Keywords: Biomimetic robotics, Cyborg intelligence
Abstract: This paper addresses the path-following challenge for a turtle-inspired biomimetic robot propelled by 3 degree of-freedom flippers. First, a high-fidelity hydrodynamic model is established through computational fluid dynamics (CFD)- driven parameterization. Then a hierarchical control framework combining model predictive control (MPC) and fuzzy logic control (FLC) is proposed to address nonlinear hydrodynamic interactions, environmental uncertainties, and actuator mapping complexities. The upper-layer MPC optimizes reference forces and moments via receding horizon optimization, while the lower-layer FLC adaptively translates these references into actuator-specific commands through fuzzy rule-based signal allocation. Finally, comparative simulations in a U-shaped and 8-shaped path following task demonstrate the proposed strategy’s superiority, achieving a 70% reduction in average cross-track error and 50% faster convergence than conventional PID and sliding mode controllers under randomized hydrodynamic disturbances.

Keywords: Biomimetic robotics
Abstract: The concept of aerial-aquatic robots offers an innovative solution for devices capable of operating in both air and water. While previous research on such robots has primarily focused on conventional technologies, such as fixed-wing and multi-rotor aircraft, the flying fish—a species adept at both swimming underwater and gliding above the surface—has not been extensively explored as a biomimetic model, particularly with regard to its motion patterns involving collapsible pectoral fins. In this study, we developed a novel bio-inspired robot with a detailed dynamic model that integrates collapsible wings actuated by soft hydraulic systems and highlight a unique multi-modal motion pattern. Based on extensive testing of the soft hydraulic actuators and aerodynamic coefficients, simulations are conducted under various conditions and confirm the viability of the design.

Keywords: Biomimetic robotics, Neuro-Control and communication, Other related topics
Abstract: Identifying key environment-matched traits or optimal trait solutions forms a critical foundation for biomimetics research. This study investigates how Asiatic toads (Bufo gargarizans) adapt locomotor traits and cerebellum size to energy-limited high-altitude environments. Through comparative analyses of two populations (926 m vs. 3239 m, 30 individuals per population), it was found that high-elevational toads significantly exhibited shorter stride lengths, lower center of mass, and smaller cerebellum, correlating with reduced energy expenditure. These biomechanical adaptations highlight trade-offs between movement efficiency and neural control complexity, offering insights for designing energy-efficient bio-inspired robots. More importantly, this study demonstrates how intraspecific comparative methods in evolutionary biology can enhance biomimetic optimization design.

Keywords: Advanced materials and sensors for robotics, Wearable robotics, Biomimetic robotics
Abstract: Structural color hydrogels are promising material candidates for optical displays, smart sensing, and biological interfaces. However, it is still a challenge to formulate structure color hydrogels with multifunctionality (e.g. robust toughness, strong adhesion, high stretchability, and high electronic conductivity) without compromising their intrinsic optical properties. In this paper, we report a multifunctional structural color hydrogel by incorporating a conductive polymer Poly(3,4-ethylenedioxythiophene)-poly (styrene sulfonate) (PEDOT: PSS) into a mechanically robust poly (vinyl alcohol) (PVA)/poly (acrylic acid) (PAA) double network (DN) hydrogel. The as-prepared structural color hydrogel could span a wide range of mechanical properties by simply tuning the polymer composition and the number of freezing-thawing cycles. The dynamic hydrogen bonding interactions endow the hydrogel sensor with reversible adhesion to different substrates. In particular, the structural color hydrogel shows high sensitivity in optical and electronic sensing capability, and can accurately detect some human movements. This work provides useful insights into the development of functional structural colors for wearable soft electronics.

Keywords: Biomimetic robotics, Advanced materials and sensors for robotics, Micro/Nano robotics
Abstract: In operations involving movement and gripping, it is crucial to minimize the preload stress at the contact interface to prevent damage to the surface of the object being manipulated. However, negative pressure suction cups typically require significant preload to expel media from the chamber and ensure adequate adhesion between the edge and substrate for effective suction. This necessity limits the application of such suction cups on sensitive surfaces. Gecko-inspired adhesive films can adhere under minimal preload conditions; yet, they offer limited normal adhesion force. Consequently, this study designs a negative pressure suction cup featuring an annular inclined micro-wedge structure at its lip, which enables the establishment of dry adhesion at low preload levels while maintaining conformity with the contact surface. When the edges conform properly, the suction cup provides substantial adhesion force. The fabrication of such cross-scale annular inclined wedge structures poses a challenge, addressed herein through a novel method combining tip-based micro-nano shape regulation with lathe processing. This approach facilitates the creation of annular wedge structures at the periphery of the suction cup, generating dry adhesion forces that maintain the seal of the negative pressure suction cup and reduce the preload required for establishing negative pressure. Furthermore, this fabrication technique is applicable for large-area micro-nano structural manufacturing.

Keywords: Micro/Nano robotics, Biomimetic robotics
Abstract: Magnetrically controlled soft robots demonstrate significant potential in the precise treatment of intestinal diseases due to their accessibility in narrow cavities, remote-controllable motion, and precise targeting capabilities. However, the intestinal application of magnetic soft robots faces formidable challenges posed by complex environments, including large deformations induced by intestinal peristalsis and protrusions formed by mucosal folds. Inspired by the jumping characteristics of mudskippers and their locomotion adaptability to complex intertidal terrains, this paper proposes a mudskipper-inspired soft jumping robot. Under an external magnetic field, the robot achieves controllable jumping motions at varying heights, enabling it to effectively traverse fold-induced protrusions in the intestine and accurately reach target lesion sites. Dynamic and kinematic models are established to analyze and predict the jumping behavior of the bio-inspired soft robot. Experimental and simulation results demonstrate that the fabricated mudskipper-inspired soft jumping robot exhibits excellent jumping capabilities and motion controllability. This research provides a theoretical framework for designing magnetic soft robots adaptable to intestinal peristalsis and fold-rich environments. The proposed approach holds promise for enabling precise drug delivery in complex intestinal environments and paves a new path for the precise treatment of intestinal diseases.

Keywords: Neural-machine interface
Abstract: Stroke rehabilitation remains a significant clinical challenge, particularly for patients with chronic motor impairments who exhibit limited recovery under conventional therapies. This paper introduces BrainFormer, a novel electroencephalography (EEG)-based deep learning model combining convolutional neural networks (CNNs) and Transformer architectures to classify neural signals with high precision. Specifically, BrainFormer targets two critical biomarkers for stroke rehabilitation: (1) Steady-State Visual Evoked Potentials (SSVEP) for detecting visual attention and (2) alpha wave suppression as an indicator of motor-related cortical activation. The model achieves 92% accuracy in SSVEP classification and 100% accuracy in binary detection of alpha wave suppression, demonstrating its robustness for EEG decoding. Furthermore, a low-cost, portable 16-channel EEG acquisition system is introduced and designed for clinical and home use, enabling scalable implementation. These results highlight BrainFormer’s potential as a foundation for a realtime, adaptive neuroregulation system that can be integrated into future rehabilitation technologies, addressing the gap in patient engagement during therapy.

Keywords: Rehabilitation robotics, Neuro-Control and communication, Neural-machine interface
Abstract: Brain-computer interface (BCI) plays an important role in interactive rehabilitation training. Action recognition in rehabilitation training is the basis and key to realize interactive rehabilitation training. With the development of deep learning, motion image deep learning models with higher decoding accuracy continue to emerge. However, extracting single features, such as surface electromyography (sEMG) and electroencephalogram (EEG) signals, may lead to poor decoding performance in muscle fatigue states. Therefore, this paper proposes a dual-scale fusion neural network (DFNN) for accurate recognition of hand rehabilitation training actions under muscle fatigue, which realizes feature learning of EEG and EMG fusion. We evaluated the performance of the model on our own multi-modal dataset collected when performing rehabilitation training actions in fatigue states, achieving an average recognition accuracy of 93.33%.

