Abstract Introduction Dynamic postural control is a key determinant of balance and an essential factor in fall prevention, with center of pressure (COP) adjustments reflecting the motor mechanisms that ensure balance during forward-reach tasks. A ceiling-mounted fall impact mitigation robot has been introduced for balance-challenging tasks, including the Functional Reach Test (FRT); however, its effect on postural control in older adults remains unclear. Objective To investigate whether a ceiling-mounted fall impact mitigation robot supports changes in dynamic postural control during a challenging forward-reach task in healthy older adults. Methods Fifteen older adults (73.0 ± 7.4 years) were randomly assigned to experimental ( n = 8) or control ( n = 7) groups. All participants completed a pre-test FRT, followed by progressively challenging reaches starting at 98% of the individual maximum (+ 2% per successful reach) until failure, followed by a post-test FRT; the experimental group performed the post-test FRT with and without the robot, whereas the control group performed the post-test FRT without the robot. The COP in the anteroposterior direction was recorded from dual force plates. The primary and secondary outcomes were the anterior COP excursion and FRT distance, respectively. Results In the experimental group, COP excursion significantly increased after the challenging tasks ( p = 0.021), with a significant increase in the COP excursion between the pre-test and post-test FRT in the robotic condition ( p = 0.046). Conclusion The ceiling-mounted fall impact mitigation robot immediately increased the anterior COP excursion during a challenging forward-reach task, thus supporting changes in dynamic postural control in healthy older adults.
Soft robots exhibit rich deformation and contact interactions that are particularly suited to crawling locomotion. At the same time, these properties make modeling and control challenging, especially in systems with continuous deformation and strong ground contact. While simulation-based reinforcement learning has been explored, transferring learned policies to physical systems often suffers from model inaccuracies. As a result, direct real-to-real reinforcement learning has attracted increasing attention, although existing demonstrations for soft crawling robots have largely been limited to one-dimensional sagittal-plane motion. In this study, we extend real-to-real reinforcement learning to two-dimensional planar crawling using an electrically driven tendon-wire soft robot. A soft caterpillar robot equipped with four independently actuated motors enabling twisting and lateral deformation was trained directly on the physical system for approximately 2.5 h. The learned policy enabled the robot to reach a target located about 500 mm away in approximately 50 s, while exhibiting diverse crawling behaviors. Successful goal reaching was also observed from randomly initialized starting positions. These results demonstrate that reinforcement learning can effectively exploit the complex deformation and contact dynamics of soft robots by directly operating a physical robot in the real world, enabling the generation of steerable crawling behaviors.
In recent years, spherical wheels have attracted significant attention for their structural characteristics and are used for step climbing and omnidirectional movement. Although many sphere-driven robots employ friction drive systems and thus face challenges such as traveling on uneven surfaces, roller slippage, and the intrusion of debris between the sphere and the rollers, further improvements in running performance are expected. If a kinematic theory applicable to arbitrary configurations of sphere-based mechanisms is established, their use and applications are expected to expand further. In this paper, we express the general kinematics of mechanisms driven by multiple rollers using forward kinematic equations and discuss their comprehensive relationship with related studies. In addition, we validate the kinematics using a sphere measurement device, assuming various rotational motions of the sphere.
The automation of food topping in manufacturing has lagged due to the challenge of balancing weight accuracy with operational speed. Conventional robots often operate slowly, repeatedly regrasping food items to adjust the weight. To address this limitation, we propose a high-accuracy, high-speed topping robot that grasps a larger quantity of food and dispense it while simultaneously measuring its weight. A novel robotic hand has been developed that integrates fingers for grasping food and a conveyor for dispensing it. This hand eliminates the need for regrasping prior to topping and allows for repeated topping actions after a single grasp. Dynamic conveyor control including adjustments to rotation speed, direction, and target weight enhances topping precision. Experiments conducted with a prototype robot after carefully determining the control parameters demonstrate that the proposed system effectively improves both topping accuracy and processing speed.
