Keywords: Robotics, Machine Learning, Decision-making systems
Abstract: Autonomous safe navigation in unstructured and novel environments poses significant challenges, especially when environment information can only be provided through low-cost vision sensors. Although safe reactive approaches have been proposed to ensure robot safety in complex environments, many base their theory off the assumption that the robot has prior knowledge on obstacle locations and geometries. In this paper, we present a real-time, vision-based framework that constructs continuous, first-order differentiable Signed Distance Fields (SDFs) of unknown environments entirely online, without any pre-training, and is fully compatible with established SDF-based reactive controllers. To achieve robust performance under practical sensing conditions, our approach explicitly accounts for noise in affordable RGB-D cameras, refining the neural SDF representation online for smoother geometry and stable gradient estimates. We validate the proposed method in simulation and real-world experiments using a Fetch robot.

Keywords: Control Technologies, Robotics, Software Design
Abstract: Four-wheel independent drive and steering (4WIDS) robot possesses high maneuverability, with exploiting its stemming 8-DoF (Degrees of Freedom) control inputs, for navigation itself in complex, obstacle-rich environments. In such scenarios, Model Predictive Path Integral (MPPI) control, a sampling-based approach to model predictive control, has emerged as a powerful technique. While MPPI can effectively handle nonlinear dynamics and non-differentiable cost functions, it confronts the fundamental challenge of curse of dimensionality, where the number of required samples grows exponentially with the dimension of the control space. Although dimensionality reduction of the control space is a known countermeasure for suppressing the computational burden of MPPI, a systematic analysis for rationally designing the reduced space remains as an open issue. This research therefore addresses this issue by first proposing a novel metric, Maneuverability Ellipsoid of the 4WIDS, to quantify the robot’s maneuvering capability with regard to the multiple DoFs of the robot control inputs. Based on this ellipsoid, we numerically analyze a sampling method for selecting variables that contribute higher maneuverability in the MPPI framework. The robot maneuverability index is also proposed that is quantified by the size and shape of the manipulability ellipsoid. Through simulations, we demonstrate that this index significantly correlates with the success rate of robot navigation.

Keywords: Robotics, Control Technologies, Automation
Abstract: This paper presents an integrated navigation framework for Autonomous Mobile Robots (AMRs) that unifies environment representation, trajectory generation, and Model Predictive Control (MPC). The proposed approach incorporates a quadtree-based method to generate structured, axis-aligned collision-free regions from occupancy maps. These regions serve as both a basis for developing safe corridors and as linear constraints within the MPC formulation, enabling efficient and reliable navigation without requiring direct obstacle encoding. The complete pipeline combines safe-area extraction, connectivity graph construction, trajectory generation, and B-spline smoothing into one coherent system. Experimental results demonstrate consistent success and superior performance compared to baseline approaches across complex environments.

Keywords: Machine Learning, Rescue Systems, Automation
Abstract: Traditional post–earthquake inspections are slow, costly, and subject to human error. This paper presents an autonomous structural inspection system that combines a low–cost unmanned aerial vehicle (UAV) with a Proportional–Derivative (PD) flight controller, a convolutional neural network (CNN) for crack detection, and a gesture–based user interface for intuitive operation. Implemented on a DJI Tello platform, the system achieves 98 % validation accuracy on a dataset of 15,594 images while maintaining stable flight and executing predefined inspection trajectories via hand gestures. Results indicate the feasibility of integrating UAVs and deep learning to optimize post–seismic inspection workflows.

Keywords: Automation, Robotics, Hardware Design
Abstract: In recent years, grippers that can make high-speed contact and grip objects have been proposed utilizing the high back-drivability of the low reduction ratio geared motors or direct drive motors. However, with simple rotational back-drive phenomena alone, the impact force applied to the rotation axis when the fingertips come into contact with the environment cannot be avoided. Repeated impact forces applying on the actuator could lead to significant damage. Therefore, in this study, the center of percussion (CoP) theory is applied to the finger design of two-fingerd rotational opening-and-closing gripper, which has two highly-backdrivable actuators to independently drive them. This finger shape design theory can lead to the impact force mitigation. In this paper, the impact transmission ratio is considered and the results of experiments that demonstrate the validity and usefulness of the design theory based on the dynamics was presented. The finger design theory was modeled using Newton-Euler equations. As an initial step in the research, the collision experiments at a maximum speed of 1 m/s with a single finger attached to a robot arm were executed. As a result, the proposed CoP-based finger achieved an impact transmission ratio of approximately 0.2, indicating that about 80 % of the impact force was mitigated. Comparisons with other finger shapes further demonstrated the experimental validity and effectiveness of the proposed design.

Keywords: Human-robot Interaction / Collaboration, Software Design, Robotics
Abstract: This paper describes a method to reduce noise contained in signal obtained from a general microphone by using sensor fusion with a bone conduction microphone and a general microphone. A bone conduction microphone is a microphone that directly detects the vibrations in the throat when a person speaks. Hence, bone conduction microphones are less susceptible to external noise than general microphones. However, as bone conduction microphones have different acoustic characteristics than general microphones, the sound quality will deteriorate. Based on the above prospects, we are researching the sensor fusion technology with a bone conduction microphone and a general microphone. In this paper, we formulate the framework of the proposal approach and conduct some experiments to check the effectiveness of the proposal approach.

Keywords: Robotics, Software Design, Integration Platforms
Abstract: The Robot Operating System 2 (ROS 2) has become the middleware backbone of modern robotics and cyber-physical systems, offering flexibility, modularity, and high-performance communication via the DDS protocol and eProsima’s Fast-CDR serialization library. However, this re- liance on implicit type contracts between publishers and subscribers introduces critical attack surfaces. In this paper, we present the first systematic study of type confusion vul- nerabilities in ROS 2 deserialization, exposing a previously unexplored attack surface in robotic middleware. Through our fuzzing approach, we show that injecting malformed or mismatched message types into topics expecting a different format can trigger Fast-CDR deserialization failures. These failures propagate as uncaught exceptions resulting in process crashes and node-level outages.

Our findings reveal a previously undocumented flaw in ROS 2’s trust model for topic integrity, where the absence of runtime type enforcement or input validation leads to exploitable denial- of-service conditions. Through targeted fuzzing and case studies using standard ROS 2 messages, we evaluate the exploitability of this vulnerability in both simulation and physical robotics environments. This work underscores the need for secure-by- design messaging to ensure the reliability and safety of robotic middleware and cyber-physical systems.

Keywords: Robotics, Human-robot Interaction / Collaboration, Automation
Abstract: Safe physical human-robot interaction (pHRI) in industrial settings requires robust and accurate tracking of key points on the human co-worker's body and the robot structure. However, vision-based single-camera tracking solutions often face many challenges, such as limited field of view (FoV), detection range, occlusions, and inconsistent detection. This paper proposes a complementary multi-sensor tracking scheme that integrates RGB-D and near-infrared (NIR) cameras to improve human motion tracking accuracy while ensuring compliance with ISO/TS 15066 safety requirements. For the first time in pHRI, we deploy an infrared-based tracking system, originally designed for driver assistance and accident prevention, to complement RGB-D cameras, which provide detailed pose estimation at near range but suffer from a narrow FoV. A safety-oriented complementary approach is developed to fuse human tracking data from both systems into robot control, integrating a well-established safety paradigm based on the Safe Motion Unit (SMU) framework. The proposed system is experimentally validated in real-world collaborative robotic workspaces across various pHRI scenarios. Results demonstrate its effectiveness in respecting human safety constraints, even under challenging operating conditions, without unnecessary performance restrictions. The complementary vision-based approach improves tracking accuracy, expands FoV, and enhances reliability, making it a promising solution for certifiable, human-aware collaborative robotics in various industrial settings. The video documentation can be seen at https://youtu.be/xWksc_vhuew.

Keywords: Robotics
Abstract: The following contribution presents a fully additively manufactured inductive displacement sensor applied to a model of a translationally movable robot flange. In addition to demonstrating the feasibility of 3D-printed coils, the focus is particularly on their inductive properties and the variation of inductance as a function of translational displacement. For this purpose, two coils are entirely additively manufactured and integrated into the flange of a 3D-printed robot model. Through translational movement of the flange, the overlap between the two opposing coils changes, enabling position detection by measuring the series inductance. The advantage of such additively manufactured approaches lies in their adaptability to diverse application requirements and their capability to realize complex geometries.

Keywords: Human-robot Interaction / Collaboration, Decision-making systems, Robotics
Abstract: The object reference recognition task in human-robot interaction (HRI) consists of identifying the object to which a human is referring, based on communicative cues, including gaze and pointing, which is particularly challenging under ambiguous non-verbal behavior. This paper proposes a bio-inspired multimodal fusion algorithm to enable robots to recognize object references based on human gaze and pointing gestures. The proposed method integrates and encodes sensory inputs into a dynamic neural field, allowing the robot to adaptively resolve ambiguities in object referencing. The model was evaluated in an experimental setting where the participants interacted with a Furhat robot. The results showed that the system identified referenced objects with higher accuracy when both gaze and pointing cues were combined. Additionally, subjective evaluations using the Godspeed questionnaire indicated that participants perceived the robot more favorably when it engaged in joint attention behaviors. These results highlight the potential of dynamic neural models in improving intuitive and seamless HRI by addressing non-verbal ambiguity in shared workspaces. Future work will explore improved gaze-tracking techniques and closed-loop interaction models to enhance system robustness and adaptability.

Keywords: Human-robot Interaction / Collaboration, Assistive Robotics
Abstract: Interfaces for shared control of assistive mobility robots are often limited to either joysticks or wearable devices. While recent works have showcased the potential of wearables to promote physical activity, their setup can be cumbersome. This paper explores the potential of non-intrusive methods for controlling robotic wheelchairs, advancing the development of more user-friendly mobility solutions. Using pressure sensors embedded in the wheelchair seat and backrest, our objective is to assess whether a data-based approach can offer advantages over model-based controllers. Our baseline for the model-based controller is the state-of-the-art control methods based on the measured pressure distributions. We compare to this baseline the control performances achieved with data-based approaches. Such methods have the advantage to not require a calibration step. We collected a novel open-source dataset with six different drivers. The dataset, gathered using a commercial pressure mat, can be readily applied to the control of other robotic systems. We successfully demonstrate controllability without the need for wearables or other external systems, paving the way for a zero-shot approach. The dataset and sample code are available at: https://github.com/tchernin/posture-control.

Keywords: Robotics, Intelligent Transportation Systems, Machine Learning
Abstract: Smartphone-induced “texting while walking” poses growing safety risks not only in public shared spaces but also in robot navigation scenarios where humans and robots coexist. To mitigate these risks, recent studies have developed pedestrian behavior detection models that aim to recognize when people are distracted by their smartphones. However, these models still suffer from high false-positive rates and reduced detection accuracy when visually similar poses or occlusions occur. To address this issue, we propose a Vision-Language Model (VLM)-based behavior detector that exploits VLMs pretrained on large image-text datasets and capable of global-context inference. Specifically, we leverage LLaVA-7B and systematically evaluate three prompt-engineering schemes—chain-of-thought and self-consistency under zero-shot settings, and few-shot prompting under few-shot settings. We conducted the dataset generation experiment in a typical indoor hall with a centrally placed table that intermittently occluded the robot's view. During each session, four to six participants walked freely while performing nine everyday actions, resulting in 11,815 annotated pedestrian images captured from the robot's perspective. Experimental results show that our VLM-based pipeline significantly reduces false-positive detections and improves both precision and overall F1-score compared to a conventional pose-based LSTM baseline. These gains demonstrate that combining large-scale VLM reasoning with specially designed prompts can overcome long-standing misclassification issues in existing approaches. Our curated dataset and prompt-analysis results provide a foundation for extending VLM-based perception to a wide range of camera-based monitoring and navigation systems.

