonstrated through a design example using the Lotka-Volterra competition equations as a representative nonlinear system.

Keywords: Robust Control
Abstract: This paper proposes an H-infinity controller design method for achieving type 2 output sensitivity characteristics in type 1 servo systems. This method improves the performance of suppressing output variations caused by plant uncertainties, independent of servo characteristics. Closed-loop systems with designed controllers achieve two distinct characteristics (i.e., type 1 servo and type 2 sensitivity) within the low-frequency range using a single controller. Designing such H-infty controllers is difficult, and the details of the design method have not been fully clarified. Kawakami {it et al.} (2024) proposed a design method using an input weight function with integral characteristics. However, the dynamics of the plant model affect the output sensitivity characteristics achieved by the method therein. To address this problem, this paper proposes using the inverse dynamics of the nominal plant model in the integral weight function. The inverse dynamics reduce the effect of interfering dynamics on the achieved output sensitivity characteristics. The effectiveness of the proposed method is shown through a numerical example.