Abstract Minimally invasive surgery (MIS) has transformed surgical practice by reducing patient trauma and improving postoperative outcomes. In laparoscopic surgery, these benefits have been further enhanced by the clinical adoption of teleoperated robotic systems, most notably the da Vinci Surgical System, which provides improved dexterity, motion scaling, and ergonomics in confined environments. As surgical robotics advances, its application is expected to extend beyond conventional MIS to microsurgical procedures requiring levels of precision and stability beyond those achievable manually. However, the clinical adoption of robotic assistance in microsurgery remains limited, particularly for minimally invasive procedures in highly constrained workspaces. Teleoperated leader–follower robotic architectures offer a promising solution for robot-assisted minimally invasive microsurgery (MIMS) by enabling precise motion scaling and tremor suppression while preserving intuitive surgeon control. Ophthalmic MIMS requires dexterous manipulation within an extremely confined intraocular workspace under millinewton-level interaction forces. Although snake-like and continuum instruments have been explored to improve access and distal dexterity, achieving multi-degree-of-freedom (DOF) motion within a submillimeter outer diameter remains challenging. These challenges stem from inherent trade-offs among bending range, shaft stiffness, wire routing, pretension, and buckling stability. This work presents the design and miniaturization of an ultra-fine multi-DOF robotic instrument for vitreoretinal surgery. The proposed instrument integrates 2-DOF distal bending (pitch and yaw), shaft rotation (roll), and a microgripper within a 0.7 mm outer diameter. To support miniaturization while maintaining manufacturability and structural integrity, a novel surface-constrained, V-type disk-stacked bending mechanism is introduced. Wire passability through reduced-diameter guide holes is geometrically verified at maximum disk tilt, and shaft stiffness and Euler buckling are analyzed using second-moment-of-area models under a conservative 10 mN lateral tip load. Prototypes with outer diameters of 0.9 and 0.7 mm were fabricated and tested. The 0.7 mm instrument demonstrated smooth pitch–yaw bending and reliable grasping, with bending hysteresis of approximately ± 5°. Shaft deflection during bending and grasping remained below 0.06 mm, while rotational whirling produced displacement amplitudes of 0.1–0.23 mm. These results highlight key design trade-offs and provide experimentally validated guidelines for the development of ultra-fine robotic instruments for ophthalmic MIMS.
While articulated robots are very versatile and can perform a variety of tasks, they are often redundant systems when viewed from a task-based perspective. In exchange for versatility, the development of task-specific hardware can be created at low cost and often has a beneficial impact on social implementation. In this study, we focused on a robot for dressing assistance, which is known as a complex task, and performed imitation learning using an articulated robot. We then created a task-specific, low-DoF hardware prototype based on the low-dimensional feature space acquired through imitation learning. The results of the prototype demonstration showed that the system can acquire motions that can be adapted to a round-back posture and a standard posture by changing the motor speed based on the information stretched and compressed in the time axis direction. This research has a great impact on the social implementation of technology because it enables the construction of low-DoF task-specific systems based on the extraction of only the features necessary for task execution by learning a complex task that appears to be high-dimensional with a high-DoF manipulator.
This paper presents a dynamic object collision avoidance algorithm for a cane-type accompanying robot designed to support individuals who can walk independently but experience anxiety. The proposed method ensures a consistent light-touch contact support point while preserving the user’s voluntary walking behavior in dynamic environments. Our approach comprises two core components. First, we propose a tentative target point selection algorithm based on a dynamic cost map. This map reflects not only static objects but also dynamic objects, allowing the robot to determine an optimal target from candidate points positioned around the user. Second, we develop a velocity command generation method using Model Predictive Control tailored for accompaniment tasks. This formulation incorporates dynamic obstacle constraints to compute optimal velocity commands in real-time. To evaluate the effectiveness of the proposed algorithm, we conducted experiments in a living lab and with a real robot. The results demonstrate that the proposed framework achieves safe and smooth accompaniment while effectively avoiding dynamic objects.