Keywords: Automation, Machine Learning, Robotics
Abstract: Stable operation of oil refineries is essential for ensuring a continuous supply of petroleum products, which underpin the foundation of modern society. However, various anomalies such as leaks can occur even during regular operation, making periodic inspection indispensable. To reduce the burden on human operators, automated visual inspection using mobile robots has been attracting increasing attention as a promising alternative to manual inspections. One promising approach, which we refer to as anomaly-aware change detection, is to compare videos captured during past and current inspections to identify scene changes specifically caused by anomalies. However, the robot cannot perfectly retrace its previous path, resulting in viewpoint misalignment between the two videos, which significantly degrades the performance of naive frame-wise comparison methods. To address this issue, we propose a novel inspection method that reconstructs a 3D model of the refinery from the past inspection video using Structure from Motion and 3D Gaussian Splatting, and then renders novel view images from the same viewpoints as those in the current inspection video. This allows us to obtain geometrically aligned image pairs, enabling anomaly-aware change detection that is robust to viewpoint misalignment. Field experiments conducted in an operational oil refinery achieved an F1-score of 0.806, significantly outperforming conventional methods and demonstrating the effectiveness of our approach.

Keywords: Robotics, Machine Learning, Automation
Abstract: In this paper, we propose an aircraft autopilot specifically designed for autonomous landing flights. The autopilot is trained through behavior cloning using a dataset of control command outputs provided by a human pilot in response to image inputs captured from the cockpit view. However, in unknown environments with different lighting conditions, even a trained autopilot struggles to determine appropriate control command outputs for visually different image inputs. To address this challenge and improve generalization capability across varying lighting conditions, we apply semantic segmentation to the original RGB images to classify runway pixels, and use the resulting segmentation images as inputs to the autopilot. Unlike RGB images, the segmentation images correctly classify only the runway regardless of lighting changes, producing visually consistent representations across all environments. Simulation experiments demonstrate that the proposed autopilot achieves improved generalization compared to the previous RGB-based autopilot, successfully landing on the runway in unknown evening and nighttime environments.

Keywords: Machine Learning, Automation, Decision-making systems
Abstract: This study proposes an automatic training data selection method for Autoencoder-based defect detection in hammering inspection, designed to address the severe class imbalance between normal and defective sounds. The proposed method employs a physically grounded indicator, Acoustic Energy per Impact, to automatically select and collect only the sound data considered normal. An Autoencoder is then trained exclusively on the collected normal sounds to identify defects based on reconstruction errors. To evaluate the effectiveness of the proposed method, experiments were conducted using concrete specimens with cracks. The results demonstrate that the proposed method achieves higher defect detection performance than a conventional approach, even under highly imbalanced class conditions.

Keywords: Robotics, Mechatronics Systems, Automation
Abstract: Industrial robots often perform handling and assembly tasks in industrial applications. This places high demands on dynamics and positioning accuracy. However, when handling low-weight workpieces, a relatively high robot mass usually leads to a limitation of the motion dynamics. The reason is that the manipulator's gears and drives are often arranged in series, as in the case of SCARA or articulated robots. This results in high moments of inertia for the actuated axes, located at the start of the kinematic chain. The overall stiffness is determined by the weakest element of the structure in that cases. Manipulators with parallel kinematic structures, such as delta robots, can provide a remedy here. However, when using manipulators of this type, problems arise when integrating them into production systems. They have an unfavourable ratio of workspace to installation space. Installation above the process area is mandatory and therefore strongly restricts the arrangement of other components in the production system. Hybrid structures, consisting of open and closed kinematic chains, offer alternative solutions. Based on such a hybrid approach, a concept of a novel kinematic structure for high dynamic applications is presented and analysed in this publication.

Keywords: Hardware Design, Robotics, Mechatronics Systems
Abstract: Controlling laser beam to be directed toward a tracking target with high accuracy is necessary in free-space optical communications and laser processing. If an optical center of a high-speed camera for tracking and a rotational center of a laser scanning system are aligned, three-dimensional calibration becomes unnecessary and wide-area laser tracking projection becomes easier, but it is difficult to precisely align their optical centers only manually. In this paper, we propose an interactive calibration method to precisely align the optical centers using a circular projection on screens with slits on one side, placed in different positions. Image processing visualizes the displacement of the circular parameters on each screen and enables highly accurate alignment in the translational direction. Evaluation experiments have demonstrated sub-pixel accuracy of calibration and fast and accurate laser tracking projection.

Keywords: Control Technologies, Mechatronics Systems, Machine Learning
Abstract: This paper presents a comprehensive framework for trajectory generation based on data-driven model predictive control (MPC) specifically tailored for industrial systems. The proposed method effectively combines a transfer function model, which is identified from operational data, with a neural network to address and mitigate modeling errors. In contrast to conventional online MPC approaches, the method formulated in this paper treats trajectory generation as an offline optimization problem, wherein the position sequence is directly optimized. Multiple performance metrics are jointly optimized, with the cost function weights being automatically tuned through machine learning-based optimization techniques, all while adhering to explicit tracking error constraints. Experimental validation conducted on a linear guide system demonstrates that the proposed method achieves both high tracking performance and smooth, practical trajectories, eliminating the need for manual parameter adjustments. This framework offers a robust and adaptable solution for advanced trajectory design applicable across a variety of industrial applications.

Keywords: Assistive Robotics, Human-robot Interaction / Collaboration, Entertainment and Educational Systems
Abstract: In this study, we developed a fingertip motion teaching suit using pneumatic artificial muscles (PAMs) to support the acquisition of optimal release timing in dart throwing. The proposed device is equipped with bidirectional bending-type PAMs placed near the metacarpophalangeal (MP) joint of the index finger and on the back of the hand. It provides haptic guidance for finger flexion and extension through the contractile force generated by pressurization. A control algorithm is proposed to pressurize the PAM at the pre-analyzed optimal release timing while accounting for actuator delay, so that the dart is released from the fingertip grip at the intended moment. By presenting the appropriate finger extension timing synchronized with the throwing motion, the system enables intuitive training and aims to improve dart-throwing skills. To validate the effectiveness of the device, dart-throwing experiments were conducted under three conditions—before, during, and after wearing the device—with healthy adult participants. The landing error and variability in release timing were evaluated. As a result, the release timing during device use became closer to the optimal reference timing compared to the pre-wear condition, and the average landing error also decreased. In the post-wear condition, the landing error was the smallest among the three conditions, and three out of four participants retained the improvement in release timing observed during device use. These findings suggest the presence of a retention effect and indicate that the proposed PAM-actuated device can function as an effective feedback tool for teaching motor timing.

Keywords: Rehabilitation Systems, Hardware Design, Human-robot Interaction / Collaboration
Abstract: Interpersonal tactile interaction plays a crucial role in emotional regulation and physiological responses, with affective touch demonstrating significant benefits in stress reduction and affective modulation. Existing approaches to replicating affective touch have relied on manual stroking or robotic brushing mechanisms, both of which present limitations in consistency, precision, and real-world applicability. This study introduces a Wearable Microblower System that utilizes airflow-based tactile stimulation to provide precisely controlled and reproducible affective touch stimuli. The system is designed to generate affective touch by producing stroking sensations at CT-optimal velocities and pressures, ensuring effective activation of CT afferents. A systematic performance evaluation confirmed that the device delivers tactile forces within the CT-optimal range, suggesting its potential feasibility for affective haptic applications. Moreover, the proposed system enables continuous and natural stroking through a control strategy that incorporates sequential activation of multiple stroking points. Our evaluation results indicate that the proposed device offers a quantifiable and reproducible means of delivering affective touch, with potential applications in stress alleviation, affective computing, and virtual reality–mediated haptic experiences. Future work will explore empirical validation through comparative studies and further integration with immersive technologies.

Keywords: Virtual / Augmented / Mixed reality, Software Design, Human Factors
Abstract: We present a VR-based pre-training system that estimates user-specific joint torques in real time by coupling a VR interaction environment (SIGVerse) with a biomechanics simulator (DhaibaWorks). The system visualizes internal load together with postural information using four encodings (color map, bar graph, exemplar posture, and exemplar+self posture) and enables rehearsal of lifting posture without handling real weight. In a within-subject study (11 participants), a simulated box-lifting task was evaluated using (i) time-integrated lumbar torque normalized by body mass and (ii) two 7-point Likert ratings (perceived comprehension and perceived load reduction). Across conditions, we did not observe a reliable reduction in normalized torque after training. Perceived load reduction differed across feedback modes, whereas perceived comprehension showed no clear between-condition differences. These findings indicate that visualizing internal load can influence users’ perception of effort, although the present short session did not yield measurable torque changes. The proposed platform provides a safe pre-training route for learning low-strain movement strategies and a foundation for adaptive human–agent interaction that can leverage real-time estimates of human physical state.

Keywords: Virtual / Augmented / Mixed reality, Human-robot Interaction / Collaboration, Human Factors
Abstract: Using a head-mounted displays (HMDs) and an acceleration device to present the sensation of ascent in a virtual reality (VR) environment offers various advantages, such as enhancing immersion and aiding in the treatment of acrophobia. However, existing studies have primarily used large-scale ascent sensation presentation devices, as smaller devices face limitations in the achievable ascent distance in VR. In this study, a shoe-type acceleration device developed in previous research was utilized to examine the presentation of long-distance ascent sensations. This study demonstrated that combining the device with VR visuals was more effective in presenting ascent sensations compared with using VR visuals alone. Furthermore, in environments where the operational range of devices, such as shoe-type devices, was limited, increasing the ascending speed was found to be desirable. However, discrepancies between the VR visuals and operation time of the device were identified as a challenge in presenting long-distance ascent sensations. Additionally, although increasing the initial velocity of the VR visuals enhanced the perception of ascent, a better ascent experience required more than simply synchronizing the movement of the device with the VR visuals. The appropriate handling of the VR visuals after the operation of the device ended was crucial. The findings of this study clarify key challenges for future research on presenting ascent sensations using small wearable devices.

Keywords: Robotics
Abstract: This paper studies state estimation for continuum robotic fingers during in-hand manipulation, where accurate pose estimation relative to the environment is required in feature-sparse scenes. To address this requirement, we adopt a SLAM-based formulation that estimates the robot pose and a local map from exteroceptive sensing. Continuum fingers lack encoder-based joint angle measurements, while conventional SLAM assumes feature-rich environments that are rarely available inside the hand. We propose a SLAM-based estimator that fuses exteroceptive proximity sensing with a constant-curvature kinematic prior by replacing encoder angles with virtual joint angles from the model. The key idea is to leverage designed in-hand self-body elements, namely the opposing fingers and the palm, as stable reference geometry to maintain observability in feature-spares environments. We evaluate our method through free motion and grasping simulations, and analyze the effect of presence and shape of the palm on estimation accuracy. The proposed estimator outperforms a kinematics-only baseline by suppressing bias, reducing a position error of an end effector, and improving map quality. We demonstrate that three-dimensional contoured palms enhance observability, and a composite wavy palm yields the smallest errors without temporal drift. These results indicate that designed in-hand geometry enables effective state estimation for continuum fingers in feature-sparse grasping scenarios, supporting reliable in-hand manipulation.

Keywords: Robotics, Hardware Design, Automation
Abstract: Electric prosthetic hands should be lightweight to decrease the burden on the user, shaped like human hands for cosmetic purposes, and designed with motors enclosed inside to protect them from damage and dirt. Additionally, in-hand manipulation is necessary to perform daily activities such as transitioning between different postures, particularly through rotational movements, such as reorienting a pen into a writing posture after picking it up from a desk. We previously developed PLEXUS hand (Precision–Lateral dEXteroUS manipulation hand), a lightweight (311 g) prosthetic hand driven by four motors. This prosthetic performed reorientation between precision and lateral grasps with various objects. However, its controller required predefined object widths and was limited to handling lightweight objects (of weight up to 34 g). This study addresses these limitations by employing motor current feedback. Combined with the hand’s previously optimized single-axis thumb, this approach achieves more stable manipulation by estimating the object’s width and adjusting the index finger position to maintain stable object holding during the reorientation. Experimental validation using primitive objects of various widths (5–30 mm) and shapes (cylinders and prisms) resulted in a 100% success rate with lightweight objects and maintained a high success rate (≧80%) even with heavy aluminum prisms (of weight up to 289 g). By contrast, the performance without index finger coordination dropped to just 40% on the heaviest 289 g prism. The hand also successfully executed several daily tasks, including closing bottle caps and orienting a pen for writing.