Keywords: Rehabilitation robotics, Other related topics
Abstract: In the early stages of stroke rehabilitation, rehabilitation training can effectively promote the recovery of lower limb function in patients. To achieve bilateral coordinated assistance for the lower limbs, we designed a novel deep learning model to perform end-to-end prediction of the affected-side knee joint torque using bilateral joint angles and unaffected-side EMG signals. This model integrates Gated Recurrent Units (GRU), Token Statistics Self-Attention (TSSA), and an iTransformer encoder architecture. Experimental results demonstrate that the proposed fusion model achieved the best predictive performance among the six subjects, with average Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) of 0.8518 N·m , 0.9183 N·m , and 0.8248 N·m, respectively. Compared with other mainstream models, it exhibits superior prediction accuracy and generalization capability.

Keywords: Neuro-Control and communication, Rehabilitation robotics, Other related topics
Abstract: 功能性电刺激 （FES） 是一种有效的 脑卒中后手功能康复技术， 然而，它的进步受到效率低下和 传统校准方法的可扩展性差。这 论文提出了一种基于 计算 建模来解决这个瓶颈。核心贡献 在于实现“逆向设计”——从期望 功能的 最佳刺激方案的手势——通过建立 从刺激参数到 神经激活状态。我们利用招聘指数 （RI） 和选择性指数 （Sel） 用于量化神经 激活。 通过计算机优化，该框架提供了一个 以目标为导向的范式，用于精确和高效开发 下一代 FES 系统。

Keywords: Wearable robotics, Biosensors and bioactuators, Neuro-Control and communication
Abstract: This study enrolled 42 patients with stroke accompanied by upper limb hemiparesis and 42 healthy controls, utilizing functional near-infrared spectroscopy (fNIRS) to acquire hemodynamic signals during a finger-to-nose task and investigate differences in brain network and connectivity characteristics. Results revealed that, during the task-state, the stroke group exhibited significantly reduced average node degree, clustering coefficient, and network density in neural activity frequency bands compared to the healthy group (p<0.01), specifically manifesting as attenuated functional connectivity between the supplementary motor area (SMA), right premotor cortex (PMC), and other brain regions (p<0.01). Intragroup analysis demonstrated increased average node degree, clustering coefficient, and network density in the stroke group during the resting-state (p<0.05), predominantly characterized by enhanced connectivity in left hemispheric regions. These findings suggest fNIRS-monitored neural frequency band physiological information may serve as a biomarker for early post-stroke neurovascular coupling abnormalities, which requires further validation.

Keywords: Regenerative medicine, Neural-machine interface, Bio-Cell assembly and tissue fabrication
Abstract: The reconstruction of damaged neural circuits is critical for neurological repair after brain injury. Classical brain-computer interfaces (BCIs) allow direct communication between the brain and external controllers to compensate for lost functions. Importantly, there is increasing potential for rehabilitative BCIs to input information into the brain to restore damage, but their effectiveness is limited when a large injured cavity is caused. Notably, it might be overcome by transplantation of brain organoids into the damaged region. We constructed organoid-brain-computer interfaces (OBCIs), an innovative BCIs mediated by implantable organoids. We assess the prolonged safety and feasibility of the OBCIs and explore neuroregulatory strategies. The results indicated that OBCI stimulation promotes progressive differentiation of grafts and enhanced structural-functional connections within organoids and the host brain, promising to repair the damaged brain via regenerating and regulating, potentially directing neurons to preselected targets and recovering functional neural networks in the future.

Keywords: Other related topics, Biosensors and bioactuators, Biomimetic robotics
Abstract: Untethered motion, omnidirectional locomotion, and crossing obstacles on walls are important requirements for wall-climbing robots. For realizing all these requirements, a lightweight wall-climbing robot with hybrid adhesion has been proposed in this paper, whose weight is only about 167.7 g. Omnidirectional locomotion can be realized by using two threaded rods distributed in a cross shape to combine vertical and horizontal motions. The hybrid adhesion combined a flat adhesion and an electroadhesion to enhance the adhesion force, which is beneficial to realizing untethered motion and improving load capacity of the foot. The robot can cross obstacles by raising its feet as high as possible. This integrated robot is expected to achieve detection, surveillance and rescue tasks in complex environments.

Keywords: Cyborg intelligence, Other related topics
Abstract: In order to improve the mobility and adaptability of search and rescue robots in complex and unstructured terrain environments, this paper proposes a reconfigurable tracked search and rescue robot. The robot has multiple configurations for common post-disaster terrain conditions. Through the auxiliary obstacle-crossing module, the robot can ensure that at least one set of tracks is in contact with the ground and pull the robot to cross the obstacles, which improves the obstacle-crossing ability of the robot, and through the curvature-adaptation module, the track modules can always maintain the maximum contact area with the ground, which improves the robot's traction force and mobility ability. Finally, simulations and tests are conducted, and the robot has excellent mobility and obstacle-crossing ability for unstructured environments such as variable curvature and multiple obstacles compared to existing robots.

Keywords: Cyborg intelligence, Other related topics
Abstract: To enhance the adaptability of conventional wall-climbing robots to curved surfaces, particularly pipe structures, a wheel-type wall-climbing robot is proposed by leveraging the advantages of passive mechanisms in autonomous system design. The robot features two mechanical cantilevers connected via a differential center. Each cantilever is equipped with two magnetic spherical wheels incorporating passive magnet supports, enabling the system to conform passively to both concave and convex surfaces with varying curvatures. The minimum magnetic force required for a single magnetic wheel is determined through the static analysis of the robot. Magnetic field simulations are then conducted to verify whether the magnets can generate sufficient force to meet this requirement. Finally, a prototype was built, and motion performance tests were conducted to show its performance.

Keywords: Micro/Nano robotics
Abstract: In recent years, micro-miniaturization, swarm intelligence, and intelligent control have become mainstream trends in robotic systems research. Micro-robot systems are gaining attention due to their flexibility, robustness, and scalability. However, limited computing and storage resources pose challenges for applying traditional path planning and collaboration algorithms in such platforms. In addition, many existing solutions assume simplified environments and reliable communication, which are often unrealistic. This thesis investigates a collaborative diffusion-based search strategy tailored for micro-robot systems. To address resource limitations, a DBT-RRT* path planning algorithm is proposed. It integrates a target biasing mechanism and bidirectional search with dynamic step size, while applying the triangle inequality for post-processing path optimization. The algorithm achieves efficient planning in complex environments with minimal computational cost. Its effectiveness is validated through experiments.

Keywords: Advanced materials and sensors for robotics, Biomimetic robotics, Micro/Nano robotics
Abstract: Unmanned aerial vehicle (UAV) grasping systems have expanded the range of UAV applications, such as item delivery, dynamic perching, infrastructure inspection, search and rescue operations, industrial maintenance, precision agriculture, scientific sampling. However, prevalent gripper-based and fixture-based mechanisms are limited by restricted applicability, potential for causing significant mechanical stress on targets, and high demands on target properties. This study presents and fabricates a liquid crystal elastomer (LCE)-based adhesive patch for UAV grasping. Utilizing the pronounced temperature-dependent viscoelasticity changes of LCE, the patch enables stable adhesive grasping and controllable, heat-triggered release of objects. The LCE adhesive patch is synthesized through an efficient one-step click chemistry reaction and can be customized for diverse scenarios using a molding technique, facilitating its easy deployment on various UAV platforms. Only a simple heating module is required to actuate the adhesion/release cycle. This novel adhesive grasping mechanism, characterized by deployment flexibility and grasping reliability, broadens the application scope of UAV grasping systems significantly, offering innovative solutions for tasks including object retrieval, precision delivery, and dynamic perching.