Abstract In this study, we propose a knowledge-selective transfer reinforcement learning method that simultaneously achieves heterogeneous domain transfer and knowledge selection in autonomous robots. In recent years, autonomous robots capable of recognition, decision, and action in their environments have been developed, and their utilization is advancing in a wide range of fields such as disaster response and logistics support. Reinforcement learning (RL) enables adaptive learning in unknown environments and is instrumental in realizing technologies such as autonomous robots. However, RL requires extensive exploration, leading to the issue of long training times. To address this, transfer reinforcement learning (TRL) has been introduced to reduce the training time by reusing previously learned knowledge. Nevertheless, the transfer effectiveness depends on the knowledge reused, creating a need for appropriate knowledge selection. Therefore, we propose heterogeneous domain SAP-net (HDSAP-net), which enables knowledge transfer by utilizing heterogeneous-domain knowledge sets acquired from various agents and tasks, thus extending the existing Spreading Activation Policy Network (SAP-net). In its algorithm design, HDSAP-net incorporates inter-task mapping using linear interpolation to bridge discrepancies between the heterogeneous domains. Furthermore, by analyzing the behavior of HDSAP-net, we formulated a network design method using optimal transport cost and implemented a new activation-value control method, thereby improving its performance. This enables TRL, wherein knowledge sets derived from various robots are shared among various agents, which was challenging with the conventional SAP-net. Verification experiments using the physics simulator Webots involving a mobile robot, robotic arm, and drone demonstrated that HDSAP-net significantly improves the learning efficiency when compared with that of conventional RL. The proposed method reduced the learning time by 32.1% to 82.0%, confirming its capability of autonomously discovering and utilizing effective knowledge across various domains.
We developed a non-invasive ultraviolet (UV) stimulation method to control the movement of a bio-intelligent cyborg insect by utilizing its natural sensory and motor pathways. This approach allowed the insect to retain its own decision-making ability while its movement direction could be guided. However, the control relied mainly on body motion data, making it difficult to understand how the insect perceived its environment. In this study, we investigate the relationship between physiological data and behavioral data during insect perception and propose a perception-driven control strategy. The proposed method combines insect physiological data, including low-frequency neural amplitude features and heartbeat activity, together with body motion data to estimate the insect’s environment-associated internal perception using machine learning under different environmental conditions, such as natural, UV, chemical, heat, and food. The inferred environment-associated internal perception is used within a closed-loop bio-intelligent cyborg insect control strategy to modulate its behavior. The results show that physiological data and behavioral data are linked to the insect’s environment-associated internal perception, and that perception-driven estimation can improve movement control, demonstrating the potential of the perception-driven control strategy for bio-intelligent cyborg insects in low-power robotic applications.
Robotic weeding has been attracting attention to reduce the burden of weeding in agriculture. In weeding by a robot, it is required to detect weeds correctly. In addition, it is expected that robots can remove weeds without damaging surrounding crops by detecting the stems of weeds and approaching the stems. In this paper, YOLO11, an instance segmentation method, is combined with rule-based image processing based on weed shape features to achieve a real-time stem detection method. The proposed method uses YOLO11 instance segmentation to detect the contour of the weed region, and then uses convexity defects in the weed region to detect the stem position. The proposed method was applied to a dataset taken at a farm, and both Precision and Recall of the stem detection achieved 0.886.
Soft grippers have attracted attention as a method for grasping objects with various shapes, sizes, and weights in an enveloping grasp with a simple structure. Soft grippers with various structures have been proposed, such as multi-finger soft grippers using pneumatic actuators and fingerless grippers such as jamming grippers. However, they have been developed with a focus on a variety of objects that can be grasped. Therefore, when the shape error of the object to be grasped is limited to a certain range, there is still room for the optimization of the shape of the grasping part and other aspects. In this study, we propose a soft gripper driven by a spherical actuator that offers a high degree of freedom in gripping part design. The spherical actuator part can open and close widely for any width, thickness, number, and shape of gripping parts at the rim, allowing the design to be optimized according to the characteristics of the object to be grasped. We propose a method for designing a soft gripper specialized for each grasped object by taking advantage of this degree of freedom. This proposal contributes to the development of a specialized approach focused on enveloping grasps for specific objects, in contrast with a generalized approach aimed at grasping a wide variety of objects using a single system.