Keywords: Robotics, Hardware Design, Human-robot Interaction / Collaboration
Abstract: In this paper, we present a soft robotic hand structure capable of enhancing object grasp retention force and achieving stable object grasping under teleoperation without haptic feedback. The soft robotic hand consists of a hand skeleton equipped with a soft glove device actuated by extra-thin McKibben artificial muscles. In this paper, a particle layer was integrated into the hand skeleton, and the jamming transition was implemented by applying a vacuum to this particle layer. The jamming transition increases the driving force required to actuate the fingers of the hand skeleton, thereby improving the object grasp retention force of the soft robotic hand. Finally, as a demonstration, teleoperation was performed using leader–follower control without haptic feedback. The results show that activating the jamming transition after grasping an object enables the hand to maintain a stable grasp regardless of the leader’s posture.

Keywords: Robotics, Hardware Design, Control Technologies
Abstract: 遠隔操作システムは、 人間の労働者に危険をもたらす環境および 完全自律型ロボットが重大な 課題。 このようなシステムに組み込まれたロボットハンドは、 人間のオペレーターに匹敵する能力を有する 幅広いタスクを遂行するために。

この研究では、 に統合するための腱駆動ロボットハンド 遠隔操作システム。 ターゲットタスクには、器用な操作と 通常、人間が行う力を多用する操作。 これらの機能を実現するために、まず タスクの実行に必要なパフォーマンス要件と グリップを最大化するリンク比を解析的に導き出した 効率。 これらから、さらにリン

Keywords: Rehabilitation Systems, Machine Learning, Medical Training
Abstract: The functional independence measure (FIM) is widely used to evaluate patients' physical independence in activities of daily living. However, traditional FIM assessment imposes a significant burden on both patients and healthcare professionals. To address this challenge, we propose an automated FIM score estimation method that utilizes simple exercises different from the designated FIM assessment actions. Our approach employs a deep neural network architecture integrating a spatial-temporal graph convolutional network (ST-GCN), bidirectional long short-term memory (BiLSTM), and an attention mechanism to estimate FIM motor item scores. The model effectively captures long-term temporal dependencies and identifies key body-joint contributions through learned attention weights. We evaluated our method in a study of 277 rehabilitation patients, focusing on FIM transfer and locomotion items. Our approach successfully distinguishes between completely independent patients and those requiring assistance, achieving balanced accuracies of 70.09--78.79% across different FIM items. Additionally, our analysis reveals specific movement patterns that serve as reliable predictors for particular FIM evaluation items.

Keywords: Control Technologies, Machine Learning, Robotics
Abstract: This work proposes a novel framework to accelerate motion planning in previously unseen environments with obstacles by modeling the distribution of the collision-free configuration space using Wasserstein Generative Adversarial Networks with Gradient Penalty (WGAN-GP). To effectively manage multimodal data, we condition the WGAN-GP using a Variational Autoencoder (VAE) embedded in a continuous latent space. The configuration space is approximated through a set of Gaussian distributions, allowing the dataset to be segmented into multiple localized models. This strategy not only enhances the learning efficiency but also reduces convergence time. We utilize the reconstructed configuration space to evaluate motion planning performance in previously unseen scenarios. Experimental results highlight the potential of our approach to significantly accelerate planning in unknown environments while maintaining the generation of near-optimal trajectories.

Keywords: Machine Learning, Robotics
Abstract: Sparse rewards are a persistent bottleneck for robotic manipulation with Reinforcement Learning (RL), primarily because RL agents must discover long-horizon, multi-step behaviors while receiving infrequent and weakly informative feedback. Recent work uses pre-trained Vision Language Models (VLMs) to provide dense per-step rewards, yet most approaches score only a single image against a goal text, ignoring whether the recent change actually moves the system toward success. We argue that this omission impairs exploration (e.g., goal-like detours, wrong-way progress, action aliasing) and propose to make time explicit in VLM rewards by adding a directional signal that evaluates short-horizon change. Concretely, we pair visual change over a few steps with a text description of the desired change, and finetune lightweight heads with RL; the resulting directional signal is combined with a standard positional signal into a single shaping reward. We evaluated our approach in six MetaWorld manipulation tasks with fixed goals. This directional shaping improves running average success at a fixed budget to 78.2%, versus 63.8% for the best-tuned positional baseline (improvements were observed in five of six tasks). Ablations identify key design choices for the proposed directional term to be effective and show its synergy with the positional term when supplying dense VLM rewards, demonstrating improved exploration and sample efficiency.

Keywords: Automation, Mechatronics Systems, Control Technologies
Abstract: In robotic visual servoing, when prior measurement or imaging is impractical, control must rely on a single image of the initial view. Diffusion models can generate 3D shapes from a single image; however, the blind spot area involves uncertainty, which may degrade alignment accuracy if used directly in model-dependent control manipulation. This study proposes a hybrid visual servo control method that operates under the assumption of this uncertainty. From a single image, multiple 3D shape candidates are sampled using a pre-trained generative model, and then progressively controlled while retaining them. First, rough positioning is performed using PBVS (Position-Based Visual Servoing) with multiple shape candidates. Next, the system compares the candidate images rendered with the observed image and selects the best model based on geometric error and visual similarity. Finally, IBVS (Image-Based Visual Servoing) uses the selected model to refine slight alignment errors with high precision. This proposed method achieves high-precision approach and alignment from minimal input of a single image, providing a framework that resolves the problems of shape uncertainty and control error caused by 3D generation. Experiments show that the convergence success rate improved as the number of shape candidates increased and that high-precision alignment was achieved through the staged integration of PBVS and IBVS.

Keywords: Integration Platforms, Automation, Robotics
Abstract: Autonomous task-oriented robots are increasingly in demand across various domains; however, few existing systems address the challenge of autonomous high-resolution terrestrial scanning for construction and inspection purposes. This paper presents a task-oriented autonomy framework integrated with the Spot quadruped robot, enabling autonomous exploration, mapping, and deployment of a FARO terrestrial laser scanner. We introduce two novel algorithms for selecting optimal scanning positions: SCANSAFE (Scanpoint Navigator using Spatially-Aware Filtering and Evaluation), which prioritizes coverage of open space relative to prior scans, and PATHSAFE – Path-Aligned Trajectory Heuristic for Scanpoint Allocation with Filtering and Evaluation method, which places scan points along the robot’s traveled path. These approaches are evaluated against two existing strategies: Next-Best-View Greedy (NBV-Greedy) and Frontier, as well as a manually guided baseline. Tested in multiple environments, the proposed algorithms successfully identified valid scanning points. On average, the SCANSAFE method generated 23.4% fewer scan points than NBV-Greedy, 44.4% fewer than Frontier, and 2.0% more than the manual baseline. The PATHSAFE method showed average reductions of 32.8% compared to NBV-Greedy, 51.6% compared to Frontier, and 10.4% compared to the manual approach. Both methods improved efficiency, reduced operational overhead, and increased safety in hazardous or constrained environments.

Keywords: Robotics, Hardware Design, Mechatronics Systems
Abstract: Mobile robots capable of operating in environments where human work is difficult, such as elevated locations or rough terrain, have been developed, including rough-terrain mobile robots capable of three-dimensional locomotion by fixing end of a tether to the environment and winching it in. However, such robots still face issues, for example, automation of tether end fixation and release. To address these issues, we consider elastic telescopic arm (ETA) as an effective solution. In this paper, we propose a robot capable of three-dimensional movement using multiple ETAs. The proposed robot extends several ETAs mounted on its body, grips rigid environmental structures with grippers at their tips, and moves the body to arbitrary positions in three-dimensional space by controlling extension and contraction of the arms. To verify the feasibility of this principle, we conducted contraction experiments on the arm while under tensile load. Furthermore, to demonstrate the usefulness of the ETA in rough-terrain traversal, we performed a proof-of-concept experiment using a prototype equipped with a single ETA mounted on a four-wheeled mobile robot. The experiment confirmed that the use of the ETA enabled the robot to climb stairs that it could not climb on its own.

Keywords: Renewable and sustainable energy, Environment / Ecological Systems, Integration Platforms
Abstract: Hydroelectric power plants constitute the primary source of electricity generation in Guatemala, representing the largest share of the national energy mix, making the inspection and maintenance of infrastructure such as large water inlet pipes (penstocks) crucial. These pipes, which are several kilometers long and have limited access, present hazards for human inspectors. This work introduces a low-cost, modular robotic rover designed for manual inspection of penstocks of varying sizes and conditions. The robot features a tracked locomotion system with an adjustable mechanism for stable alignment and is operated via a web-based joystick interface, enabling real-time anomaly detection. It supports mapping, localization, and data logging for post-inspection analysis. The modular design allows for disassembly and transport through tight spaces, followed by quick reassembly inside the pipe. The rover has successfully conducted multiple inspections in diverse hydroelectric pipelines of different diameters and materials, under diverse sediment accumulation conditions.

Keywords: Robotics, Control Technologies, Integration Platforms
Abstract: Aerial manipulators offer a compelling solution for maintenance tasks in hazardous or hard-to-reach environments. Contact-based operations, such as pipeline crack repair, demand not only precise trajectory tracking but also stable and controlled force application when interacting with cylindrical surfaces subject to friction. These dual objectives are challenging due to the aerial system’s inherent instability, strong nonlinearities, and sensitivity to disturbances. While impedance control enables direct interaction, it replaces the position control loop that can reduce robustness under dynamic conditions. In contrast, conventional admittance control wraps around existing loops but suffers from limited adaptability and challenging gain tuning. This paper proposes a unified force-compliant control framework for a quadrotor equipped with a hybrid manipulator. An adaptive backstepping–adaptive fast terminal sliding mode controller (AB–AFTSMC) governs the inner trajectory loop, ensuring reliable tracking performance. Over this structure, a fuzzy-admittance outer loop modulates the desired reference trajectories based on desired force, enabling compliant interaction during contact without altering the underlying control architecture. Interaction forces are estimated using a lightweight disturbance observer (DOB), enabling low-computation, sensorless feedback without the weight or complexity of force sensors. Validation is carried out in MATLAB Simscape through a 3D physics-based model. The results demonstrate reliable tracking, compliant force build-up, and consistent free-flight-to-contact transitions.

Keywords: Integration Platforms, Decision-making systems
Abstract: In the automotive industry, as market demands continue to grow, the complexity of vehicle systems has been increasing accordingly. Vehicle manufacturers are thus confronted with the challenge of offering more design options. Model-Based Systems Engineering (MBSE) has been introduced to manage the complexity of product development. Meanwhile, its integration with Multidisciplinary Design Analysis and Optimization (MDAO) provides a harmonized methodology for the design and analysis of complex systems, offering a system-level virtual integration environment to support decision-making processes. However, current MBSE modeling languages such as SysML v1.x are extensions of the graphical modeling language UML and are considered semi-formal. Their limited capability to extend and integrate with modeling languages across different disciplines significantly restricts their effectiveness. SysML v2, built upon the formal modeling language KerML and supporting standardized APIs, enhances model interoperability. By utilizing SysML v2 to develop a domain-specific modeling language tailored for multidisciplinary engineering, it becomes possible to effectively maintain data consistency between system design and analysis processes, establish a virtual integration platform, and ensure interoperability with other environments. This paper presents a SysML v2-based multidisciplinary system virtual integration framework, implemented by constructing a coupled multidisciplinary problem model in the context of automotive systems.

Keywords: Integration Platforms, Machine Learning, Renewable and sustainable energy
Abstract: This paper presents the design and experimental validation of a modular intelligent platform for energy-efficient lighting and monitoring in controlled environment agriculture (CEA). The proposed architecture integrates dimmable LED lighting with neural network–based spectral optimization, machine learning–based light intensity estimation, image-based plant monitoring, and IoT-enabled data acquisition. The platform emphasizes a unified control framework that bridges perception, control, and application layers, enabling adaptive real-time decision-making and modular scalability. The system was implemented and tested in a greenhouse environment, demonstrating a 28% reduction in energy consumption per gram of dry biomass while maintaining plant health and productivity. The results underscore the effectiveness of the proposed architecture in advancing intelligent control and automation strategies for sustainable horticultural systems.