Keywords: Biomimetic robotics, Other related topics
Abstract: Traditional firefighting methods face significant limitations in terms of rapid response and precise operation for fires on high-rise buildings. This paper introduces a UAV-based fire suppression system integrated with a robotic arm, designed to accurately deploy fire suppression munitions (FSMs) through windows. First, we discuss the strategy of throwing FSMs and propose the throwing control algorithm during flight. Second, we design the mechanical structure of the robotic arm, integrate the control system of the UAV and the robotic arm, and construct the prototype. Finally, we conduct a series of experiments to validate the system's feasibility. The results demonstrate that the system can achieve precise deployment through the target window at various heights and speeds, with an average success rate of approximately 69%. This novel bionic projectile-throwing method provides a new approach for high-rise fire suppression.

Keywords: Cyborg intelligence, Biomimetic robotics, Other related topics
Abstract: Efficient motion planning algorithms are essential in robotics. Optimizing essential parameters, such as batch size and nearest neighbor selection in sampling-based methods, can enhance performance in the planning process. However, existing approaches often lack environmental adaptability. Inspired by the method of the deep fuzzy neural networks, this work introduces Learning-based Informed Trees (LIT*), a sampling-based deep fuzzy learning-based planner that dynamically adjusts batch size and nearest neighbor parameters to obstacle distributions in the configuration spaces. By encoding both global and local ratios via valid and invalid states, LIT* differentiates between obstacle-sparse and obstacle-dense regions, leading to lower-cost paths and reduced computation time. Experimental results in high-dimensional spaces demonstrate that LIT* achieves faster convergence and improved solution quality. It outperforms state-of-the-art single-query, sampling-based planners in environments ranging from R^8 to R^14 and is successfully validated on a dual-arm robot manipulation task. A video showcasing our experimental results is available at: https://youtu.be/NrNs9zebWWk.

Keywords: Other related topics, Neural-machine interface, Biomimetic robotics
Abstract: Enabling robots to efficiently search for and identify objects in complex, unstructured environments is critical for diverse applications ranging from household assistance to industrial automation. However, traditional scene representations typically capture only static semantics and lack interpretable contextual reasoning, limiting their ability to guide object search in completely unfamiliar settings. To address this challenge, we propose a language-enhanced hierarchical navigation framework that tightly integrates semantic perception and spatial reasoning. Our method, Goal-Oriented Dynamically Heuristic-Guided Hierarchical Search (GODHS), leverages large language models (LLMs) to infer scene semantics and guide the search process through a multi-level decision hierarchy. Reliability in reasoning is achieved through the use of structured prompts and logical constraints applied at each stage of the hierarchy. For the specific challenges of mobile manipulation, we introduce a heuristic-based motion planner that combines polar angle sorting with distance prioritization to efficiently generate exploration paths. Comprehensive evaluations in Isaac Sim demonstrate the feasibility of our framework, showing that GODHS can locate target objects with higher search efficiency compared to conventional, non-semantic search strategies. Website and Video are available at: https://drapandiger.github.io/GODHS

Keywords: Cyborg intelligence
Abstract: This paper introduces a novel safety-aware event-triggered disturbance compensation (SAETDC) mechanism to ensure system safety and stability in the presence of external disturbances and multiple obstacles. The proposed methodology improves the dual robustness of both safety and stability within standard model predictive control (MPC) frameworks, particularly when subjected to external disturbances, and balances computational resource and system performance. Firstly, we construct an augmented system by mapping barrier states (BaS) to an initial input-affine system. This transformation ensures that the stability of the augmented system directly implies the safety of the original system. Secondly, a disturbance observer is proposed to compensate for disturbances and result in residual disturbances. Based on these residual disturbances, tightening strategies for state and input constraints are introduced. This enhances the robustness of MPC with respect to safety and stability. Thirdly, an event-triggering mechanism is proposed based on these residual disturbances and state deviations. This mechanism conserves computational resources and prevents Zeno phenomena even under residual disturbances, ensuring that the triggering conditions are not continuously met at the boundaries of the safe set. The proposed method has been validated across various nonlinear input-affine systems, including a planar double-integrator, a cart-pole swing-up, a differential wheeled robot, and a robot manipulator. Simulations and experiments demonstrate the fulfillment of safety and stability even in the presence of disturbances and obstacles.

Keywords: Cyborg intelligence, Biomimetic robotics, Micro/Nano robotics
Abstract: In recent years, deep reinforcement learning (DRL) has made significant breakthroughs in quadruped robot locomotion control, particularly in enabling high-speed and highly robust dynamic behaviors that conventional model-based approaches struggle to achieve. Despite these advances, most existing DRL-based methods use velocity commands as control references, which often result in limited gait diversity and a lack of global navigation capability. To overcome these limitations, this work proposes a novel reinforcement learning framework that leverages a sequence of target points as navigation references. This framework allows the robot to autonomously plan its gait and velocity based on the navigation goal, thereby achieving more flexible and adaptive locomotion. Experimental results demonstrate that the proposed approach outperforms traditional velocity-command-based strategies in terms of both locomotion speed and gait diversity.

Keywords: Biomimetic robotics
Abstract: Existing reinforcement learning methods typically rely on precise environment modeling and trajectory planning, and can only handle a single type of obstacle. To address this limitation, this paper proposes a depth information-based reinforcement learning motion control method for quadruped robots, enabling the robot to perceive the terrain via an onboard depth camera and move autonomously. In addition, based on a dual-distillation framework, privileged information is transformed into policy inputs that can be extracted in real-world environments, and subsequently utilized to generate the final action commands. Experimental results show that the quadrupedal robot is able to maintain stable locomotion in complex terrains, and both simulation and physical experiments validate the effectiveness of the proposed method.

Keywords: Other related topics, Cyber-Physical bio-system
Abstract: Social acceptance has become a critical factor in behavioral design for mobile robots during navigation, where robotic systems should avoid causing discomfort or panic in crowds. To address the problem where existing methods fail to assess crowd states and result in uncontrolled traversal through social crowds during robot movement, we propose a dynamic decision-making navigation approach for robots based on behavioral states of social crowd. Building upon social spatial relationship theory, the concept of a “group” is introduced by defining a crowd engaged in normal social interactions as a cohesive whole. Mobile robots need to avoid crossing these social groups in order to respect the social needs of the crowd in navigation and reduce intrusion. Secondly, in order to increase the success rate of robot navigation in crowds and reduce pedestrian intrusion, a dynamic reward function is designed to adjust the robot’s decision-making behavior in real time based on the crowd’s movement speed and density. Furthermore, differentiated reward and penalty mechanisms are implemented for social group and individual pedestrians respectively, ensuring the robot can intelligently optimize its navigation path. Experimental validation demonstrates that our method not only improves the success rate of robot navigation but also reduces the rate of pedestrian intrusion.

Keywords: Biomimetic robotics
Abstract: Current motor-driven flapping-wing micro aerial vehicles (FWMAVs) face challenges in achieving a weight comparable to that of real insects. This paper presents the design and manufacturing of a tailless four-winged FWMAV, which is driven by two motors through two flexure- hinge transmission mechanisms to actuate two pairs of wings. Our work enhances the assembly efficiency of FWMAVs through optimized in both design and manufacturing processes. Inspired by the flapping wing system of flying insects in nature, we adopt a compliant flexure-hinge transmission mechanism, enabling lightweight construction, improved assembly efficiency, and wing reversal capability. A rapid smart composite microstructure (R-SCM) process is proposed to accelerate the fabrication of compliant transmission mechanism, reducing 90% of the fabrication time compared to the standard SCM process. The final FWMAV prototype weighs 6.2 g, has a wingspan of 17 cm, and a single-sided flapping angle of 103°. In tethered flight tests using a pendulum setup, the prototype overcame gravity under a 3.4 V input voltage, achieving motion above the pendulum's horizontal plane within 0.247 s, validating the design and manufacturing approach.