The maneuverability of a dual propulsion VTOL drone may be enhanced by the two types of rotors mounted on it, one for rotary-wing mode and the other for fixed-wing mode, when they are used simultaneously. This paper investigates how much altitude gain can be achieved within a specified time from a high-speed cruise flight by using the two types of rotors together, and discusses the flight states and the forces applied to the aircraft during the ascent. A dynamic model of the drone is developed by modeling the aerodynamic and thrust forces. Using this model, numerical optimization is performed to obtain a rapid ascent maneuver that maximizes the altitude gain at a specified end time, with the rotor battery voltages and the elevator angles during the maneuver as the optimization variables. The altitude increase achieved by the rapid ascent maneuver is much larger than that obtained in the conventional fixed-wing mode flight. In the early stages of ascent, both aerodynamic lift and rotor thrust forces are used together to accelerate the drone upward. In the later stages, the drone continues its ascent using the rotor thrust forces while suppressing large pitch angles caused by aerodynamic pitch-up moments due to a negative angle of attack. These results demonstrate the enhanced maneuverability of a dual propulsion VTOL drone in achieving rapid ascent by simultaneously utilizing both types of rotors.
Deep learning-based image semantic segmentation techniques have made great strides in recent years. However, they still need large amounts of finely annotated image data, and generalizing the model from known classes to unknown ones remains a challenge. Most of the work on few-shot semantic segmentation techniques deals with the support set by directly utilizing images and masks for feature fusion. That tends to make the model not pay enough attention to the less-sample category, leading to missed detections. To alleviate this problem, in this paper, we propose the Region Select Enhancement Network, a novel structural model composed of base and meta learner, based on the perspective of metric learning and data enhancement. We employ an additional base learner to individually recognize targets within the base class, utilizing the recognition results of the base class as background-guided features for the final target. We then effectively fuse the outputs of the base learner and meta learner to produce accurate target images. Notably, unlike the common meta learner, we add a separate target category selection enhancement branch to the meta learner, augmenting the target features with known information from the support set. This further reduces background interference, thereby improving the model’s generalization ability. We conducted experiments on Cityscapes- 3^i , a few-shot outdoor dataset constructed from labeled images in the Cityscapes dataset, to validate the effectiveness of our method.
While articulated robots are very versatile and can perform a variety of tasks, they are often redundant systems when viewed from a task-based perspective. In exchange for versatility, the development of task-specific hardware can be created at low cost and often has a beneficial impact on social implementation. In this study, we focused on a robot for dressing assistance, which is known as a complex task, and performed imitation learning using an articulated robot. We then created a task-specific, low-DoF hardware prototype based on the low-dimensional feature space acquired through imitation learning. The results of the prototype demonstration showed that the system can acquire motions that can be adapted to a round-back posture and a standard posture by changing the motor speed based on the information stretched and compressed in the time axis direction. This research has a great impact on the social implementation of technology because it enables the construction of low-DoF task-specific systems based on the extraction of only the features necessary for task execution by learning a complex task that appears to be high-dimensional with a high-DoF manipulator.