Keywords: Integration Platforms, Machine Learning, Robotics
Abstract: Recently, AI-based action models are expected to be applied to the robot systems. This study proposes the robot AI ecosystem for AI-based action models. This extends the platform sharing action models to the model circulation system through sharing demonstration data and model building. Specifically, we will focus on building advanced models through model circulation among the demonstration data of the robot behaviors and model training by the model builders in the ecosystem. Experiments in two scenarios to build the new generation action models through model circulation were performed to demonstrate the advancement of the action models. The results of experiments confirmed that open collaboration among the robots and the model builders in the ecosystem is effective for creating the advanced action models.

Keywords: Integration Platforms, Machine Learning, Virtual / Augmented / Mixed reality
Abstract: In this work, we propose a bedside monitoring system for ICU patients that integrates patient data from different medical devices and collect information about physical activity of the patient. We investigate how extra information can be utilized to provide more robust, context-aware and personalized patient monitoring and early warning with the integrated data for inhospital deterioration. Furthermore, we explore visualization techniques to present analysis results to clinical caregivers in a more intuitive and interpretable way.

Keywords: Machine Learning, Automation, Entertainment and Educational Systems
Abstract: An automatic method for generating sewing patterns corresponding to dress images is proposed in this study. In garment production, the creation of sewing patterns, the blueprints for garment construction, from design sketches is a highly complex process that demands substantial expertise and experience. Most existing studies focus on learning from entire garments; however, they face the challenge of reduced shape reproduction accuracy for small parts with diverse shapes, such as collars and sleeves. The proposed method segments a garment image into three main parts—bodice, sleeve, and collar—and inputs each part into a specialized sewing pattern generation model, enabling faithful reproduction of even small and complex garment parts. A custom training dataset consisting of garment images and their corresponding sewing pattern images used in actual garment production is constructed. In addition, a part segmentation model and part-specific GAN based sewing pattern generation models are developed. The proposed method is capable of adapting to diverse garment shapes and variations across parts, thereby enhancing both the accuracy and efficiency of sewing pattern creation in garment production workflows.

Keywords: Robotics, Machine Learning, Automation
Abstract: The construction industry faces severe labor shortages, driving the need for robotic automation solutions. However, effective deployment of construction robots requires robust environmental perception capabilities, particularly accurate identification of diverse objects in complex, dynamic construction environments. Closed-set object detection methods are limited to predefined categories, proving inadequate for the highly varied object types encountered on construction sites. This paper introduces EVLOD (Ensemble Vision-Language Open-vocabulary Detection), an ensemble framework that integrates multiple state-of-the-art vision-language models to enable open-vocabulary object detection in construction scenarios. EVLOD employs a voting-based fusion strategy that combines predictions from GroundingDINO and DINO-CLIP detectors, utilizing their complementary strengths while mitigating individual model weaknesses. The ensemble approach incorporates confidence voting, object name voting, and bounding box voting to produce reliable detections with reduced false positives. Evaluated on a comprehensive dataset of 825 Unmanned Aerial Vehicle (UAV)-captured construction images with 5,020 annotated objects, EVLOD achieves an Average Precision (AP) of 0.49 when Intersection over Union (IoU) equals 0.5, representing a 36.1% improvement over the best-performing baseline. The method effectively reduces detection noise from 5,495 to 3,232 detections.Qualitative analysis reveals primary limitations in detecting small-scale objects and low-contrast elements.

Keywords: Machine Learning, Environment / Ecological Systems, Robotics
Abstract: This study proposes a method for accurately estimating the mixing ratio of V.harveyi and soil bacteria by analyzing motility in inverted microscope videos and extracting 24 features. Using an XGBoost model, the proposed method outperformed SVM and 1D-CNN approaches. The proposed analytical method will be implemented into a MD-based screening system integrating an inverted microscope, automated stage, and robotic micromanipulator to enable real-time automated classification, selection, and retrieval of antagonistic bacteria during microscopic observation.

Keywords: Robotics, Integration Platforms, Software Design
Abstract: Inspecting sewers represents a significant challenge as these environments pose considerable safety risks to human operators. In this view, drones capable of autonomous flight can be used to perform inspection tasks reducing human exposure. However, sewer environments are typically confined, featureless, and poorly lit, hence, standard algorithms for the localization in GNSS-denied environments, such as Visual-Inertial Odometry (VIO), often fail. In addition, drone localization is further complicated by rotor-induced turbulence, and vibrations, that affect sensor measurements. This paper presents a low-cost multisensor-based method for robust pose reconstruction of Unmanned Aerial Vehicles (UAVs) to enable reliable navigation in visually degraded, GPS-denied environments. The proposed framework leverages environmental geometry, specifically obstacle and wall distances, to estimate relative motion and correct drift via a speed control strategy that maximizes the distance from any obstacle. The approach is validated through both simulation and real-world experiments, demonstrating its effectiveness in representative scenarios.

Keywords: Robotics, Software Design
Abstract: This paper presents a system architecture for the robotic manipulation and reshaping of 3D deformable objects. The inherent complexity of manipulating objects with evolving shapes requires the construction and efficient integration of components that are reusable and remain functional under operational scenarios that involve high variability. Such original components are developed and integrated in this paper, notably including a mapping subsystem to aggregate and analyse sensor data for extracting general manipulation and sensing heuristics, as well as a deformation modelling component to predict the effect of candidate robot actions on the shape of an object of interest. A simulation interface is further implemented on top of robotic simulators, allowing a modular creation of environments to evaluate the performance of the aforementioned components and to generate synthetic datasets of 3D deformable object manipulation tasks. These general-use components are integrated with application-specific task and motion planning components in real and simulated environments.

Keywords: Machine Learning, Robotics
Abstract: Because of labor shortages, robots are expected to provide work assistance in a variety of settings, including the home environment. At home we often deal with flexible objects, but flexible objects are characterized by their tendency to change position and shape. Because of this nature, data dealing with flexible objects involves uncertainty. Although deep learning has been used to perform a variety of complex tasks, the deterministic nature of conventional RNN makes it difficult to handle data with a probabilistic structure. In this study, we propose a method based on deep predictive learning that enables real-time motion generation and predicts the variance of joint angles, which facilitates learning of probabilistic structures and can handle dynamic changes. Experimental results show that the robot is able to generate motions that are adaptive to flexible objects with dynamic position changes.

Keywords: Control Technologies, Robotics
Abstract: Grasping soft and individualized food and agricultural products without causing damage is a significant challenge in robotics. This task requires balancing two conflicting demands: applying sufficient force to lift the object and avoiding excessive force that could cause damage. Conventional approaches include learning-based manipulation and sequential control based on slip detection. However, the former requires prior training, while the latter takes time to adjust the grasping force. Therefore, these methods are not suitable for environments where object properties change frequently or for high-throughput operations. To address these issues, we propose a parameter adaptation method for deformation suppression that does not require learning and enables high-speed processing. The proposed method reduces the grasping force according to object deformation, which is detected by optical proximity sensors. A key benefit of optical proximity sensors is high-speed data acquisition, which enables real-time adjustment of grasping force without stopping the motion, leading to faster task completion. Furthermore, the deformation information obtained from the proximity sensor is converted into a virtual force, and the grasping force is adjusted based on the virtual dynamics framework. This enables seamless integration with pre-grasp control strategies that gently approach the object.

Keywords: Human-robot Interaction / Collaboration, Medical Training, Entertainment and Educational Systems
Abstract: This study proposes a foundational method for objectively classifying whether mothers can easily perceive fetal movements, a crucial indicator in perinatal care. Traditional fetal movement assessments rely on subjective maternal reports or wearable sensor-based methods susceptible to external noise.

This paper applies the previously established Multi-Resolution Feature (MRF) method to independently quantify both fetal and maternal deformation from simulated fetal movement videos. We constructed a novel fetal-maternal deformation map using these deformation values as axes. The results demonstrate that this map can visually represent the relationship between fetal and maternal deformation. Furthermore, exploratory analysis of the data distribution on the map suggested the existence of distinct fetal movement patterns: movements accompanied by significant maternal abdominal wall deformation (likely perceived by the mother) and movements where the fetus is active but less transmission occurs to the mother (likely difficult for the mother to perceive). This method is expected to be a stepping stone towards establishing a new approach for more detailed assessment of fetal health and, consequently, improving the quality of perinatal medical care.

Keywords: Control Technologies, Robotics, Hardware Design
Abstract: This paper proposes an articulated-arm underwater robot with a suction mechanism, which includes thrusters for both suction and propulsion. A robotic arm encounters fluid resistance during manipulation. However, generating a suction force using a propulsion thrusters enable stable manipulation. This suction force depends on the arm’s movement. In this research, the suction force required to attach the wall was estimated based on a dynamic model that includes fluid resistance. The experiment revealed the relationship between the input value of the electric speed controller (ESC) for thrusters and the suction force. Furthermore, the experiment showed that the developed robot can maintain suction to a pool wall with this estimated suction force. Results confirmed that the proposed dynamic model can estimate the required suction force for 10 out of 12 arm-end trajectories.

Keywords: Control Technologies, Robotics, Software Design
Abstract: Biomimetic autonomous underwater vehicle (BAUV) with multi-link mechanism is widely used in aquatic life observation and environmental surveys due to its low power consumption and high maneuverability. An environmental survey requires a path following system that automatically follows specific points. However, the path following system of BAUV is limited, and its evaluation with multi-link mechanism robots has not yet been clarified. The path following system in BAUV requires prior simulation because the model differs depending on the type of biomimetics. In this study, we propose a path following system for BAUVs with a multi-link mechanism and evaluation in underwater simulation. In this result, it was possible to design a path following system suitable for BAUV, determine parameters using a simulator, and evaluate control methods.

Keywords: Hardware Design, Control Technologies, Mechatronics Systems
Abstract: Water hydraulics offer low viscosity, high-pressure capability, and environmental compatibility, making them ideal for teleoperated systems in extreme environments. However, achieving stable low-flow control and reliable sealing remains a key challenge owing to the complexity of current valve designs, and their limited precision at low-flow rates. To address this issue, we developed a compact needle valve—called Series Elastic Actuated Needle Valve (SEANV)—which incorporates a series elastic element, along with a sealing equilibrium model to predict its sealing behavior. Preliminary experiments using a single-acting actuator confirmed improved position tracking; however, these did not quantitatively characterize the flow control performance. Therefore, we conducted a detailed experimental evaluation of the flow and sealing characteristics of SEANV under various single-acting actuator conditions. The results demonstrate the enhanced low-flow resolution, achieving an RMSE of 0.516 × 10^3 mm^3/s, and robust sealing performance, with leakage maintained within ±50 mm^3/s, thereby validating the applicability of the proposed model to compact water-hydraulic systems.

Keywords: Robotics, Integration Platforms, Hardware Design
Abstract: In recent years, with the rapid advancement of space development, there has been increasing demand for constructing solar power generation systems on the Moon using modular robots. This study has been proposing a modular construction method aimed at building tall structures on the lunar surface. Their modules adopt a truncated octahedron shape one of a space-filling polyhedron allowing their configuration and placement to be determined geometrically by exploiting its space-filling property. This approach can also be treated as a space quantization and establishes a unified platform in which module development, configuration, and positioning can be derived with comparatively simple calculations. In this paper, we present the results of a field experiment using newly prototyped modules, along with the previously proposed construction method and module design. The results demonstrate the feasibility of the proposed method and highlight the potential for space quantization for robot motions based on space-filling properties.