Keywords: Other related topics
Abstract: Conventional spherical robots that employ rigid monolithic shells face constraints in environmental adaptability. This paper proposes a novel morphing spherical shell based on arc-shaped four-bar linkage mechanism, which enables the shell's diameter to change within the range of 40-50 cm, enhancing the robot's adaptability in complicated environments. The morphing spherical shell constitutes an independent system exhibiting high compatibility with various driving methods. It also provides adequate placement space for internal components with a high internal space ratio of 70% in its contracted state. And the soft outer covering film of the shell ensures the sealing and waterproof properties. This paper primarily encompasses the geometric model of the morphing spherical shell, mechanical structure of the morphing spherical shell system and prototype experimental validation through deformation testing.

Keywords: Prosthesis and exoskeleton robotics, Wearable robotics
Abstract: Replicating both power grasp and precision pinch capabilities in myoelectric prostheses remains challenging, as most current designs require multiple actuators per finger, which increases the size, weight, and control complexity. Hence, a compact finger mechanism that can switch seamlessly between these two fundamental grip modes only a single motor embedded in the palm is yet to be realized. This study proposes a novel finger mechanism that supports multiple finger motions with a minimal number of motors. Notably, the singular configuration of the linkage enables selective activation of the proximal interphalangeal joint using only one motor, thus achieving two distinct movements: “grasping” and “pinching.” We designed a compound four-bar linkage mechanism integrated with a linear actuator such that reversing the rotational direction of the motor yielded two different flexion trajectories. Furthermore, by placing small motors within the palm and utilizing 3D-printed resin components reinforced with metal screws, the dimensions of a human finger could be approached without compromising strength. A prototype demonstrated smooth and reliable switching between “grasp” and “pinch,” thereby indicating enhancing user comfort and reducing fatigue.

Keywords: Wearable robotics, Biomimetic robotics
Abstract: In recent years, research on wearable robots that assist or replace human physical functions by attaching devices to the human body has advanced, particularly in body augmentation that offers additional fingers or arms through robotics. Although these robots enhance physical capabilities, their grasping functions are often confined to specific tasks, hindering widespread adoption. Therefore, this study targeted artificial augmented fingers, a representative form of body augmentation robots, and sought to develop a system capable of performing multiple types of grasps without restriction to a single task. To achieve this, grasp types involving augmented fingers were systematically classified, and a mechanism for an artificial augmented finger system capable of executing these grasps was developed and evaluated.

Keywords: Biomimetic robotics, Other related topics
Abstract: Robotic assistance for dressing tasks, particularly manipulating deformable garments like trousers, remains a significant challenge. This paper presents the design and preliminary evaluation of a novel, bio-inspired soft gripper specifically for grasping trousers that are being worn. The gripper utilizes 3D-printed flexible fingers made from TPU 95a, which mimic human finger movements to create and secure fabric crease. It is powered by a single DC motor that operates through a worm gear and spur gear chain for energy efficiency. The design prioritizes safe human-robot interaction (HRI) by incorporating material compliance and current-based force control. Experimental results demonstrate a correlation between motor current threshold and grasping force, with the gripper achieving sufficient force (approx. 25 N at 400 mA) for potential trouser manipulation tasks. Overall, the proposed gripper contributes to bridging critical gaps in assistive dressing technology, paving the way for safer and more effective robotic dressing support.

Keywords: Prosthesis and exoskeleton robotics, Biomimetic robotics, Wearable robotics
Abstract: Individuals with severe injuries or congenital malformations resulting in limb amputation face significant challenges in daily activities. Recent advancements in upper-limb prosthetic design have accelerated due to the adoption of novel technologies, and research efforts have focused on developing prosthetic prototypes with the objective to replicate the human-like wrist movements. This paper proposes a differential parallel mechanism with three-degrees-of-freedom (3-DoF), designed to replicate natural wrist movements such as Flexion/Extension (F/E), Pronation/Supination (P/S), and Radial/Ulnar (R/U) deviation to achieve a human-like range of motion (ROM) within a compact structural layout. A comprehensive kinematic model of the system is presented, and the link dimensions are determined using a Modified Monte Carlo (MMC) optimization algorithm. Finally, a computer-aided design (CAD) model was developed to simulate the proposed configuration.

Keywords: Rehabilitation robotics, Medical surgical robotics, Wearable robotics
Abstract: Pressure ulcers represent a significant healthcare challenge with high prevalence rates and substantial caregiver burden. This paper introduces an autonomous robotic system for pressure ulcer care that operates through a novel hierarchical Vision-Language-Action (VLA) architecture. Unlike end-to-end approaches, our modular three-component design enhances interpretability and adaptability while reducing dependence on large domain-specific datasets. The system integrates an RGB-D camera for spatial perception, Qwen3 language model for instruction interpretation, and a ROS-controlled DENSO Cobotta robotic arm for precise interventions. Through in-context learning techniques, our approach effectively overcomes domain knowledge limitations in both vision and language components. Laboratory experiments with medical mannequins demonstrate the system achieves an 87.51% success rate across diverse tasks, including joint movement, object manipulation, and simulated wound irrigation. Comparative analysis reveals clear performance improvements through in-context learning, with success rates increasing from near-zero to over 85% for complex tasks. As the first implementation of a VLA-driven robotic system for wound care, this research addresses the critical gap between surgical robotics and nursing assistance, potentially enhancing treatment consistency while reducing healthcare worker workload in clinical settings.

Keywords: Medical surgical robotics
Abstract: In spine surgery, robotic system can achieve precise surgery planning and instrument positioning. However, millimeter-level displacements caused by respiratory motion may reduce guidewire placement accuracy. This study presents a spine surgical robot that is integrated with a real-time tracking system and a dynamic compensation system based on Kalman filter. The animal experiments (porcine model) demonstrated that the placement accuracy in the dynamic tracking group (83% level-A satisfaction) was improved compared with that in the no tracking group (42%). The result validated the system's clinical potential while maintaining similar operation time (32.5 min vs. 31.0 min) and overall workflow efficiency.

Keywords: Medical surgical robotics, Biomimetic robotics, Advanced materials and sensors for robotics
Abstract: This study presents the design and implementation of a force-feedback-enabled aspiration fluid control system for ophthalmic surgery. The system features a dual-pump collaborative architecture that combines the rapid negative pressure response of a venturi pump with the high-precision flow control of a peristaltic pump. An intelligent mode-switching mechanism enables smooth transitions between pressure and flow control modes to meet varying surgical requirements. A segmented force-feedback foot pedal is integrated to support multimodal interaction, enhancing the surgeon’s perception of system status and operational intuitiveness. Through closed-loop control algorithms and multi-sensor feedback, the system achieves precise regulation of aspiration pressure and flow. Experimental results demonstrate good performance in control accuracy, responsiveness, and human-machine interaction, indicating strong potential for application in next-generation intelligent minimally invasive surgical systems. However, further optimization is needed to improve long-term stability and reliability.

Keywords: Biomimetic robotics
Abstract: Heading perception is essential for navigation in both biological and artificial systems. However, many existing heading estimation methods rely on geometric feature matching or scanline-based techniques, which lack biological plausibility and often degrade under complex conditions. Inspired by the primate dorsal visual pathway and neural mechanisms of the head direction system, this paper proposes a bio-inspired heading perception method that integrates visual motion processing with a computational model of head direction cells. The visual subsystem extracts optic flow via motion-sensitive processing along the V1–MT pathway and estimates angular velocity by mimicking MST-area template matching. This motion cue is then integrated by a continuous attractor neural network, emulating head direction cell dynamics to maintain a stable heading representation. Experiments on controlled indoor and natural outdoor datasets, including iRat and KITTI, demonstrate that the proposed method achieves accurate heading estimation and outperforms existing biologically inspired and engineering-based baselines. The system offers both biological interpretability and practical effectiveness, contributing to bio-inspired navigation systems through a principled fusion of biological vision and computational neuroscience.