Current research has focused on developing robots that can imitate the musculoskeletal structure of human body to achieve skillful movements similar to those of a human. Several studies have investigated the end-point trajectory of musculoskeletal systems, but a sensor is essential for trajectory tracking. The previously proposed muscular internal force feedforward control method achieves point-to-point convergence to a desired position without any sensor; however, it lacks explicit control over the specific end-point trajectory from the initial to the desired position. Given that the end-point trajectory depends on the shape of the potential field generated by the control input, the trajectory sometimes takes a detour depending on how the desired position is set. To resolve this limitation, in this study, a novel shaping method is proposed for the end-point trajectory of musculoskeletal systems using the muscle redundancy. The end-point trajectory is shaped by optimizing a control input that cannot be uniquely determined due to the muscle redundancy based on the objective function of joint torques. As an advantage in the proposed method, the end-point trajectory can be shaped without sensors. Furthermore, the control input can be prevented from saturating the output limit by formulating the problem as a constrained minimization problem. The effectiveness of the proposed method is presented using simulation results. The proposed method definitely converges at the desired position without depending on the shaping result of the end-point trajectory because the target is a musculoskeletal system with muscular arrangements that converge to the desired position.
Abstract We propose an occlusion robust 2D scan matching method utilizing template matching and ray tracing. Scan matching is a fundamental technique for mobile robots in many application scenarios, such as localization, loop closure in SLAM, and object detection. Although it is an extensively researched topic, dealing with the occlusion of 2D LiDAR scans remains challenging due to the limited discriminative information in 2D range scans. We therefore leverage ray tracing within SLAM to explicitly label unknown (occluded) vs. empty (free) cells, and incorporate this occlusion information into the matching score. Specifically, we demonstrate the effectiveness of our approach by solving a stop position determination problem, where scan matching is implemented to recognize a reference object in a 2D map and the mobile robot determines its stop position based on the recognized reference object. The reference object template is prepared in advance by cropping the reference object region from a 2D map. When the position of the reference object changes, the robot creates a new 2D map of the updated environment by using ray tracing, detects the template on the new map, and calculates the new stop position based on the detection result. The technical challenge of this strategy is that the shape of the reference object on the map changes when its position or the set of observable sides changes due to occlusion. Therefore, we propose a template-matching-based detection method that takes occlusion into account. By assigning lower penalties to unknown areas while penalizing empty-occupied conflicts, the matcher becomes less sensitive to occlusion and more selective against false positives. In a verification experiment, we confirmed that our algorithm could detect the reference object with occlusion more robustly than a baseline template matching method in a simulated environment. In object recognition with an occlusion rate of approximately 50% to 75%, the proposed method was able to reduce the recognition error rate by 1/5 to 1/3 compared to the baseline method.
This paper presents a bio-inspired soft climbing robot designed to overcome the limitations of conventional climbing robots in unstructured, irregular, or deformable environments. Mimicking the arboreal locomotion of sloths, the robot employs an embracing-based anchoring mechanism driven by a tensegrity-structured spiral system, capable of generating gripping forces exceeding 20 N through topological interlocking. Coupled with a telescopic actuation module utilizing tunable tower springs, the robot achieves adaptive body deformation and bidirectional locomotion. Experimental results demonstrate stable vertical climbing, passive adaptation to curved and variable-diameter surfaces, active gap-crossing via elastic energy release, and reliable performance on deformable substrates. This work establishes a new paradigm for robotic locomotion in complex 3D environments by integrating bio-inspired topological constraints with elastic actuation.