Keywords: Decision-making systems, Robotics, Control Technologies
Abstract: This paper aims to achieve flexible placement onto tray-type storage areas using a mobile manipulator. In a previous study, tray-packing was accomplished by preparing object masks in advance before tidy-up and planning the arrangement by matching these masks with the available tray space. However, such an approach faced difficulties in handling unknown objects and postures for which prior information could not be obtained, and was unable to take the current state of the tray into account, resulting in limited flexibility. The proposed system estimates masks for previously unregistered objects from an RGB-D image, verifies and corrects object posture to a default, builds a tray mask from the current state of the tray, and computes collision-aware, space-efficient placements via a 2D irregular packing algorithm. The proposed system performed a tidy-up on five objects placed in various postures, achieving a success rate of 93%. These results indicate improved flexibility, robustness, and practicality for real-world tidy-up compared to the previous system.

Keywords: Integration Platforms, Robotics, Software Design
Abstract: Indoor robot navigation in unfamiliar environments requires accurate mapping, contextual understanding, and semantic reasoning to interpret user intent. While ROS 2 navigation frameworks provide strong 3D SLAM capabilities, they lack an integrated, modular approach that unifies perception, semantic representation, and high-level reasoning in a single, real-time system. We present Navigation AI Impulse by Scene Graph (NAIS), the first ROS 2 framework to continuously fuse SLAM, dynamic open-vocabulary scene graph construction, and foundation-model reasoning into a deployable and modular pipeline for mobile robots. NAIS operates online, using VLMs to describe objects and LLMs to infer spatial relations and resolve ambiguous natural language requests. This unified design links environment understanding directly to navigation planning, enabling context-aware goal selection without manual configuration or static class lists. Demonstrations on a TurtleBot 4 in office environments show NAIS handling high-level, context-dependent commands such as “I’m thirsty” or “I’m tired”, illustrating its potential for robust, adaptable indoor navigation.

Keywords: Robotics, Automation, Software Design
Abstract: Small, resource-constrained swarm robots require scene understanding that is both semantic and metric, yet most SLAM pipelines either ignore semantics or demand heavy sensors. We propose an online hybrid factor-graph optimisation (FGO) framework that jointly estimates continuous robot poses and discrete terrain labels using only low cost wheel encoder and IMU data. Continuous and discrete variables are modelled as nodes in a single factor graph; maximum-a-posteriori inference is carried out by an alternating optimisation scheme executed inside a fixed-size sliding window, allowing constant time updates on embedded hardware. The method closes three longstanding gaps: (1) a unified probabilistic formulation for hybrid state estimation, (2) an online solver that scales with mission duration, and (3) automatic construction of a dynamic semantic topological map that captures both spatial layout and label transitions. The resulting graph supports high level navigation and situational awareness without external infrastructure. We validate the approach in a 2D simulation comprising six terrain regions, random walks of 150 steps, and realistic odometry and classification noise. These results demonstrate that hybrid FGO can endow minimalist robots with robust, semantics-aware mapping capabilities, paving the way for long-duration exploration and cooperative task planning in GPS-denied, sensor-limited environments.

Keywords: Intelligent Transportation Systems, Decision-making systems, Robotics
Abstract: Autonomous vehicle (AV) planners must satisfy complex, often conflicting, safety constraints, traffic laws, and comfort norms. Conventional methods like formal logics and optimal control may fail under rule conflicts, while learning-based policies lack the necessary formal guarantees for certification. This paper introduces a unified rulebook framework that encodes heterogeneous driving rules as differentiable violation metrics structured by total order over equivalence classes (TORQ). This removes rule incomparability and enables lexicographical optimization of trajectories. The specification integrates into real-time control using robust High-Order Control Barrier Functions (HOCBFs) and Control Lyapunov Functions (CLFs) solved via Sequential Quadratic Programming (SQP). A recursive relaxation algorithm maintains the hierarchy of the rules, allowing violations of only the lowest-priority rules necessary to resolve conflicts. Extensive simulations, including urban intersections and lane drift scenarios on roads, demonstrate that the system consistently prioritizes high-level safety mandates. By combining formal specification, real-time synthesis, and verification, this framework offers a robust, certifiable, and transparent approach to AV behavior planning.

Keywords: Robotics, Integration Platforms, Software Design
Abstract: This paper presents a modular manipulation pipeline for CALL-M, a mobile robot developed at CNRS-AIST JRL for semi-autonomous pick-and-place operations in convenience store environments. The system leverages a ROS 2-based architecture integrating 3D perception, grasp detection, and motion planning using Moveit 2. The pipeline comprises modular stages-point cloud acquisition, object selection, grasp estimation and trajectory generation-coordinated by a centralized task manager. Validation in both simulation and real-world scenarios demonstrated successful grasps. While simulation confirmed reliability under ideal conditions, real-world trials revealed challenges due to sensor noise, workspace constraints, and misalignments in grasp pose generation. Despite this, the system's modularity and adaptability make it a scalable solution for manipulation in semi-structured environments.

Keywords: Robotics, Integration Platforms, Decision-making systems
Abstract: In recent years, general-purpose robots have been introduced into domains that require delicate manipulation, such as materials synthesis experiments. Although advances have been made in automating specific processes, generalized robotic pick-and-place operations still pose a challenge due to the diversity of target objects and the need for appropriate force control. This study proposes a method for the zero-shot estimation of the target grasping force that utilizes knowledge about human motions embedded in a vision-language model (VLM). Robot manipulation images are converted into human action images using a style transfer approach based on a fine-tuned variational autoencoder, enabling the VLM to better infer grasping force requirements. The VLM, specifically GPT-4o, is prompted to estimate the target grasping force in discrete categories (i.e., no grasp, light grip, and firm grip). The experimental results show that converting robot images into human representations improves the accuracy of grasping timing estimation. Also, using human images improves target object recognition accuracy. Furthermore, the inclusion of target object information in the prompt improves estimation accuracy across all input image types. These findings highlight the effectiveness of utilizing a human-knowledge-trained VLM for robotic force control and open a new direction for general-purpose, cost-efficient manipulation without relying on large-scale robot force datasets.

Keywords: Robotics, Automation
Abstract: We address the problem of trajectory optimization in scenarios where multiple grasp targets are available for the same object. We introduce SwitchOpt, an adaptive optimization strategy that dynamically switches between grasp targets during trajectory optimization. Instead of committing to a single candidate, SwitchOpt monitors progress step by step using a merit function that captures trajectory quality and constraint satisfaction. A prediction horizon is used to assess whether the current trajectory is likely to improve further, while a minimum-stay mechanism ensures sufficient refinement before considering a switch. Whenever a switch is considered, SwitchOpt reconstructs candidate trajectories by combining the current head with interpolated tails toward alternative grasp targets, and evaluates each of these full trajectories with the same merit function over the prediction horizon. If the best candidate is predicted to outperform continuing toward the current target, that target is selected as the new goal and its reconstructed trajectory is used as the new initialization, allowing the solver to continue from a promising adapted trajectory. This principled selection strategy balances local exploitation of the current target with structured exploration of alternative grasp poses, maintaining optimization continuity between switches. Experiments in simulation demonstrate that SwitchOpt improves final trajectory quality, accelerates convergence, and increases feasibility in multi-target trajectory optimization.

Keywords: Machine Learning, Automation, Software Design
Abstract: For industrial automation, robots have to robustly pick objects with a diverse range of physical properties, such as shape, weight and surface optical properties. To realize such a purpose, this research proposes a method for depth estimation of transparent objects having complex optical properties, such as refraction and reflection from a single viewpoint. While con- ventional RGB-based or depth-completion approaches struggle to provide reliable predictions of a depth image for such transparent objects, we propose a novel monocular framework that simultaneously estimates the depth and surface normals of transparent objects from a single polarization image. our method leverages the rich cues provided by polarization and achieves a computationally efficient depth estimation that re- quires neither analytical models of light reflection nor multi- view setups. To obtain accurate ground-truth labels for a transparent object, the proposed method uses depth and normal maps generated by existing models as pseudo ground-truth, enabling effective learning without manual labels. Experimental results demonstrate that the proposed lightweight framework achieves competitive accuracy in real-world environments.

Keywords: Robotics, Mechatronics Systems, Environment / Ecological Systems
Abstract: Traditional mobile robot localization techniques relying on on-board sensors often face significant limitations, such as visual occlusion, overreliance on visual features, and high computational costs. To address these challenges and monitor the entire environment in real time, this paper proposes an innovative localization framework that uses pressure sensors embedded under the floor to estimate a mobile robot pose based on its weight distribution. In order to reduce computational complexity, a mathematical model based on the Kirchhoff–Love plate theory is established to describe the ground deformation under external loads. This model is then used to simulate and analyze the stress distribution using the finite element method (FEM), forming the basis for the localization. The robot pose is iteratively estimated using a particle filter-based approach, which dynamically adjusts based on observed and predicted pressure distributions to arrive at the optimal result. Using environmental feedback rather than relying on on-board sensors, our approach eliminates the need to equip robots with dedicated localization hardware, reducing cost and system complexity.

Keywords: Human-robot Interaction / Collaboration, Virtual / Augmented / Mixed reality, Telecommunication Systems
Abstract: Robot remote control systems proposed for society have included approaches that utilize VR equipment, but concerns related to cybersickness still persist. In this paper, we develop a robot remote control VR simulation system that implements spatial and temporal noises, robot vibration during moving, and network connection lag-induced delays and stalls. Using this system, we investigated factors influencing the onset of cybersickness and operational workload during remote control. The experimental results suggest that communication lag may influence the onset of cybersickness. It is considered that the unpredictable nature of stalls prevents user adaptation. Regarding workload, individual spatial or temporal noise had little effect. On the other hand, a combination of robot oscillation and communication lag led to a stronger perception of workload, suggesting that the interaction of these factors increases the operational burden. Designing a human-friendly control interface of a teleoperation mobile robot by using these insights is our future project.

Keywords: Software Design, Integration Platforms, Robotics
Abstract: This paper describes an ISO/IEC 25010-based evaluation methodology for designing Digital Mock-Up (DMU) toward nuclear decommissioning. Tokyo Electric Power Company Holdings, Inc. (TEPCO) is planning to develop a DMU to facilitate safe and reliable decommissioning operations using remote-control robots since the conditions inside the reactor buildings remain many unknown unknowns. To effectively implement the DMU, it is essential to carefully select available software technologies, install them, and adopt a software system integration framework to organize the DMU. Therefore, appropriate evaluations must be conducted. In this paper, we proposed a quantitative method for evaluating the functionality and systemization of the DMU based on a software quality requirements and evaluation model ISO/IEC 25010. Also, we reported a case study in which the proposed evaluation methodology was applied to a fundamental DMU developed based on a typical use case.

Keywords: Intelligent Transportation Systems, Robotics, Machine Learning
Abstract: In this study, we propose a method for road surface estimation and obstacle detection using a fisheye stereo camera. In obstacle detection using stereo cameras, obstacles are detected based on distance information obtained through stereo matching. However, there are regions where stereo matching cannot obtain reliable disparity. Moreover, direct obstacle detection using deep learning cannot detect obstacles that are not included in the training data. Therefore, we first detect obstacles on the road surface using the relative depth obtained from monocular depth estimation. Then, by focusing only on the obstacle regions for distance measurement, we aim to detect all obstacles. Experiments demonstrate the ability to detect only obstacles with high accuracy.

Keywords: Integration Platforms, Environment / Ecological Systems, Hardware Design
Abstract: This study presents a technique to enhance Compton camera imaging used for radioactive source localization in security applications. A key issue in Compton imaging is "backscattering events," where gamma-ray interactions are misordered—such as when the absorber records a hit before the scatterer—leading to incorrect Compton cone reconstruction and image artifacts. To address this, the research applies the Time-of-Flight (TOF) principle to determine the correct interaction sequence. A detector system comprising GFAG scintillators, SiPMs, and a custom ASIC/DAQ setup achieves a 296 ps coincidence time resolution, sufficient to resolve 4.4 cm-scale interaction order. TOF values are calculated per event after correcting for systematic effects, allowing identification and rejection of backscattering events. Applying this TOF-based filtering significantly improves the signal-to-noise ratio. The resulting data, processed with the MLEM algorithm, yields an angular resolution (ARM) of 13.3°–16.5° (FWHM), aligning well with Geant4 simulations. This demonstrates TOF filtering as an effective method for enhancing Compton imaging quality.