Keywords: Cyber-Physical bio-system, Cyborg intelligence
Abstract: Autonomous obstacle avoidance in dynamic and unknown environments is essential for mobile robots, forming a prerequisite for many high-level tasks. However, existing methods are often constrained by the limited computational and energy resources of mobile platforms, making it challenging to ensure both obstacle avoidance performance and operational efficiency. In contrast, the human brain demonstrates remarkable adaptability and exploration capabilities in unfamiliar environments while consuming an extremely small amount of energy. Inspired by this, we propose a novel brain-inspired obstacle avoidance method that integrates brain-like perception, data fusion, and control strategies, offering both computational and energy efficiency. We employ a bio-inspired event camera to efficiently capture motion cues of objects in the environment and extract salient features using a spiking-driven self-attention-based event variational autoencoder (SDSA-EVAE). To compensate for the lack of depth information, we incorporate LiDAR to complement visual input through a multimodal fusion algorithm inspired by neuronal diversity. This functionality is achieved using a reinforcement learning framework that leverages a hybrid architecture of artificial neural networks (ANNs) and spiking neural networks (SNNs). Comparative simulations with baseline models demonstrate the effectiveness of our approach, while physical experiments validate its practical applicability. Furthermore, deployment on a neuromorphic processor confirms the energy efficiency of the method.

Keywords: Biosensors and bioactuators, Medical surgical robotics
Abstract: 骨科手术机器上 0154;中，六维力 传感器对于高精 4230;力反馈至关重ව 1;。 但是，在骨骼穿 6879;等动态场景中ᦁ 2; 光纤布拉格光栅 （光纤光栅） 传感器经常受到 响应延迟和惯性 4341;起的滞后，这 力控制精度。为 0102;解决这个问题ᦁ 2;本研究 专注于光纤六维 1147;传感和 其在骨钻过程中 0340;动态误差补偿 最小二乘法支持 1521;量回归 （LSSVR）。比较 到具有外生输入 0340;传统自回归 （ARX） 模型，LSSVR 提供卓越的精度 1644; 健壮性，从 Historical 输出来实现实时 4917;偿。实验的 结果表明，这种 6041;法显著 提高所有 六维力矢量，过 0914;小于 大于 13%，稳定时间低于 15 毫秒，并

Keywords: Micro/Nano robotics, Advanced materials and sensors for robotics, Other related topics
Abstract: Abstract— Photopolymerization printing, known for its high precision and surface quality, is widely used across various fields. However, traditional methods face limitations in 4D printing, particularly in rapid prototyping and flexibility. This paper proposes a photopolymerization system integrating a dynamic spatial illuminator and an upright microscope, enabling one-step high aspect ratio molding and freeform printing. This approach enhances printing efficiency, flexibility, and allows rapid shape changes under external stimuli, providing a new path for 4D printing applications in smart materials, flexible sensors, and adaptive structures.

Keywords: Micro/Nano robotics, Biosensors and bioactuators, Other related topics
Abstract: This study introduces a light-induced AC electroosmotic micropump based on non-uniform illumination that drives liquid with patterned illumination instead of solid electrodes. Removing physical electrodes simplifies device fabrication and allows dynamic flow control through projected light pattern. Finite-element simulation results show that the flow velocity of micropump is strongly influenced by parameters such as AC voltage amplitude, frequency, solution conductivity, microchannel thickness, and illumination gradient. Specifically, directional net flow is achieved by breaking the symmetry of the electroosmotic vortex pattern through a linearly decreasing light intensity profile. The system operates under the low-frequency AC regime of optoelectronic tweezers, enabling fast switching between fluid pumping and particle dielectrophoresis by tuning the frequency. The proposed light-induced AC electroosmotic micropump shows great potential for integration into lab-on-a-chip devices, particularly where flexible, programmable, and contactless flow control is required. This work provides a foundation for the development of optically reconfigurable microfluidic systems.

Keywords: Micro/Nano robotics
Abstract: Heterogeneous microrobotic collectives formed by agents with different functions have attracted extensive attention. The sorting of different agents is a critical step in biomedical applications, such as biological analyses and diseases diagnostics. In this work, we propose a method to enable self-sorting of magnetic microparticles in heterogeneous collectives under oscillating magnetic fields. Magnetic hydrogel particles with different sizes are fabricated as agents. The effects of particle size on oscillating behaviors of particles are analyzed. The correlations between the locomotion velocity of collectives and the field strength are investigated, determining the range of field strengths for self-sorting. Moreover, experiments are conducted to verify the self-sorting method of heterogeneous particle collectives.

Keywords: Biomimetic robotics
Abstract: The oscillatory pitch motion at the leading edge of a millimeter-scale flexible sheet on the water surface can generate undulatory locomotion for swimming, similar to a honeybee vibrating its wings for propulsion. The influence of various parameters on such swimming strategy remains unexplored. This study uses magnetic milliswimmers to probe the propulsion mechanics and impact of different parameters. It is found that this undulatory propulsion is driven by capillary forces and added mass effects related to undulatory waves of the milliswimmers, along with radiation stress stemming from capillary waves at the interface. Modifying the parameters such as actuation frequency, pitch amplitude, bending stiffness, and hydrofoil length alters the body waveform, thus, affecting the propulsion speed and energy efficiency. Although undulatory motion is not a prerequisite for water surface propulsion, optimizing body stiffness to achieve a proper undulatory waveform is crucial for efficient swimming, balancing energy consumption, and speed. The study also reveals that the induced water flow is confined near the water surface, and the flow structures evolve with varying factors. These discoveries advance the understanding of undulatory water surface propulsion and have implications for the optimal design of small-scale swimming soft robots in the future.

Keywords: Biomimetic robotics, Prosthesis and exoskeleton robotics
Abstract: Bipedal robots, due to their unique structure, possess rapid and stable locomotion capabilities. Stable and efficient movement can significantly enhance task completion efficiency. However, these robots face challenges in flexibility and adaptability to complex environments due to limitations in their drive mechanisms. Traditional methods heavily rely on complex joint structures and electronic sensor feedback systems to ensure motion stability, which increases implementation complexity. Inspired by the human ability to walk steadily in complex environments, with the ankle joint playing a crucial role in maintaining stability, this study proposes a novel walking stability control method for bipedal robots. By designing a mechanically coupled perception-control structure based on mechanical logic computation, the robot simulates the real-time transmission of neural signals triggered by ground reaction force (GRF) in human reflex loops, thus simplifying traditional control methods. Through the collaboration of a flexible ankle joint and foot drive modules, the robot achieves real-time control of the ankle joint via the foot perception system. This paper presents a simplified motion control method for bipedal robots. The method, based on a mechanically coupled perception-control structure, can assist robots in achieving stable locomotion over complex terrains. This approach has potential applications in tasks requiring mobility over uneven terrains, such as search and rescue.

Keywords: Biomimetic robotics, Cyborg intelligence, Other related topics
Abstract: Inverse kinematics contributes significantly to the execution of trajectory planning and control operations in robotics. Traditional inverse kinematics algorithms, predomi- nantly utilized in serial robotic arm control, often lack precision and have limited applicability in parallel-legged robotic systems. To enhance the accuracy of solving inverse kinematics for planar bipedal robots, this paper introduces a design for a parallel tripedal robot and establishes its inverse kinematics model in a two-dimensional plane. Subsequently, a Sequential Quadratic Programming (SQP) algorithm integrated with a Sequential Habitat Genetic Algorithm is devised to address the inverse kinematics equations of the robot, constrained by a linear inverted pendulum model, enabling the tripedal robot to navigate various environments without the need for feedback. Compared to conventional SQP methods, this algorithm expands the solution scope and enhances computational accuracy. Additionally, this paper proposes a hierarchical NSGA-SQP algorithm, which, when compared to the traditional NSGA-SQP algorithm, further augments solution precision. Ultimately, to validate the accuracy and versatility of the proposed algorithm, simulation experiments were conducted, and the algorithm was deployed in real-world scenarios, confirming its feasibility and applicability.