Abstract Polymerase chain reaction (PCR) testing was widely used for diagnosing infectious diseases, particularly during the novel coronavirus disease (COVID-19) pandemic. The automation of this process has been limited by a severe lack of compact microtube cappers/decappers that accommodate a wide array of microtubes. To address this issue, we developed an automatic microtube capper/decapper (AMC/D) system, which we call the single-AMC/D (S-AMC/D) because it handles individual tubes one at a time. Subsequently, recognizing that microtube handling is a fundamental operation not only in PCR testing but also in various clinical examinations and biological experiments, and considering the strong demands of clinical laboratory and biological laboratory personnel, we developed a manually operated microtube equipped with the AMC/D. We also identified a significant demand among laboratory automation system developers and automatic pipetting equipment manufacturers for multistation AMC/Ds capable of handling numerous microtube caps simultaneously to increase throughput. In this study, we propose a multistation automatic microtube capper/decapper (M-AMC/D) with improved throughput for automating clinical experiments and bioexperiments that use microtubes. We adopted two design concepts to develop this M-AMC/D: According to the first one, eight microtubes can be opened and closed at one process. According to the second one, eight microtubes can be driven by staggering the timings of their opening and closing operations so that the loads generated by the torques of multiple opening and closing operations are not applied to the drive system or structural system at the same time. Based on the design concepts, we performed the basic design of an eight-station AMC/D (8-AMC/D) with a new driving mechanism using cams and cam followers. The installation space (249 mm × 93 mm) of this system is approximately 35% of the installation space required for the eight S-AMC/Ds (124 mm × 78 mm × 8 units). Finally, we developed a prototype of the proposed M-AMC/D and conducted a function confirmation experiment and a verification experiment to improve the throughput. The experimental results of throughput improvement showed that the operation time of the 8-AMC/D was 110.5 s—a significant reduction of 343.3 s (or 75.7%) from the S-AMC/D operation time of 453.9 s. Thus, the effectiveness of the proposed 8-AMC/D was verified. The proposed device enables simultaneous handling, opening, and closing of eight microtubes in a compact and efficient configuration. Although initially motivated by the urgent demand for PCR testing during the pandemic, the proposed system addresses a fundamental challenge in laboratory automation involving press-type microtubes and is applicable to a broad range of laboratory procedures involving microtubes.
Teleoperated robots are revolutionizing remote work for individuals with severe motor dysfunction by expanding the scope of employment opportunities. However, the complexity of teleoperation often results in tasks requiring significant time to complete, and concerns have emerged regarding a decline in the quality of subjective experiences. This study introduces an assistive method that automatically switches between pre-stitched multi-view task-assistance layouts, which are designed to enhance spatial continuity, and non-stitched zoomed-in views, which are conducive to specific task processes. Therefore, this study focuses on the fundamental problem of determining whether the advantages of each layout can be effectively combined via automatic switching. We explore the challenges involved in achieving increased efficiency while enhancing the quality of subjective experiences. A key design constraint is that no additional input must be required for view switching, assuming that operators with severe motor dysfunction may have limited control over their upper limbs. This study involved 11 participants, including individuals with severe motor dysfunction, who assessed the proposed method utilizing a set of indices derived from a scenario modeling item delivery within a nursing facility. A trained assistant operator performed automatic switching to evaluate the feasibility of the approach. A fundamental concern with this method is that sudden changes in the size and position of a view within the pre-stitched multi-view interface may require the operator to re-recognize the remote environment, potentially conflicting with their intentions. However, results indicate that the proposed method enhances task performance in terms of task failure rate and completion time, potentially owing to the high consistency of the layouts before and after switching. Furthermore, the results indicate that the mental demand and frustration caused by external assistance are crucial factors for optimizing both task performance and subjective satisfaction. These findings contribute to expanding the scope of assistance modalities by switching between various types of operational support and increasing the potential for remote work utilizing teleoperated robots for diverse individuals.
Farmers working on steep terrain need comfortable tools to improve their working conditions. Olive harvesting, which uses long unbalanced tools for reaching high branches, causes back pain. The posture of farmers, during harvesting operations, has been studied to determine the factors that are relevant in the evaluation of physical fatigue. The paper describes an exoskeleton designed to assist farmers during harvest and pruning. The exoskeleton, worn by the farmer, offers an additional arm to support the harvester. The proposed exoskeleton uses a spring to better balance the weight of the tool on the user’s body, reducing the weight on the upper limbs. The kinematics of the selected architecture is analysed. The exoskeleton is discussed with an emphasis on its advantages and disadvantages. A group of operators has tested the exoskeleton on the field. Myoelectric tests have been performed to evaluate the fatigue. The results are positive, the exoskeleton reduces the perceived load, allowing sufficient dexterity to work outdoors. The research shows that it is possible to create inexpensive exoskeletons for agriculture. The proposed exoskeleton can also be successfully used for branching, painting, and cleaning glass.