Keywords: Robotics, Human-robot Interaction / Collaboration, Automation
Abstract: Recent advancements in industrial robotics have led to the introduction of updated safety standards, including ANSI/RIA R15.08-1/2 and ANSI/A3 R15.08-2, which establish guidelines for Industrial Mobile Robots (IMRs). However, many aspects of these standards have not yet been explored in the context of IMR safety component development. This paper presents the design and integration of safety-oriented software components that align with these standards, with the goal of enhancing safety and efficiency of IMRs in intralogistics environments. We propose a multi-layer motion planning architecture, a planning and control framework for the manipulator, and a whole-body impedance controller that treats the IMR as a unified system. Additionally, we integrate components for robot intent communication, human perception, and human-robot spatial interaction to improve safety and user experience. Validation experiments in industrial settings demonstrate the effectiveness of the developed components in ensuring safe and adaptive IMR operation in shared and collaborative human-robot workspaces.

Keywords: Robotics, Software Design
Abstract: Despite the multitude of excellent software components and tools available in the robotics and broader software engineering communities, successful integration of software for robotic systems remains a time-consuming and challenging task for users of all knowledge and skill levels. And with robotics software often being built into tightly coupled, monolithic systems, even minor alterations to improve performance, adjust to changing task requirements, or deploy to new hardware can require significant engineering investment. To help solve this problem, this paper presents Coral, an abstraction layer for building, deploying, and coordinating independent software components that maximizes compositionality to allow for rapid system integration without modifying low-level code. Rather than replacing existing tools, Coral complements them by introducing a higher-level abstraction that constrains the integration process to semantically meaningful choices, reducing configuration complexity without limiting adaptability to diverse domains, systems, and tasks. We describe Coral in detail and demonstrate its utility in integrating software for scenarios of increasing complexity, including LiDAR-based SLAM and multi-robot corrosion mitigation tasks. By emphasizing practical compositionality in robotics software, Coral offers a scalable solution to a broad range of robotics system integration challenges, improving component reusability, system reconfigurability, and accessibility to both expert and non-expert users. We release Coral open source.

Keywords: Software Design, Integration Platforms, Machine Learning
Abstract: This paper presents an embedded multi-modal vision system for drone detection, combining an event-based camera, an IMU, and an RGB sensor. The method leverages an attentional mechanism on the event stream and is robust to rotations along all three axes (roll, pitch, and yaw) of a rotating platform. The event-based sensor enables localization of fast moving objects, while the RGB camera provides classification, with the entire system optimized for embedded computational constraints. Performance analysis, with and without attention mechanisms and across various algorithmic variants, assesses the trade-off between computational cost and detection accuracy. The study identifies optimal operating situation for each configuration, validated on an outdoor test data samples.

Keywords: Software Design, Integration Platforms, Machine Learning
Abstract: How can we extract complete geometric models of objects that we encounter in our daily life, without having access to commercial 3D scanners? In this paper we present an automated system for generating geometric models of objects from two or more videos. Our system requires the specification of one known point in at least one frame of each video, which can be automatically determined using a fiducial marker such as a checkerboard or Augmented Reality (AR) marker. The remaining frames are automatically positioned in world space by using Structure-from-Motion techniques. By using multiple videos and merging results, a complete object mesh can be generated, without having to rely on hole filling.

Keywords: Robotics, Integration Platforms, Mechatronics Systems
Abstract: This paper brings the architecture for a ROS-based Hardware Abstraction Layer (HAL) to interface with the devices of a mechatronic end-effector. A HAL is a middleware between an application and the hardware, responsible for abstracting low-level operations for the client applications. A key feature of this approach is the scalability and modularity delivered by the ROS framework. To illustrate this, the paper presents a case study related to the modeling and integration of a critical-safety gripping function of the given robotic end-effector. This function is modelled and simulated using the timed automata formalism in UPPAAL, which is then implemented within the HAL using the proposed architecture. The system’s HAL controller is implemented and tested directly on the Toradex Verdin hardware. The results highlight how this design approach enables the early verification of critical mechatronic functions in robotic systems.

Keywords: Hardware Design, Machine Learning
Abstract: Additive and low-cost manufacturing techniques promise increased access to robotic actuation at the cost of mechanical precision. In this work, we employ principled Design of Experiments (DoE), including Taguchi orthogonal arrays, in parallel to sequential experimentation enabled by Bayesian Optimization (BO) for co-optimization of design and manufacturing parameters across two design case studies. We optimize for a combination of gear ratio and backdrivability in a 3D-printed compound Wolfrom bilateral gearbox. We also optimize for crack pressure and steady-state pressure differential of an injection-molded silicone check valve. Using BO, we find a 3D-printing compatible gear design with a gear ratio of 63.6 that backdrives without ever needing more than 0.35 Nm of input torque. This represents a 49% increase in ‘score’ over the Taguchi method. Similarly, we find a BO valve with lower combined crack and steady-state pressure errors than the Taguchi trials, decreasing cumulative error by 55%. Tracking model uncertainty throughout training, we conclude that further model training is necessary to reach optimal results in both cases. We further conclude that BO via the Ax platform is not yet as “plug-and-play" as Taguchi arrays.

Keywords: Control Technologies, Robotics
Abstract: Many industrial robots utilize feedback control to perform precise tasks such as welding. The trajectory tracking performance of feedback control critically depends on the accuracy of sensor information, which often suffers from link deflection and homing error of the joint displacement sensors. To address these problems, this paper introduces lightweight and highly rigid passive link systems (PL) for acquiring measurement information with reduced deflection effects. To optimally integrate these redundant measurements, a sensor fusion method is developed, based on the resolution ellipsoid, to achieve precise estimation of the end-effector position. An optimal design methodology for the PL is also proposed, and the effectiveness of both proposals is validated experimentally. Moreover, to address biases observed in the experimental data relative to the true values, we propose a bias removal method based on Recursive Least Squares (RLS). We demonstrate through experimental data that this method enables precise position measurement of the end-effector.

Keywords: Hardware Design, Virtual / Augmented / Mixed reality
Abstract: Liquid-to-gas phase change actuators are pneumatic actuators that utilize liquid-to-gas phase transitions for actuation, and have attracted attention as high-output, wirelessly drivable actuators because their heating method directly corresponds to their actuation method. In this study, we propose a shape-changing haptic display that can be fabricated through computer-controlled automated processes, utilizing actuator arrays of liquid-to-gas phase change actuators arranged on a two-dimensional plane. The developed automated fabrication system enables high-precision fabrication of actuator arrays with diverse geometric patterns such as grid and honeycomb types from design patterns, and we constructed a system capable of independent control of each pouch-type actuator through selective wireless driving using CO2 lasers and galvanometer scanners. This allows users to safely experience shape changes by directly touching the actuator arrays. We also constructed a theoretical model of polygonal actuators and established a method for calculating the minimum liquid volume for arbitrary shapes. Through pressure distribution measurement experiments, we demonstrated that the system can generate approximately 7.4 N of force per actuator and achieve selective driving without affecting neighboring actuators. This research provides a new approach to shape-changing haptic displays that requires fewer components and offers excellent scalability.

Keywords: Hardware Design, Robotics, Mechatronics Systems
Abstract: In this study, we develop an antagonistic joint using a hyper-extension actuator for the development of robots that can perform human-like movements that are smooth and flexible.Previous antagonistic joints using pneumatic artificial muscles could not maintain joint rigidity when no air pressure was applied. The hyper-extension actuator greatly elongates when air pressure is applied and has a small radial-expansion rate. By applying this actuator to the antagonistic joint, joint stiffness can be maintained even without air-pressure application. The developed antagonistic joint can achieve target joint angles and stiffness by controlling the applied pressure to the actuator, as defined by a model. The pressure-control performance of the joint angles and stiffness is verified using an actual device. The results demonstrate that the expected and measured values are in agreement.

Keywords: Hardware Design, Renewable and sustainable energy, Large-scale Infrastructure Systems
Abstract: To enhance the energy-intensive mixing of highly viscous biomass in solid-state anaerobic digestion (SS-AD), a novel peristaltic bioreactor inspired by intestinal peristalsis was developed. The reactor uses rotational flow and negative pressure inflow to enhance the mixing efficiency. A prototype's performance was evaluated using a simulated biomass (total solids = 40%), and the mixing progress was quantified using a mixing index. The experimental results showed that the proposed rotational flow achieved complete mixing, unlike conventional axial flow. Furthermore, the application of negative pressure accelerated the mixing time from 5.2 min to 3.1 min. These findings validate our intestine-inspired approach, demonstrating its potential for developing compact and energy-efficient SS-AD systems.

Keywords: Medical Training, Medical Devices, Machine Learning
Abstract: To facilitate efficient laparoscopic surgical education, a system was developed that utilizes machine learning to classify surgical skill levels—novice, intermediate, and expert—based on the motion dynamics of surgical instruments. This system not only categorizes surgical proficiency but also incorporates SHAP, an explainable AI technique, to provide insights into the rationale behind each classification result. For the machine learning dataset, the movements of four surgical instruments were recorded using a motion capture (mocap) system during total nephrectomy training sessions conducted on 46 cadaveric specimens prepared for laparoscopic surgery. The entire nephrectomy procedure was divided into three distinct processes: colon mobilization (Process 1), renal vascular dissection (Process 2), and incision and removal of the remaining tissues (Process 3). Surgical skill analysis was performed separately for each phase. Surgeons were categorized into three groups based on their prior experience with laparoscopic procedures: novices (0–9 cases), intermediates (10–49 cases), and experts (50 or more cases). A total of 111 features were extracted from the instrument motion data for each phase, and comparative analyses were conducted across the three groups. Multiple machine learning approaches—including Support Vector Machine (SVM), Principal Component Analysis followed by SVM (PCA-SVM), and Random Forest—were employed to develop models for classifying surgeons into three distinct skill levels. The classification performance of these models was subsequently validated. The results revealed that features related to efficiency and speed significantly contributed to differences in surgical skill levels. The developed system enables quantitative comparison and visualization of specific instrument characteristics. Furthermore, it provides feedback to surgeons by visualizing classification results and explaining the underlying reasoning, thereby supporting more effective sur

Keywords: Medical Training, Medical Devices, Machine Learning
Abstract: To promote efficient laparoscopic surgery education, a system utilizing machine learning has been developed to quantify surgical skill levels based on the movement of surgical instruments. In this system, the movements of surgical instruments operated by surgeons during laparoscopic cadaver surgery training are recorded using an optical motion capture system, and kinematic features are extracted from the recorded data. These extracted features are then used as input for machine learning models, with expert-evaluated scores—based on the Global Operative Assessment of Laparoscopic Skills (GOALS)—serving as the training data. The entire nephrectomy procedure was divided into three distinct processes: colon mobilization (Process 1), renal vascular dissection (Process 2), and incision and removal of the remaining tissues (Process 3). In this study, interpretable kinematic features were extracted from instrument movements during the colon mobilization phase (Phase 1). These features were used to train three regression models: Principal Component Analysis followed by Support Vector Regression (PCA-SVR), Partial Least Squares regression (PLS), and Ridge Regression. The models aimed to predict GOALS scores across five key domains: depth perception, bimanual dexterity, efficiency, tissue handling, and autonomy. Model performance was evaluated using 5-fold nested cross-validation repeated 100 times. Among the models, Ridge Regression consistently demonstrated high accuracy, with median mean absolute errors (MAEs) below 0.82 in most domains. This system is expected to contribute to more effective surgical education by providing multidimensional, objective feedback on surgical performance.

Keywords: Medical Devices, Rehabilitation Systems, Assistive Robotics
Abstract: This study proposes a method of estimating the forward tilt angle using an illuminance sensor, focusing on changes in the illuminance value of the seat surface due to changes in the posture during wheelchair use, for the purpose of preventing wheelchair accidents. The effectiveness of the proposed method was tested by fixing the position of the wheelchair and installing an illuminance sensor at the center of the rearmost part of the seat surface. The proposed equation was proposed by measuring illuminance values using a board as a preliminary experiment, and then three healthy subjects measured illuminance values five times per person as in the preliminary experiment and compared them with the proposed equation. The comparison showed that the proposed equation and the measured values had errors due to the curvature of the back, so a correction method that takes into account the curvature of the back during forward tilt was studied and the results were discussed. The results showed that the proposed correction method reduced the error of the forward tilt angle on average, and the error accuracy of the forward tilt angle estimation was within 3 °, with an average error reduction rate of approximately 23.5%.