Keywords: Biomimetic robotics
Abstract: Recently, the jumping ability of humanoid robots has made significant progress. However, in practical scenarios, robots always need to complete jumping movement carrying additional payloads. This puts more serious challenges on system: it demands higher torque output yet faces the performance limitations of the motor.To solve these challenges, this paper proposes a novel knee joint design that integrates a linear actuator and parallel springs. The linear actuator constructs a triangular mechanism with variable side length through ball screw. This realizes the variable reduction ratio characteristics of the joint. Compared with fixed reduction ratio, it provides higher torque output when taking off and better cushioning performance when landing.The introduced parallel spring mechanism works synergistically with the linear actuator, which effectively reduces the torque demand of motor, thereby enhancing the overall reliability of the system. Finally, we conducted jumping tests based on a single knee joint platform with a 10 kg load and verified the effectiveness of the design.

Keywords: Biomimetic robotics, Other related topics, Neuro-Control and communication
Abstract: Most humanoid robots walk with crouched legs, in an effort to avoid the singularity problem in articulated robot control. In comparison, a human walks with straight legs. We used a reinforcement learning approach to enable human- like straight-leg walking for humanoid robots. First, we train different models to control the humanoid robot to walk and stand using reinforcement learning. Then we train a switch- ing model to achieve smooth switching between walking and standing. After training in simulation, we deploy the model on a real humanoid robot. We bridge the gap between simulation and reality by combining visual characteristics for the robot state estimation. Thus, the humanoid robot can move stably on complex terrains with straightened legs, including walking in different directions, adapting to external disturbances, and standing still on discrete obstacles and slopes. Straight-leg locomotion is more human-like and can reduce 23% motor torques.

Keywords: Biomimetic robotics
Abstract: Lunar robots serve as the core equipment for surface exploration and operational tasks in lunar exploration projects, holding invaluable importance in deep space exploration and lunar scientific research. This study addresses the constraints of space and weight limitations for future lunar robots by proposing a humanoid robot composite joint based on single joint-single drive, incorporating both wheeled and legged motion modes. To achieve independence between the two motion modes, this paper employs a planetary gear system combined with electromagnetic brakes to design a switchable transmission linkage. Furthermore, a dynamic model of the planetary gear system and a mathematical model of motor drive are established, leading to a motion control framework that includes torque feedforward. The results of co-simulation experiments indicate maximum tracking errors of 0.95% for the wheeled mode and 1.3% for the legged mode, validating the effectiveness of the control strategy. Prototype experimental results demonstrate that the proposed wheel-legged composite joint achieves stable switching between motion modes and independent motion output, confirming the feasibility of the design. In the future, the design of this composite joint can provide a reference for developing lunar robot movement systems.

Keywords: Biomimetic robotics
Abstract: 月球低重力和复杂地形的影响，传统的轮腿机器人难以平衡地形适应性和运动效率，从而限制了它们在月球任务中的适用性。为了增强月球表面的多任务作能力，这论文提出了一种新型仿生轮腿移动机器人系统，采用四杆联动轮腿由双电机同轴嵌套驱动的结构 配置。这种设计确保了高地形适应性和运动效率。此外，建立了机器人的运动学和动力学模型，以及轮腿的安全作工作空间系统被分析。为了解决由复杂的月球地形引起的姿态扰动一种自平衡控制基于位置反馈和力反馈的策略开发了循环。仿真结果表

Keywords: Biosensors and bioactuators, Cyber-Physical bio-system
Abstract: Optogenetics can enable cells to achieve photosensitivity through heterologous expression of photosensitive proteins, and thus has been widely applied in bio-syncretic visual perception, control of bioactuators, and treatment of diseases. However, there is a lack of comprehensive characterization of optogenetically engineered cells currently with respect to their physical properties. In this work, we developed a platform by integrating the patch-clamp system and atomic force microscope (AFM) on the inverted fluorescence microscope, and characterized the differences between the optogenetically engineered cells and unmodified cells in terms of multidimensional information including fluorescence distribution, photoinduced whole-cell current, and Young's modulus of cells. Experiments show that the platform can characterize the expression level of transmembrane proteins on a single cell and dramatic changes in its photoresponsivity and Young's modulus. The comparisons of the results demonstrate that the optogenetically engineered cells are significantly different from those unmodified, and they are not just capacitated to respond to light irradiation, but also become softer due to the expression of transmembrane proteins. The relationship between the expression level of transmembrane proteins and the corresponding variations of photoresponsivity and Young’s modulus of a cell will be further studied. The platform to detect the multidimensional information of optogenetically engineered cells is of great significance for investigating cellular functionalization mechanisms and also paves an avenue for the application of optogenetics in bio-syncretic engineering.

Keywords: Medical surgical robotics, Micro/Nano robotics
Abstract: Microswarm 机器人技术已成为一项很有前途的技术用于生物医学应用，特别是在导航方面复杂的环境，其特征是复杂的人体内的生理结构。 然而 当前的控制方法主要局限于小规模系统，在以下方面面临着严峻的挑战保持稳定高效的制导远距离运动控制。本文提出了一种新型能够实现稳健的远程控制系统以及跨扩展作域的精确运动。 这 集成磁体系统产生旋转磁诱导螺旋轨迹迁移的磁场磁化粒子朝向指定目标。我们的混合动力控制框架战略性地结合了姿态约束与视觉反馈以保持群体完整性，同时实施基于高斯的密度评&

Keywords: Micro/Nano robotics, Medical surgical robotics
Abstract: Thrombosis causes millions of deaths globally. Existing treatments face limitations in efficiency and invasiveness. We propose an α-Oriented Magnetically Controlled Helical Microrobot (α-MCHM) using oblique-angle magnetization (α = 0°–75°) and multi-field synergistic control (rotational + gradient magnetic fields). Fabricated via DLP 3D printing (diameter: 1–5 mm), the α-MCHM achieves dual-field propulsion. Experiments show the 30°-MCHM attains a maximum speed of 25.86 mm/s (281% higher than conventional 0°-MCHM) and thrombus removal rate of 1.28 mm³/s (753% improvement). Navigation time in a vascular leg model is reduced by 37.27%. This work enables faster, safer vascular interventions.

Keywords: Medical surgical robotics
Abstract: The clinical flexible gastroscope and instrument exhibit limited degrees of freedom (DoFs) and operational dexterity, making mucosal dissection challenging during the complex flexible endoscopic procedure, particularly in endoscopic submucosal dissection (ESD). To address these issues, this work proposes a dual-armed flexible endoscopic robot that consists of a continuum endoscopic arm with enhanced triangulation capabilities and two miniature flexible instrument arms with improved dexterity. The endoscopic arm has an outer diameter of 16mm with 3-DoFs. The instrument arm features only 2.6mm with 5-DoFs. It enables 360° unrestricted distal independent rotation of the forceps, unlike the entire rotation of typical flexible instruments. Ex-vivo experiments on a porcine stomach have been performed to verify the robot’s effectiveness in ESD procedures.

Keywords: Other related topics
Abstract: This paper presents an interactive vision-based system that can guide an underwater robotic arm by performing click operations on a calibration board within the camera’s field of view. Users can directly select target points on the two-dimensional image interface, and then the system calculates the corresponding three-dimensional coordinates of the target point in the system’s base coordinate system through a pre-calibrated conversion process. Different from traditional manual positioning methods, our method does not require operators to explicitly input coordinates or use joysticks for control, thus simplifying the operation burden. Tank experiments have verified the feasibility of the system. Within the short-distance operation range, the average error of single-point repeated positioning for each corner point is between 0.35 - 0.45 mm with a small standard deviation, and the average error of path repeated positioning is 0.1 - 0.2 mm, indicating that the system has accurate and stable positioning. The proposed scheme is suitable for common hand-eye robotic arm setups and only requires a calibration board, making it highly practical in scenarios such as underwater maintenance or sampling.

Keywords: Cyborg intelligence, Neural-machine interface
Abstract: Dogs play a vital role in fields such as border control, drug detection, disaster rescue, and disease detection, owing to their exceptional olfactory system. However, traditional training methods are reliant on behavioral observation, which is inefficient and demands high skill from trainers. This study proposes a neural signal analysis approach based on brain-computer interface (BCI) technology to enhance training efficiency by analyzing the neural mechanisms underlying dogs' sniffing behavior. The Phase Locking Value (PLV) is used to quantify the phase characteristics of Electrocorticography (ECoG) in dogs under free-moving conditions. A scent classification model based on Bayesian optimization and AdaBoost ensemble learning is constructed. Experimental data demonstrate that this method effectively mitigates motion artifact interference and successfully classifies neural signals corresponding to two different scents. This research provides a theoretical foundation and technical approach for the development of a novel canine olfactory BCI system, offering significant practical value in reducing training costs for working dogs and expanding the application scenarios of olfactory detection.

Keywords: Other related topics
Abstract: 传统的单目视觉里程计在很大程度上依赖于 特征点匹配的质量和密度 连续帧，通过 几何方法。这种依赖关系呈现了这样的系统 在高速场景或缺乏的环境中脆弱 足够的纹理，特征提取经常失败 和潜在客户 运动估计精度差。克服这些 漏洞，本文提出端到端深度 学习 使用增强图像进行相机姿态估计的框架 数据。我们的架构首先采用基于 CNN 的功能 提取器，通过注意力机制增强，以 提高 它能够从图像中辨别显着信息 对。这些 然后将特征输入Bi-LSTM网络，以有效地 对相机运动的时间动态进行建模。决赛 stage 使用完全连接的

Keywords: Other related topics, Cyber-Physical bio-system, Biosensors and bioactuators
Abstract: In this work, we address the limitations of static facial representations in multimodal affective computing by introducing a temporal sequence of face patches extracted from video frames. These visual features are fused with EEG-derived differential entropy features using a token-level fusion framework, enabling dynamic modeling of cross-modal temporal cues. Our model demonstrates improved fine-grained emotion recognition across valence, arousal, dominance, and liking dimensions.

Keywords: Other related topics, Rehabilitation robotics, Wearable robotics
Abstract: Freezing of gait (FoG) in Parkinson’s disease (PD) significantly impacts patients’ mobility and daily lives. Video-recorded walking tasks have become the gold standard for assessing FoG. However, existing video-based methods encounter challenges such as the need for multiple cameras, susceptibility to occlusion, and inability to predict FoG events. We propose a novel framework that uses inertial measurement unit (IMU) data as a guiding modality during training to distill predictive knowledge into a video-only model for inference. Our hierarchical contrastive learning strategy effectively transfers high-frequency motion patterns from IMU signals to visual skeleton (SKE) representations. This achievement lies in utilizing 1) intra-modal contrastive learning to extract stride and angular velocity patterns with temporal consistency constraints and 2) cross-modal alignment to transfer these discriminative IMU-based features to skeleton encoders and to alleviate visual occlusion issues through adaptive noise filtering. Validated using expert-annotated videos from 35 PD patients, our hierarchical contrastive learning algorithm achieved an IMU level accuracy of 79.6%. This method utilizes privacy-preserving leg keypoints for FoG detection, enabling clinical assessment and long-term home monitoring. The code is available at GitHub.

Keywords: Other related topics, Biomimetic robotics, Advanced materials and sensors for robotics
Abstract: Automatic replacement of disc cutter is urgent for improving the TBM efficiency and safety in underground construction projects, while it is still challenging to precisely grasp disc cutter remotely in the harsh and complex on-site environment. In this paper, we proposed an adequate, robust and easy-to-employ method to position the end-effector of the robot to improve the disc cutter replacement task. A progressive framework was proposed to control the robot. A tactile sensing system was proposed for locating the disc cutter within the narrow cutter house, which mainly relies on an array of digital I/O sensors to percept the relative spatial position between the cutter and the gripper. A novel velocity vector generation method was proposed to control the gripper to adaptively approaching a grasped status. With the tactile sensing system, our method enables real-time control with sufficient accuracy. Experimental results demonstrate the successful gripping of disc cutters in laboratory tests. The proposed method is robust to harsh environment conditions underground and can be applied to similar robotic tasks.

Keywords: Biomimetic robotics, Wearable robotics, Rehabilitation robotics
Abstract: In working environments where heavy loads need to be carried or frequent bending is required, workers may often suffer from lumbar strains and lumbar disc herniation. To reduce the burden and prevent lumbar injuries, back-supporting exoskeletons can be worn to assist with workers, such as carrying goods. Inspired by human spinal joints, a multi-joint cable-driven spinal exoskeleton (MJSE) is designed in this paper. It is composed of multiple vertebral units, and can adapt to various human body movements with a high degree of flexibility, including flexion, extension, and side swing. First, the kinematical model of the MJSE is established, and the relationship between the angles of forward flexion and side swing and the steel cable's elongation is analyzed. Second, a stiffness model is established. Finally, the stiffness test platform and the MJSE's prototype are manufactured. Then the stiffness experiments and wearable experiments are carried out. The experimental results show that the exoskeleton has good flexibility and stability, and its motion range is within the motion range of human spine, so that it can match the movement of human spine well and won't hurt the wearer.

Keywords: Rehabilitation robotics, Cyber-Physical bio-system, Neuro-Control and communication
Abstract: To meet the increasing demand for safe and effective aerobic exercise among older adults, we developed a walking training device that incorporates a photoplethysmography (PPG) sensor and a magnetorheological (MR) fluid brake. This device allows real-time pulse rate monitoring and adjusts the walking resistance automatically based on the user’s exercise intensity derived from pulse rate. The PPG sensor, embedded in the handle of the device, demonstrated accuracy comparable to commercial PPG and ECG systems, even during movement. To improve signal reliability during walking, we applied a processing method that utilizes the second harmonic component of the heart wave, which helped reduce motion-induced artifacts. We also designed a feedback control algorithm that maintains exercise intensity within a target aerobic range (40–50 % of Karvonen intensity) by adjusting the brake resistance in real time. In trials with young and adult participants, the system achieved the intended aerobic intensity without causing overexertion. Feedback control experiments confirmed that the device could guide users toward and maintain aerobic exercise levels effectively. These findings suggest that the system can serve as a personalized training tool to support endurance enhancement and help prevent frailty in the elderly through safe, pulse rate-guided walking exercise.

Keywords: Wearable robotics, Prosthesis and exoskeleton robotics, Cyborg intelligence
Abstract: Reduced gravity simulation systems are important for astronaut training because microgravity can have some adverse effects on the human body. However, there is a lack of a small device that can simulate a reduced gravity environment and fewer analyses of the effects of microgravity on the human body. For this reason, we designed a reduced gravity simulation system based on a vertical body weight support system (BWS) and an exoskeleton. The BWS simulates the effects of reduced weight of the whole body by suspending a certain weight. The exoskeleton monitors the angle of the wearer's lower limb joints through a inertial measurement unit net on the body to map the corresponding joint moments, and ultimately realizes the reduced gravity simulation of hip joint. Based on the system described above, we wish to explore the effects of microgravity on human biomechanics while trying to ensure the simulations are as accurate as possible. One subject was recruited to participate in a series of experiment. Walking experiment was conducted at two speeds under three wearing conditions: without the BWS and exoskeleton, with the BWS and with the BWS and exoskeleton. Biomechanical data, including joint angles, joint moments, and muscle activations, were recorded by a motion capture system and surface EMG during the experiments. Based on these measurements, the effects of reduced gravity simulation environment on the walking patterns were mainly in the joint angles, joint moments, muscle activation, and control strategies of the human body. It is worth noting that the activation intensity of some muscles increased by different magnitudes and shifted in phase under the simulation condition compared to the normal. These results were unexpected for us and promising to serve as an inspiratio