Keywords: Medical Devices, Machine Learning, Human Factors
Abstract: Human Activity Recognition (HAR) using wearable sensors is increasingly applied in healthcare, sports, and intelligent environments. Performance is however hindered in the majority of the cases by absent sensor values, class imbalance, and inter-subject variability. We present a robust HAR pipeline that utilizesPrincipal Component Analysis (PCA) for reducing dimensions and Generative Adversarial Networks (GANs) for realistic imputation of absent values and minority-class oversampling. This is integrated into an improved DeepSense architecture with convolutional and recurrent layers for spatial–temporal feature learning. Comparisons on the OPPORTUNITY dataset, in terms of K-Fold, Leave-One-Session-Out (LOSEO), and Leave-One-Subject-Out (LOSO) schemes, demonstrate improved accuracy (+3.7%) and F1 score (+2.9%) over baseline DeepSense. The results highlight the applicability of hybrid imputation-augmentation pipelines in bringing HAR to practical, noisy sensing scenarios.

Keywords: Robotics, Automation, Medical Devices
Abstract: This paper introduces a robotic assistance system for minimally invasive middle ear surgery, focusing on precise access to the tympanic membrane. The system combines a 7-degree-of-freedom robotic arm with a hybrid visual servoing framework that integrates position-based and image-based control strategies. Dual visual feedback from a color camera and an endoscope enables robust 6-DoF pose estimation and sub-millimetric tool guidance. A model-based tracker and blob detection ensure accurate alignment and targeting, while a Quadratic Programming controller enforces safety constraints such as maintaining the field of view. The approach is validated through simulations and real-world experiments, demonstrating high accuracy, robustness to anatomical variability, and suitability for clinical integration. This work advances robotic microsurgery by providing a closed-loop, constraint-aware control architecture tailored for otologic procedures.

Keywords: Assistive Robotics, Rehabilitation Systems, Robotics
Abstract: We present a multimodal control system for a tendon-driven soft wrist exosuit that delivers personalized, angle- and mode-sensitive assistance during wrist flexion-extension. Our hierarchical architecture combines surface electromyography (sEMG) and inertial measurements through a three-layer controller comprising an intent recognizer, an angle-specific EMG normalization layer, and a low-level impedance controller. In a pilot study with four healthy participants performing repetitive wrist tasks at slow and fast speeds, with and without a 2.5 kg load, our controller substantially compensated the additional external load when present. Compared with unassisted trials under load, median normalized extensor activity decreased by approximately 30% (Cohen's d>1.7), and the fatigue-related drift in EMG across repetitions was eliminated. In contrast, assistance slightly increased EMG in the no-load condition. Assistance introduced a modest accuracy penalty: root-mean-square error and peak error rose by roughly 1-2° but remained below 4° on average. These results demonstrate that integrating angle-dependent EMG normalization with inertial sensing enables effective, task-specific unloading of the wrist while maintaining functional accuracy. Our findings lay the groundwork for larger studies and clinical applications of soft wrist exosuits.

Keywords: Robotics, Virtual / Augmented / Mixed reality, Human-robot Interaction / Collaboration
Abstract: Work-related musculoskeletal disorders are highly prevalent in physically demanding industries due to manual material handling tasks such as lifting, hauling, and carrying heavy loads. Occupational exoskeleton technology has emerged to help mitigate these injuries. Exoskeletons are wearable devices that replicate the structure of the human body to enable mechanical interaction between the user and the system. Active exoskeletons use powered actuators and sensors to provide versatile and adaptive assistance for demanding tasks. However, a gap remains in user interaction and intuitive control for these systems. Human-computer interaction technologies, including mixed, virtual and augmented reality, offer novel solutions to enhance user interaction and enable intuitive, customisable control strategies for active exoskeletons. This study integrates mixed-reality technologies with the XoTrunk active occupational exoskeleton to enable parameter tuning and system calibration through immersive interfaces. Two alternative mixed-reality interfaces were developed for this purpose: one that relies on manual user input and the other that incorporates computer vision for automatic adjustment. Experiments involving 15 participants were conducted to evaluate interfaces by performing setup and operational activities while wearing the XoTrunk exoskeleton. The results showed that the automatic interface achieved a higher System Usability Scale score (84.83/100) compared to the manual interface (78.5/100), indicating improved user acceptance and intuitiveness.

Keywords: Rehabilitation Systems, Assistive Robotics, Human-robot Interaction / Collaboration
Abstract: Each year, more than 12 million strokes and nearly one million spinal cord injuries occur worldwide. These conditions can cause severe hand impairments, leading to a significant loss of independence. Hand exoskeletons have emerged as a promising solution to support the restoration of grasping function in both rehabilitation settings and daily life. However, testing such devices remains a major challenge during the development phase. Existing hand models are either overly simplistic, lacking anatomical realism or articulation, or are complex active prosthetic systems that are costly, difficult to replicate, and not designed for passive actuation. Testing on healthy individuals introduces safety risks and bias, as their hands behave differently and may unconsciously assist movement. Moreover, direct patient testing is resource-intensive and limited. This paper presents the ADJUHand, a passive anthropomorphic hand model designed to facilitate testing of hand exoskeletons. The ADJUHand can be fabricated easily using a standard FDM 3D printer, a few screws and springs. Motion capture experiments demonstrate that the fingers follow anatomically accurate flexion trajectories with a total finger flexion angle of 249.9°. Furthermore, the stiffness of each finger can be adjusted to simulate varying levels of joint rigidity, as observed in individuals with spastic or stiff fingers. Additionally the ADJUHand enables finger abduction and thumb circumduction, facilitating diverse grasp configurations.

Keywords: Assistive Robotics, Mechatronics Systems, Hardware Design
Abstract: As the working population shrinks due to a decreasing birthrate and an aging society, greater workforce participation by women and the elderly is expected. However, the manual handling of heavy objects imposes significant physical burdens, limiting their involvement and worsening labor shortages. Conventional assistive devices support only specific body parts, leading to residual strain. In this study, a wearable weight offloading system was developed to reduce muscular load on both the lower back and the arms. Its effectiveness was demonstrated through lifting and carrying experiments with a 10 kg object.

Keywords: Human-robot Interaction / Collaboration, Assistive Robotics, Robotics
Abstract: The anticipation of human behavior is a crucial capability for robots to interact with humans safely and efficiently. We employ a smart edge sensor network to provide global observations, future predictions, and goal information to integrate anticipatory behavior for the control of a mobile manipulation robot. We present approaches to anticipate human behavior in the context of safe navigation and collaborative mobile manipulation. First, we anticipate human motion by employing projections of predicted human trajectories from smart edge sensor observations into the planning map of a mobile robot. Second, we anticipate human intentions in a collaborative furniture-carrying task to achieve a given room layout. Our experiments indicate that anticipating human behavior allows for safer navigation and more efficient collaboration. Finally, we showcase an integrated robotic system that anticipates human behavior while collaborating with an operator to achieve a target room layout, including the placement of tables and chairs.

Keywords: Human-robot Interaction / Collaboration, Robotics, Machine Learning
Abstract: In this work, we propose a human-robot physical interaction scheme designed to facilitate contact-rich manipulation tasks. In the proposed framework, neither the robot nor the human agent can complete the task independently, but a shared cost function aligns their efforts and drives them toward success. The robot agent, governed by a reinforcement learning algorithm, can exert forces and modulate its Cartesian impedance while continuously receiving evaluative feedback in the standard RL training paradigm. Simultaneously, the human agent applies forces via a standard PS4 joystick and receives both vibrotactile and visual feedback reflecting task performance. During training, the learning algorithm receives the superposition of its own and the human's actions, allowing it to implicitly benefit from the human's rapidly adapting strategy. We hypothesize that human agents can adapt more rapidly than RL and, when provided with feedback grounded in real measurements, can make more quantifiable decisions. During this rapid human adaptation phase, the robot concurrently acquires skills from the human, thereby accelerating training and improving overall efficiency. The proposed interaction scheme was evaluated in a realistic simulation environment involving 10 participants. Preliminary results indicate that participants receiving vibrotactile feedback adapted more quickly, enabling the robot to acquire the desired skill in only a few episodes for simple tasks. For more challenging tasks, human-trained RL agents required additional autonomous training, yet still achieved convergence far faster than PPO-only training. This co-adaptive framework combines the complementary strengths of humans and robots, providing a versatile foundation for contact-rich manipulation that may be extended to diverse tasks and robotic platforms.

Keywords: Human-robot Interaction / Collaboration, Assistive Robotics, Machine Learning
Abstract: In recent years, the increasing labor burden in Japanese agriculture has become a serious issue, driving the development of robots to assist in transporting harvested crops. This study proposes a method that recognizes the action transitions of agricultural workers and enables smooth transport assistance by utilizing three-dimensional skeletal information obtained from RGB-D images. Specifically, we employ Spatial Temporal Graph Convolutional Networks (ST-GCN) to detect the transition from "harvesting" to "loading." The recognition results are used to control the robot so that it approaches the worker before the loading action begins. The proposed method introduces a new labeling scheme tailored to harvesting and crop-loading motions, thereby improving recognition performance with ST-GCN. Evaluation experiments verified its generalization capability to different harvesting postures and workers, demonstrating an 18.7% improvement in action transition recognition accuracy compared with conventional methods. Furthermore, in robot-following experiments with the proposed method implemented, we confirmed that the system could both recognize action transitions and adjust the target following distance before the worker started loading. These results show that an ST-GCN specialized for agricultural tasks can effectively recognize harvesting action transitions, contributing to reducing the burden on workers during crop transportation.

Keywords: Software Design, Virtual / Augmented / Mixed reality, Human-robot Interaction / Collaboration
Abstract: This paper presents a novel system that empowers users to operate devices through full-body gestures they define themselves. Gesture-based control systems are seeing widespread application for controlling a diverse range of devices. However, most existing systems rely on a predefined set of gestures. This fundamental limitation restricts not only the number of possible operations but also the overall system flexibility. The proposed system overcomes the limitation. The proposed system comprises two primary phases. In the initial definition phase, a user wearing a HoloLens2 defines a new gesture by directly manipulating the posture of a virtual avatar. This approach facilitates a highly visual and intuitive method for gesture creation. Throughout this definition process, an external camera system captures the user's movements and computes the skeletal joint data. Then, in the subsequent operation phase, the user can control devices by performing the defined gesture without needing to wear the HoloLens2. For recognition, the system relies solely on the external camera system. This unique, two-phase methodology results in a highly adaptable and extensible framework for gesture-based interaction.

Keywords: Environment / Ecological Systems, Automation, Plant Engineering
Abstract: This paper describes a fully automated method for generating 3D pipe models with wall thickness from 3D point cloud data. In decommissioning of Fukushima Daiichi Nuclear Power Station, rapid 3D modeling of plant structures is essential for dose assessment and remote operation planning. In our method, pipe regions are first discriminated, then scanned along three orthogonal axes to extract pipe instances, followed by geometric fitting with wall thickness for use in shielding calculations. Finally, generated models are exported in Computer-Aided Design (CAD) and Building Information Modeling (BIM) formats for using facility management. Our method was validated on a 3D point cloud measured in a mock-up plant environment. We confirmed a high conversion rate and reduced processing time by restricting computation to pipe regions. The output models were imported into existing software without errors.