Keywords: Prosthesis and exoskeleton robotics, Wearable robotics, Rehabilitation robotics
Abstract: This paper presents an approach to the control of a 2-DoF (degree-of-freedom) augmented upper-limb robotic exoskeleton using adaptive admittance control based Fuzzy Inference System (FIS) combined with an adaptive Model Reference Adaptive Control (MARC). The exoskeleton is designed to augment the users' arms by providing reinforcement force in real-time based on the users' movements and intentions. The adaptive admittance control method enables the system to respond to varying users' dynamics, adjusting the interaction impedance to ensure smooth and safe interaction between the exoskeleton and the user. To enhance system stability and performance, the MARC controller and parameters adaptation method are implemented, ensuring that the closed-loop system maintains global asymptotic stability while adapting to changes in both the carrying payload and user dynamics. The adaptive nature of the control system allows for real-time adjustments to both the admittance control and controller gains, ensuring that the system remains robust to parametric uncertainties. Simulation results demonstrate the effectiveness of the proposed control strategy in terms of improved adaptability of controller to the carrying-payload uncertainty and user comfort during dynamic interactions. The proposed methodology paves the way for more adaptable and user-centric robotic exoskeleton devices, facilitating natural movement of the users.

Keywords: Rehabilitation robotics, Cyber-Physical bio-system, Cyborg intelligence
Abstract: Gait phase prediction is a key technology for gait rehabilitation robot control and gait assessment. At present, it has been proven that the LSTM model has better performance than traditional machine learning models in the field of gait phase prediction and recognition. This study proposes a model pack to predict gait phases with centre of mass horizontal position as input. We also explore the influence of sliding window sizes on prediction performance and confirm that the LSTM+Autoencoding-LSTM model combination with the sliding window size of 40 has the best gait phase prediction performance. The average Precision, Recall and F-score of this model pack are 0.947, 0.941 and 0.943 respectively. This study proves that the horizontal position of CoM can be used for gait phase prediction, and the performance of proposed model pack is comparable to related studies while the demand of sensors is reduced.

Keywords: Rehabilitation robotics, Cyborg intelligence, Wearable robotics
Abstract: People with disabilities often face challenges in their daily lives. In recent years, a growing body of research has focused on developing robotic systems to assist and support these individuals. Robotic assistance presents a promising solution to enhance their independence and improve their overall quality of life. In this research, we present the design of a robotic system that uses a wearable, real-time 2D/3D gaze-based interface to assist users with disabilities in locomotion, manipulation, and sensing. Our method involves the design of a gaze-based navigation interface for control of a wheelchair, and a gaze-based robotic arm interface for manipulation. In addition, a robotic arm object identification component based on force sensing is developed to extend the user's sensing capability for invisiable and unreachable objects. In this paper, the object identification component is limited to liquid classification. Experimental results demonstrate that the proposed system can effectively assist users in wheelchair navigation, object grasping and identification by a robotic arm, using only gaze-based control interfaces.

Keywords: Wearable robotics, Prosthesis and exoskeleton robotics, Rehabilitation robotics
Abstract: Many dexterous hands achieve coupled-adaptive grasping via complex mechanical structures. This paper presents a coupled-adaptive dexterous hand with a simple structure. It is composed of two identical robotic fingers based on a seven-bar mechanism and a palm. The motion of the robotic finger is analyzed. It has two modes during enveloping grasp: coupled and adaptive. The coupled relationship between finger joints is derived. A kinematic model of the robotic finger is established through the standard Denavit-Hartenberg method. An experimental platform was constructed and relevant experiments were carried out. The experimental results demonstrate the effectiveness of the designed dexterous hand in practical applications.

Keywords: Biomimetic robotics, Neuro-Control and communication, Other related topics
Abstract: Dexterous robotic manipulation remains a longstanding challenge in robotics due to the high dimensionality of control spaces and the semantic complexity of object interaction. In this paper, we propose an object affordanceguided reinforcement learning (RL) framework that enables a multi-fingered robotic hand to learn human-like manipulation strategies more efficiently. By leveraging object affordance maps, our approach generates semantically meaningful grasp pose candidates that serve as both policy constraints and priors during training. We introduce a voting-based grasp classification mechanism to ensure functional alignment between grasp configurations and object affordance regions. Furthermore, we incorporate these constraints into a generalizable RL pipeline and design a reward function that unifies affordance-awareness with task-specific objectives. Experimental results across three manipulation tasks—cube grasping, jug grasping and lifting, and hammer use—demonstrate that our affordance-guided approach improves task success rates by an average of 15.4% compared to baselines. These findings highlight the critical role of object affordance priors in enhancing sample efficiency and learning generalizable, semantically grounded manipulation policies. For more details, please visit our project website: https://sites.google.com/view/dora-manip.

Keywords: Advanced materials and sensors for robotics, Biosensors and bioactuators, Other related topics
Abstract: 3D shape reconstruction and force estimation functions of vision-based tactile sensors are utilized to facilitate robotic dexterous manipulation and force feedback. However, conventional force prediction models depend on the estimated depth while ignoring the influence of the variable contact area during robotic manipulation. In this paper, we propose a normal force estimation model for visuotactile sensors, which especially considers the coupled nonlinear effects of both indentation depth and contact area. In particular, a novel parameter identification approach is introduced, which integrates 3D reconstruction, height map, and nonlinear least squares. The experimental results and analysis using Gelsight Mini sensors demonstrate superior performance in predicting normal forces. Furthermore, the proposed method can provide effective normal force feedback in robotic grasping applications.

Keywords: Biomimetic robotics, Biosensors and bioactuators
Abstract: For biomedical and rescue applications, this paper presents the Hyd-U hand, a compact, hydraulically actuated underactuated robotic hand designed to enhance payload capability in dexterous manipulation. With a payload-to-weight ratio of 22.14 (16,kg payload, 722.6,g self-weight), it is ideal for applications requiring high force and low weight, such as field rescue and mobile robotics. The system integrates three key innovations: • Electro-hydraulic underactuation: Combines a singleacting hydraulic cylinder with truss-pulley transmission and cable-driven fingers to enable sensor-free adaptive grasping • Hybrid PID-NNIM control: Implements neural network inverse modeling (NNIM) with PWM current sensing to achieve ±0.05 mm displacement precision • Embedded high-force manipulation: Miniaturized servo driver enables responsive operation in unstructured environments Experimental results demonstrate the Hyd-U hand’s stable grasping of objects with varying shapes and weights, showcasing its repeatability and responsiveness in dynamic, unstructured environments.

Keywords: Biomimetic robotics, Medical surgical robotics
Abstract: To address the limitations of conventional teleoperation systems in modern application scenarios, this paper proposes a vision-based teleoperation system for robotic dexterous hands based on motion retargeting. The system primarily consists of three modules: hand keypoint detection, hand pose retargeting, and robot motion generation.

Keywords: Cyber-Physical bio-system, Other related topics, Cyborg intelligence
Abstract: Human-robot collaborative assembly (HRCA) has gained the attention of researchers due to the potential to combine the operator’s flexibility with the robot’s precision. In HRCA, implementing robotic dynamic collision avoidance can effectively minimize injuries and equipment damage. However, it is challenging to acquire the optimal collision avoidance motion path for the robot in real time, considering the safety of the moving operator and the efficiency of assembly. In this paper, a dynamic collision avoidance method considering real-time performance and the shortest path length for HRCA is proposed. Attractive, repulsive and tangential forces are generated aimed at enabling the robot to reach the target position while avoiding the collision with the operator. To improve the success rate of collision avoidance and reduce the path length, the force weight is introduced to generate the optimal guidance force. The genetic algorithm is applied to obtain the optimal solution of the force weight with the self-built collision avoidance training scenarios. Experiments are conducted to validate the superiority of the proposed method and its applications in the real-world HRCA scenario.