Keywords: Robotics, Control Technologies
Abstract: This paper presents a method that enables intuitive and stable in-hand manipulation in teleoperation of a multifingered robotic hand. Our method comprehensively handles the fundamental components of in-hand manipulation, such as object pose control, finger sliding on an object, and finger gaiting. This is achieved by constraining the operator's finger commands relative to the intended object motion, preventing their penetration into the object. Moreover, this method does not rely on any visual-based sensors and instead utilizes the six-axis force-torque sensors at the fingertips. The proposed method was evaluated with hardware through in-hand manipulation of a screwdriver, and the intuitive manipulation with high tracking performance of the operator's finger motion was demonstrated. Additionally, we present imitation learning results using the data collected by the proposed method. Policy learned with constrained finger motion as expert action maintained the performance even without the constraint during policy execution. This shows that the dataset is scalable as it is usable even without implementing the proposed constraint.

Keywords: Human Factors, Telecommunication Systems, Machine Learning
Abstract: This study aims to automatically classify emergency calls into serious (life-threatening) and minor (non-life-threatening) cases using acoustic features and machine learning models, thereby contributing to automated triage support in emergency response systems. Two enhancement strategies—noise reduction and data augmentation—are investigated to improve robustness in real-world call environments. Accurate triage during emergency calls is critical for optimizing resource allocation and ensuring timely medical response. Building on our previous exploratory analysis of acoustic features, this study advances toward practical deployment by addressing two key challenges: noisy real-world conditions and limited training data. To mitigate background noise and enhance feature stability, Wiener filtering was integrated into the preprocessing pipeline. Data scarcity was addressed through augmentation strategies, including moderate pitch shifting (±2 semitones) as well as comprehensive augmentation with pitch, volume, and noise perturbations. Acoustic features—including fundamental frequency statistics, Mel-frequency cepstral coefficients, and spectral descriptors—were extracted from call recordings provided by the Tokyo Fire Department. Three classifiers (Logistic Regression, Support Vector Machine, and Random Forest) were trained and evaluated using stratified cross-validation. Performance was primarily assessed by area under the ROC curve (AUC) and recall, given the critical importance of minimizing false negatives in emergency triage. Results showed that noise reduction improved robustness, while full augmentation yielded the greatest gains in predictive accuracy, with Random Forest achieving an AUC of 0.93. These findings demonstrate the feasibility of acoustic-based severity classification in emergency calls and highlight the potential of recall-oriented decision-support systems for emergency dispatchers. Future work will focus on real-time implementation and integration i

Keywords: Control Technologies, Robotics, Mechatronics Systems
Abstract: This study proposes a vibration-feedback method to enhance kinesthetic perception while reducing lumbar load during the teleoperation of construction machinery. The angular acceleration of the operator’s seat is dynamically estimated from joint torques, and salient vibration components are extracted from their decomposed terms. Using the partial Lagrangian method and the concept of motion-propagation force, we isolate the torque component at the base joint induced by the motion of a specific link and analyze its contribution to seat angular acceleration via wavelet transforms. Excavation experiments demonstrate that the extracted component significantly influences seat angular acceleration, suggesting that selective vibration feedback can preserve operational feel while alleviating operator workload.

Keywords: Control Technologies, Robotics, Decision-making systems
Abstract: This paper proposes a general framework for hierarchical dynamic games based on an Iterative Derivation of Optimal Policies (IDOP). The main theoretical result, stated in Theorem 1, reformulates the game using an extended state that includes the adjoint variables of all players. This enables a backward procedure in which the instantaneous optimal gain of each active player is computed while accounting for higher-priority strategies. A dedicated operator is introduced to compactly represent and solve the coupled Riccati equations arising from the Hamilton-Jacobi-Bellman framework. The method is generic and applicable to a broad class of hierarchical decision problems. Its effectiveness is demonstrated through two numerical examples and an experimental validation on a real two-degree-of-freedom (2-DOF) flexible serial robot.

Keywords: Mechatronics Systems, Hardware Design, Robotics
Abstract: In food processing sites, many tasks are still performed manually by workers, and automation through robotics is in high demand. However, research on the release motion of grasped objects remains limited, and the tendency of highly adhesive food materials to stick or remain on grippers has become a major obstacle to automation. In this study, multiple food release methods were compared and examined to ensure the reliable release of grasped food items. Based on the conditions of practically implemented combination weighing machines, particular attention was given to the method of vibrating the gripper fingers. we used a crank mechanism to induce vibrations for experiments involving pork slices laid on metal rods (modeled after the fingers of the Tsummori-Hand), while varying the vibration amplitude, frequency, and angle. The results demonstrated that vibrations exceeding a certain amplitude threshold effectively induced detachment of the adhered material. Furthermore, both higher vibration frequencies and larger vibration angles were found to enhance the release efficiency.

Keywords: Control Technologies, Automation
Abstract: The state estimation is important issues in system theory and control systems. Especially, estimation of the state variables of general nonlinear systems is essential but it is still difficult at the same time. This paper proposes design methods for an observer of a nonlinear system described by Takagi-Sugeno(T-S) fuzzy system. Because of the capability of representation of T-S fuzzy system, our observers are designed for a quite large class of nonlinear systems, which cover most systems in various engineering fields. Our observer design employs multiple Lyapunov matrix methods, which result from Lyapunov function candidate with the multiple integrals of the membership functions. Our resulting observers are a non-parallel distributed observer (PDO). This method drastically reduces the conservativeness of observer design conditions. In order to provide the usefulness of our proposed design approach, an illustrative example is provided. Finally, we end with concluding remarks.

Keywords: Intelligent Transportation Systems, Automation, Decision-making systems
Abstract: The deployment of mobile robots for material handling in industrial environments requires scalable coordination of large fleets in dynamic settings. This paper presents a two-layer framework that combines high-level scheduling with low-level control. Tasks are assigned and scheduled using the compositional algorithm ComSat, which generates time-parameterized routes for each robot. These schedules are then used by a distributed Model Predictive Control (MPC) system in real time to compute local reference trajectories, accounting for static and dynamic obstacles. The approach ensures safe, collision-free operation, and supports rapid rescheduling in response to disruptions such as robot failures or environmental changes. We evaluate the method in simulated 2D environments with varying road capacities and traffic conditions, demonstrating high task completion rates and robust behavior even under congestion. The modular structure of the framework allows for computational tractability and flexibility, making it suitable for deployment in complex, real-world industrial scenarios.

Keywords: Software Design, Robotics, Network Systems
Abstract: This study investigates the application of a hybrid ARIMA–Transformer time series forecasting model—previously validated in smart factory environments—to autonomous vehicle sensor data, in order to evaluate its domain scalability and practical feasibility. The hybrid architecture, which combines the linear forecasting capability of ARIMA with the nonlinear temporal modeling strength of the Transformer, demonstrated robust and reliable performance under complex and uncertain autonomous driving scenarios. Experimental evaluations using real-world sensor data confirmed the model’s superior accuracy under both normal and anomalous conditions. These findings underscore the potential of hybrid forecasting approaches in transportation and mobility systems, contributing to improved reliability in autonomous driving technologies.

Keywords: Robotics, Human-robot Interaction / Collaboration, Decision-making systems
Abstract: We present DELIVER (Directed Execution of Language-instructed Item Via Engineered Relay), a fully integrated framework for cooperative multi-robot pickup and delivery driven by natural language commands. DELIVER unifies natural language understanding, spatial decomposition, relay planning, and motion execution to enable scalable, collision-free coordination in real-world settings. Given a spoken or written instruction, a lightweight instance of LLaMA3 interprets the command to extract pickup and delivery locations. The environment is partitioned using a Voronoi tessellation to define robot-specific operating regions. Robots then compute optimal relay points along shared boundaries and coordinate handoffs. A finite-state machine governs each robot’s behavior, enabling robust execution. We implement DELIVER on the MultiTRAIL simulation platform and validate it in both ROS2-based Gazebo simulations and real-world hardware using TurtleBot3 robots. Empirical results show that DELIVER maintains consistent mission cost across varying team sizes while reducing per-agent workload by up to 55% compared to a single-agent system. Moreover, the number of active relay agents remains low even as team size increases, demonstrating the system’s scalability and efficient agent utilization. These findings underscore DELIVER’s modular and extensible architecture for language-guided multi-robot coordination, advancing the frontiers of cyber-physical system integration.

Keywords: Intelligent Transportation Systems, Machine Learning, Software Design
Abstract: This paper introduces a Scenario Compiler for converting textual concrete scenarios into executable XML files compliant with simulation standards. It serves as the final module in a structured framework for automated driving scenario generation, transforming natural language descriptions into simulation-ready XML files. The proposed approach combines schema-guided parsing with contextual inference using a fine-tuned large language model (LLM). In the first phase, the parser generates a complete and schema-compliant XML structure and fills in all directly extractable values, and inserting placeholders for missing or ambiguous information. The challenge addressed here is the frequent absence of explicit values for certain scenario parameters in the input text, which makes it difficult to generate fully specified XML solely through rule-based methods. In the second phase, the fine-tuned LLM infers and fills in these missing values by analyzing the broader scenario context, ensuring that the final output is both complete and plausible. An evaluation against non-specialized LLMs shows that the Scenario Compiler produces significantly more correctly instantiated XML elements while avoiding invalid tags or schema violations. By combining rule-based schema compliance with LLM-based reasoning, the approach automates the scenario generation process and reduces manual effort in simulation-based validation workflows.

Keywords: Human-robot Interaction / Collaboration, Assistive Robotics
Abstract: This paper presents an assistive system that enables blind and visually impaired (BVI) users to localize and grasp objects using spatialized audio cues, rendered as if the objects themselves emit sound. By integrating voice command recognition, RGB-D-based 3D localization, and head-tracked spatial audio via Apple AirPods Pro, the system transforms object positions into egocentric, directional prompts aligned with the user’s head orientation. We evaluated the system through tabletop grasping tasks with blindfolded sighted participants, comparing a spatial-audio (SA) condition against a speech-only (SO) baseline. While success rates were comparable between conditions, spatial audio significantly reduced task completion time and subjective workload and received substantially higher usability ratings. These findings suggest that spatialized object-originating sound can enhance task efficiency and user experience in near-field, non-visual interaction scenarios.

Keywords: Medical Devices
Abstract: In recent years, there has been increased interest in using wearable devices for continuous physiological monitoring, which has promising applications in personal health management and clinical practice. This study proposes a method for estimating heart rate using an earphone-type device equipped with a low-power, structurally simple piezoelectric sensor that captures pressure pulse wave signals in the external auditory canal. The proposed method applies two-stage noise removal filtering based on physiologically plausible R–R intervals (RRI) and signal amplitude components, enabling stable heart rate detection even under conditions with external disturbances. When applied to both resting and stimulating periods, the proposed method achieved an agreement rate exceeding 95% after noise removal and demonstrated a strong correlation with heart rate values obtained from ECG. Notably, motion and pain stimuli introduced noise into the signal during the stimulus periods; nevertheless, the proposed method effectively suppressed spurious peak detections arising from transient disturbances.

Keywords: Virtual / Augmented / Mixed reality, Human Factors
Abstract: This paper introduces a handwriting input method that uses the palm of the hand as a writing surface to address the limited input options currently available for MR devices. By providing users with an additional text entry method, this research aims to improve the overall input experience in mixed reality (MR) environments. We developed a prototype on the Microsoft HoloLens 2 and conducted a study comparing four input methods: aerial handwriting, palm-supported handwriting, whiteboard-supported handwriting, and the default virtual keyboard. The results demonstrated that having a physical writing surface significantly enhances input speed, accuracy, and usability compared to aerial handwriting. It also demonstrated the possibility of eyes-free input. This indicates that using body parts as writing surfaces is a promising approach for text entry on MR devices.

Keywords: Human-robot Interaction / Collaboration, Assistive Robotics, Human Factors
Abstract: It is difficult for deaf and hard-of-hearing people to obtain information from their hearing, particularly in group conversations where multiple speakers overlap. In this study, we propose a speech separation and recognition system that does not rely on a deep neural network but instead focuses on the removal of harmonic components. Specifically, we propose a method to extract the frequency components of one of the sounds from a mixed-gender audio signal by removing the harmonics of the other. The effectiveness of this system is evaluated by separating each individual voice from the mixed signal and measuring the recognition accuracy using an automatic speech recognition (ASR) system. We discuss the proposed method and validation results in terms of speech separation and recognition accuracy.